A deeper game.
A clearer test of understanding.

# Locked independent student-perspective review A

**Status:** Locked findings, 29 October 2026. This is an AI student-perspective review, not research with human participants.

## Method and limits

I reviewed the supplied `The_Power_Circuit.html`, `Student_Guide.pdf` and `Student_Worksheet.pdf`. I used `power-ui.js` and `cases.json` only to reconstruct player-facing screens, choices, feedback, disclosures and progression. I also extracted the player-content data embedded in the HTML: its displayed final tradeoffs differ from the fallback tradeoff text in the UI file, so the findings below use the embedded, displayed version.

The walkthrough reconstructed the sequence **home → tutorial → parliamentary formation → ordinary bill → formal challenge → replacement → explanation → country incidents and twists → final comparison → three tradeoffs → research desk → passport**. I followed one parliamentary circuit in detail, inspected the other two circuit descriptions for the comparison, and reviewed both moves for all nine country cases. This was a source-based walkthrough. A local Playwright attempt had no installed browser executable; the cloud browser could not reach the local page. I did not complete a rendered-browser usability test, observe scrolling or layout, measure task time, or test accessibility. The parent reviewer is handling live-browser testing.

**Blindness caveat:** My first unfiltered read of the self-contained HTML incidentally displayed the beginning of its bundled engine. I did not analyze or use that engine text and switched to player-copy-only extraction. I did not read previous reviews, the original repository, tests, answer-engine files, other agents' findings or developer history. This incidental exposure should remain recorded rather than being treated as perfect blindness.

The learner model was a first-time introductory student without assumed meanings for executive, legislature, mandate, confidence, investiture or comparative research methods. The questions below are analytic probes of what the materials make explainable. There are no human responses, scores, effect sizes, completion times or engagement ratings. My own political knowledge is not evidence of learning caused by this game. I did not check constitutional validity or implementation correctness.

## Overall judgment

The game has a coherent and worthwhile learning target: distinguish **getting office, remaining in office and passing policy**, then recognize that country-specific procedures can change a general pattern. Its strongest material is the contrast between a defeated bill and a defeated cabinet, followed by cases that challenge title-based classification and simplistic rules about elections.

The main risk for a novice is an uneven staircase. The game starts with unfamiliar institutional vocabulary, provides a highly guided recognition exercise, then expects an independent conditional comparison and a comparative research proposal. Students have good opportunities to encounter the intended ideas, but the supplied flow does not itself establish that they can explain those ideas without recognizing wording from options and feedback. Preserve the substantive complexity and all nine cases; add explanations, procedural tools and opportunities to show reasoning between those steps.

## Walkthrough evidence and likely learner experience

| Stage followed | Displayed evidence | What it supports; where a first-time learner may hesitate |
|---|---|---|
| Home and field guide | “Same political crisis. Three sets of rules. Who stays in office?” The guide opens by distinguishing who gets office, keeps office and passes policy. Modes, route length, pause, repair and save arrangements are explicit. | The central question is memorable. “Executive,” “assembly,” “cabinet,” “mandate” and “confidence” are already doing explanatory work before a novice has a plain-language cast of characters. |
| Tutorial, first action | Voters “choose representatives”; assembly has 100 seats. The instruction is “Tap the mandate token to connect the assembly to the executive.” | The visible relationship is useful. The action demonstrates the connection rather than asking the learner to decide who selects whom. “Mandate token” can be read as a game object without learning what mandate means. |
| Tutorial, second action | “The cabinet needs assembly confidence. A formal no-confidence motion succeeds.” The next action is “Cut the confidence wire.” The reveal says “Cabinet out. Assembly still here,” with occupied elected seats and a possible replacement. | Excellent concrete separation of cabinet and legislature. A learner can follow both actions by instruction; neither action reveals whether they can explain why one institution remains. |
| Parliamentary origin and bargain | The correct origin option describes voters → assembly → PM/cabinet. Civic 42 plus Regions 10 is one possible 52-seat agreement within three of six commitment slots. A PM selector explicitly says being largest is not an automatic entitlement. | Arithmetic, coalition variety and a leader from supporting parties make the relationship tangible. The negotiation modal gives actual promises, such as an audit or local rebuilding fund. A novice still needs to distinguish promises, cabinet membership and confidence support. |
| Bill and challenge | The ordinary bill is fixed at 48 yes/52 no, with “no confidence condition attached.” The later valid no-confidence vote is fixed at 58/42. The copy repeatedly explains that members may realign and these are matched comparison votes. | This supports a controlled institutional contrast. A learner who just negotiated 52 or 59 seats may reasonably wonder what their deal changed. The disclosure is clear, but it does not by itself make the bargain emotionally consequential. |
| Replacement and explanation | “Same seats. New government.” A replacement can be built from the existing assembly. The prompt asks why the policy defeat and formal challenge had different effects. Learning-mode explanation is optional; class mode uses a word minimum and says quality is for the instructor to read. | The desired explanation is well specified. Completion and repaired choices should not be treated as evidence of producing it independently. The worksheet offers space but no example of a sufficiently causal explanation. |
| Country sequence | “Same family. Different fine print.” Each of nine dossiers has an incident and a twist. The rule drawer starts closed before an initial decision; rules are explicitly permitted and free. Each screen offers an office/procedure reasoning clue. | Both moves matter, and consultation is normalized. Reading a legal target, threshold, condition and clock is a new skill set. The strongest needed scaffold is a consistent way to extract these facts, not fewer countries. |
| Final comparison | The unfamiliar republic has an elected fixed-term President and a confidence-dependent PM/cabinet; options ask which relationship fits. The written comparison asks origin/survival and one country complication. | This is a reasonable near-transfer check. It strongly resembles the earlier semi-presidential pattern, so success cannot establish broader transfer across altered titles, denominators, timing or exceptional procedures. |
| Tradeoffs and research | Actual displayed dilemmas concern coalition flood funding, presidential bargaining and opposing President/PM mandates. Benefits, costs and conditions follow a choice. The research desk asks a hypothesis, test, rival and challenging observation. | Conditional reasoning and avoidance of a “best system” ranking are strong. The move from interpreting a rule to designing a comparative test still needs teaching in its own right. |

## Seven issues, ordered by impact

### 1. The entry point assumes the vocabulary needed to understand the entry point

**Evidence:** The tutorial uses assembly, executive, PM, cabinet, mandate and confidence. The guide uses executive and constitutions immediately. The notebook does contain an extensive and useful glossary, including government, confidence, investiture, coalition and dissolution. However, the warm-up does not introduce that notebook as a support, and it does not give a short cast of roles. Definitions such as “government originates through the legislature” still assume that legislature and government are distinct familiar objects.

**Likely effect:** A novice can learn that a red cabinet box disappears while a blue assembly box remains without connecting those boxes to elected representatives, an executive team and lawmaking. Everyday use of “government” to mean all public institutions is a particularly important source of confusion.

**Scaffold:** Before the token action, give a four-role key in everyday language: voters; elected assembly/legislature; head of government; cabinet. Explain that “government” here means the executive team. Define confidence at its first use as the assembly support required for that team to keep office. Define mandate beside the token. Keep the full notebook, but link directly from unfamiliar words to the relevant short entry. Introduce a separately elected President only when that extra office appears. The general model and the national exceptions should remain intact.

### 2. Guided recognition and repair can outpace independent explanation

**Evidence:** Both tutorial actions tell the learner which connection to make or cut. Circuit predictions offer a correct relational explanation alongside very broad alternatives, such as all offices or assembly seats becoming vacant. A repaired answer unlocks the next step. Learning-mode circuit explanations and country passport notes are optional. Class-mode word checks are explicitly not conceptual grading. The final unfamiliar republic closely matches the taught semi-presidential description.

**Likely effect:** A learner may complete a stamp or circuit through elimination and feedback without being able to reconstruct its mechanism. This is a risk in the design, not a claim that students will in fact do so. Free hints and repairs are valuable and should be preserved.

**Scaffold:** Ask for a brief prediction before showing choices at one key transition: “The bill failed. Who, if anyone, must leave, and which rule makes that happen?” After feedback, ask for one correction to that explanation. Offer a light structure—**office → procedure → rule → consequence**—rather than just a word minimum. Use a small teacher-facing criterion: correct target, correct rule and correct account of what remains. At the end, add one novel case with altered titles or a changed vote rule and ask for a reason before revealing the family. Allow notebook consultation; the goal is reasoning from a rule, not memorizing its wording.

### 3. The relationship between bargaining choices and fixed shocks needs a clearer payoff

**Evidence:** Home promises “Build alliances. Take the vote.” Formation requires negotiated commitments and constrained slots. The copy clearly discloses that later 48–52 and 58–42 roll-calls are fixed and members may realign. The guide separates the optional Midnight Session, where decisions determine votes, from the matched exercise.

**Likely effect:** A novice can understand the disclaimer yet feel that the agreement they constructed was discarded. They may also conflate initial confidence support with guaranteed votes on every bill. Repeating the fixed bill/challenge sequence across three constitutions can become a button-following exercise once the pattern is recognized.

**Scaffold:** At the first shock, show a compact before/now ledger: **signed agreement: 52 seats; this bill: 48 yes; this confidence motion: 58 for removal**. Add one plain sentence about what changed politically and one about what is held fixed for comparison. End the circuit by asking “What did your bargain determine? What did the matched exercise hold constant?” After the first circuit, ask for the predicted *difference* under the next constitution before repeating the sequence. Preserve both fixed shocks and the optional strategy lab; make their learning purposes easier to see.

### 4. Country procedures need a reusable reading method, with extra support for denominators and clocks

**Evidence:** Country rules introduce local institutional names, Article 67 versus 68, section 102(1) versus 102(2), ordinary criticism versus constitutional accusation, Article 49(3), an all-member denominator and overlapping ten-day rules. The common clue—identify office, procedure and what remains—is helpful but incomplete for these procedural tasks. Taiwan's main rule drawer says dissolution can occur “within the constitutional period”; the twist's essential starting point for the dissolution clock becomes explicit in the optional twist-help feedback, which also states the answer.

**Likely effect:** Students can mistake unfamiliar national labels for unfamiliar institutional families, count the wrong vote, or infer the clock from the nearest event. Some can succeed by recognizing the help's exact answer without learning to extract a rule. This is not a reason to remove Taiwan, Germany or any other case.

**Scaffold:** Use the same five-field scratchpad on every dossier: **office targeted; action/procedure; required support; timing/conditions; what changes and remains**. Add a small national cast key, translating local names into roles. Demonstrate one all-member versus votes-cast calculation in a separate example. For Germany, display two procedure rows, so students can locate the route actually named. For France, label what the 48 votes are *for*: censure, not the bill. For Taiwan, provide a neutral rule statement that names the clock's starting event before offering answer-specific feedback, and a blank timeline on which to place vote, resignation and proposed dissolution. This supports difficult reasoning without replacing it.

### 5. The nine dossiers need more cumulative synthesis than stamps and one optional note provide

**Evidence:** Every country has two substantial moves and a distinct focus. Stamps record completion. The paper worksheet asks for **one** case that complicates a general rule; digital passport notes are optional. The final comparison also requires one country complication. These are useful prompts, but they do not organize the relationships among eighteen country moves.

**Likely effect:** A learner may retain nine local conclusions or a few article numbers while missing why the cases belong together. The opening model's memorable “no automatic election” rule may also dominate the conditional dissolution routes encountered later unless the difference is explicitly assembled.

**Scaffold:** Add a compact, reusable comparison record with one row for every country: source of office; ordinary policy defeat; specified challenge; conditional next step; one complication. Give two short cross-case pauses: UK versus Brazil on kinds of coalition support; Germany versus Taiwan on conditional dissolution; France's confidence-linked bill versus the ordinary matched bill; South Africa versus the US on why “President” does not settle survival. Ask students to name which fact changes the result. Keep all nine countries and both moves in each.

### 6. The research desk introduces a second discipline without enough bridge practice

**Evidence:** The desk asks for explanatory and outcome variables, mechanism, cases/periods, observable indicators, comparable conditions, competing processes and challenging evidence. The guide provides the X/Y template and a strong warning that a research idea is not a finding. The optional example invokes party fragmentation, constructive no confidence and majority status. Class mode specifies minimum lengths, but the preceding play primarily teaches legal mechanisms rather than constructing comparative research designs.

**Likely effect:** A student may write a broad claim such as “parliamentary systems are more stable” or copy the starting point without identifying what stability means or what observations could test the claim. This is an anticipated need for research-method teaching, not evidence that the supplied examples make false claims.

**Scaffold:** Start with one saved observation from play. Then walk through: **choose cabinet replacement or policy delay; name what could vary; give the mechanism; choose one observable measure; identify a rival; state a result inconsistent with the claim**. Explain “mechanism,” “indicator” and “hold comparable” beside the fields. Supply one completed miniature design and one intentionally weak claim to revise. Keep the four-field task and its explicit distinction between a proposition and a causal finding.

### 7. The long route needs explicit pauses and more concrete reasons to care about the storm

**Evidence:** The guide estimates 65–85 minutes for core and 80–100 for extended, plus an optional 8–12-minute strategy session, and allows two core sessions. Those are appropriately labeled design estimates. The route has stage navigation and country-stamp counts. Flood funding, audits, transport, housing and employment appear in promises and final tradeoffs, but most matched-circuit consequences are institutional status changes rather than visible effects on residents. Team roles are named, but no specific role turns are prompted.

**Likely effect:** A novice may experience considerable reading and repeated classification before seeing why the distinction between staying in office and governing matters. Students starting the final stations may also discover a larger writing task than the “passport almost complete” language suggests.

**Scaffold:** Show the whole route and suggested stopping points up front: models, country families, comparison/tradeoffs, research. Mark a natural pause after each family and display what will be written at the final stations. Add brief resident-facing consequences to the already conditional tradeoffs—for example, delayed repairs versus a smaller audited funding package—without assigning a universally best choice. Give the analyst and skeptical reporter one prompt at each pause. Validate duration and fatigue with actual students before presenting estimates as tested timing.

## Preserve all nine cases: contribution and targeted support

| Country | Learning contribution to preserve | Small scaffold that preserves its rigor |
|---|---|---|
| United Kingdom | Largest party is not automatic government; a minority cabinet can have outside confidence support while losing an ordinary bill. | Distinguish seats in the cabinet, promises of confidence support and support for this bill. Explain “supply” and “minority government” at first use. |
| Germany | Opposition to the incumbent is different from agreement on a successor; Article 68 is a different route. | Local cast key for Bundestag/Chancellor/Federal President; side-by-side procedure rows with action, threshold and possible next step. |
| South Africa | The title President does not imply a separate presidential survival relationship; the target of the motion changes the result. | Show Assembly → President and identify which offices each section targets; translate “reconstitute” into ordinary language. |
| United States | A blocked bill leaves office intact; House accusation is distinct from Senate conviction. | Two-chamber cast key and a staged accusation/trial/conviction sequence, clearly separate from the bill vote. |
| Brazil | Presidents can need coalitions for legislation; ministerial appointment power does not command party votes. | Draw separate relationships for President → minister and legislators → bill; compare with UK confidence support. |
| Chile | A coordinating minister is not automatically a PM; exceptional ministerial removal is not routine cabinet confidence. | Identify the coordinating role, then contrast ordinary appointment/removal with the specified exceptional accusation procedure. |
| France | Opposing mandates can coexist; linking confidence to a bill changes the ordinary-bill pattern. | Two-office map and a “what is this vote about?” label comparing 48 votes for the matched bill with 48 votes for censure here. |
| Portugal | A presidential mandate has conditions; all-member and votes-cast majorities differ. | Condition checklist for the first move; a 100-member denominator with abstentions visibly retained for the second. |
| Taiwan | No investiture vote does not eliminate later legislative responsibility; different deadlines can share a starting event. | Plain-language premier/Legislative Yuan key; separately labeled resignation and dissolution windows on one timeline. |

## Before, after and novel-transfer diagnostic questions

These questions were used as analytic probes, not administered to people. No answer or score is attributed to an actual or simulated student. A student's explanation, including use of the rules, would be the evidence needed in later human testing.

| Concept | Before play: elicit the starting model | After the relevant play: test its stated target | Novel transfer: test whether the relation survives changed details |
|---|---|---|---|
| Office versus legislation | “Representatives reject a repair bill. Does anyone necessarily lose office? What information would you need?” | “Why did the 48–52 bill defeat leave the executive in office when the 58–42 formal motion removed a parliamentary cabinet?” | “A fictional rule says the cabinet resigns only after a specified confidence vote. A tax bill loses 60–40, with no such vote attached. Identify what changes and what does not.” |
| Origin versus survival | “Who do you think chooses a President? Does the title tell you who can remove that person?” | “Why does South Africa's President have a different survival relationship from the separately elected President in the matched presidential model?” | “A country calls its Assembly-elected head of government ‘President.’ A valid Assembly no-confidence vote requires that officeholder's resignation. Explain the origin and survival relationships without using the title to decide.” |
| Coalition versus votes | “Does a signed coalition promise mean every member must support every bill?” | “How can the UK outside-support party oppose transport policy while the minority cabinet remains? How is Brazil's legislative coalition different?” | “A cabinet has 44 seats and outside support from 13 legislators only for confidence. They reject its housing bill but reaffirm confidence. Explain whether the bill and cabinet survive separately.” |
| Target and exceptional procedure | “Is every vote condemning an official a vote that removes them?” | “What changed between Chile's ordinary critical resolution and the constitutional accusation? Why does the change target the minister rather than automatically the President?” | “A special, valid constitutional process removes one minister in a presidential setting. Does that establish routine parliamentary confidence over the whole executive? Explain the missing relationship.” |
| Denominator and clock | “What does ‘majority’ mean if some members abstain? Which event starts a deadline?” | “Explain Portugal's denominator, Germany's different procedures and Taiwan's two time-sensitive actions in your own words.” | “An invented 61-member chamber has 30 yes, 25 no and 6 abstentions. Compare an all-member majority rule with a votes-cast majority rule. In a separate invented seven-day rule, identify the deadline from its stated starting event rather than the latest event in the story.” |
| Conditional comparison and research | “What would convince you that one system changes cabinet turnover? Is that the same as changing policy delay?” | “Choose one game observation and turn it into a conditional claim; name the outcome, mechanism, indicator, rival and challenging result.” | “Two countries differ in cabinet turnover and party fragmentation. What further evidence would help distinguish an institutional explanation from a party explanation? What would make you revise your claim?” |

The materials strongly support *encountering* the answers to the after-play questions. They provide less direct evidence for self-generated explanations and genuinely altered transfer cases. That is the gap an eventual student study should investigate, rather than treating completion points as its answer.

## What is already strong

- **One clear relational spine:** origin, survival and policy are repeatedly separated across the guide, game, notebook and worksheet.
- **A memorable contrast:** “Cabinet out. Assembly still here” and “Same seats. New government” make a common misconception tangible.
- **Substantive national variation:** all nine cases earn their place. The second moves change targets, procedures, denominators or consequences rather than simply repeating a family label.
- **A useful reference environment:** the notebook glossary, comparison table, rule drawers, free clues and primary-source links normalize consulting evidence. They should remain available during reasoning tasks.
- **Repair without a speed penalty:** first answers and revisions are preserved, hints are free, and no reward is given for speed, ideology or staying in office.
- **Honest model boundaries:** the guide, final screen and report distinguish selected examples and hypotheses from established causal effects, and separate cabinet turnover, policy delay and democratic survival.
- **Conditional tradeoffs:** the actual flood-funding dilemmas present benefits and costs of bargaining, scrutiny, leadership change and coordinating dual mandates, without a universal institutional ranking.
- **Transparent records and submission:** local saving, device transfer, backup, report export and student submission responsibilities are explained; written reasoning is not represented as automatically graded.

## What can reasonably be claimed about the end state

The design gives a learner several opportunities to explain why a bill can fail without an executive leaving, why a cabinet can change without a new assembly election, why a President's title is insufficient for classification, and why country-specific procedures matter. It also offers an appropriate beginning for conditional institutional comparison.

This review cannot establish that learners actually acquire those abilities, remain engaged for the estimated time, or can independently design comparative research. The most consequential next improvement is to connect the existing rich material to short, repeated explanations and a common rule-reading method, then test those explanations with first-time students. Keep the rigor, all nine countries, their twists and the caution about causal inference.

# Independent student-learning review B — locked

Locked on 2026-10-29. This is an AI review, not a study with students. The two entry points below are analytical perspectives, not invented respondents. No gains, success rates, completion times, or participant reactions were measured.

## Scope and judgment

I inspected the player-facing material in `blind-round-2/input/`: the HTML shell, the screen text reconstructed from `power-ui.js`, the nine displayed dossiers in `cases.json`, and all five rendered pages of `Student_Guide.pdf` and `Student_Worksheet.pdf`. I did not consult the original repository, tests, earlier feedback, other reviewers, or the engine. I did not inspect the constitutional sources for accuracy. This assesses instructional clarity, load, engagement and the evidence of learning, not authoritative constitutional correctness. Screen reconstruction is not a live browser usability or accessibility test.

The game has a strong instructional spine: distinguish getting office, retaining office and passing a bill; hold political inputs constant; then change the country-specific procedure. The paired country twists do real conceptual work. The weakest link is the inference from completing them to independently explaining or transferring the mechanism. Correct choices can be made with answer-revealing help, prominent family labels and feedback-driven repair. Those are useful learning supports, but completion and points do not establish unassisted understanding. In learning mode the causal explanations and tradeoff reasoning can be empty; in class mode a word minimum still does not establish their quality. The UI and guide candidly acknowledge the latter limitation.

Evidence references below use `U` for `power-ui.js` line numbers, `C` for `cases.json` line numbers, and PDF page numbers. These refer to the frozen review input.

## Two first-time entry points

| Entry point | What already helps | Where the learning can still stop short |
|---|---|---|
| A student knows the terms but thinks a lost bill removes the government | U93 explicitly says the 48–52 bill is ordinary and has no confidence condition. Its hint asks separately whether the law passed and whether a rule removed the officeholder. U94 then names a different formal procedure; U95 makes the student use the existing assembly. This is an excellent direct confrontation of the misconception. | The same policy answer repeats under all three rulebooks (U84), and the correct continuation is described before choosing it (U95). A student can learn the phrase “not automatically” without determining when a bill is linked to confidence. France's second move is the key counterexample; the final comparison should make the student connect it back to the ordinary-bill rule explicitly. |
| A curious first-year wants to know why this matters | The storm-relief setting, housing, employment and transport promises make political bargaining tangible (U77, U180). The warm-up is short, unscored and says “No political background needed” (U79). The notebook defines government, confidence, investiture, censure and dissolution (U177). | The warm-up asks the learner to perform two instructed clicks rather than explain a consequence. Much of the vocabulary needed later is in a separate notebook. “Confidence and supply,” “reconstitute,” “tender resignation,” and “normal democratic institutional operation” are demanding without an immediate gloss. The city needs relief, but the main story rarely makes the consequences of a blocked bill concrete. The research desk then requires a sizeable jump into comparative-method reasoning. |

Neither perspective justifies predicting how actual students will behave. Both identify things that a real pilot should observe.

## Positives to preserve

- **A clear mechanism instead of a list of labels.** “The wires matter more than the job titles” (U77), “Origin” and “Survival” (U100/U177), and the separate bill/cabinet/assembly status strip (U82) reinforce the same distinction.
- **An honest matched comparison.** U80/U92 and Guide p1 disclose the fixed 48–52 and 58–42 shocks and the fictional majority/commitment assumptions. The live Midnight Session separately lets bargains change the votes and then holds that result constant across constitutions (U152–160). This is a good basis for explaining a controlled comparison.
- **Substantial within-family variation.** The second moves introduce outside confidence support, constructive replacement versus dissolution, different targets, accusation versus conviction, exceptional ministerial removal, a linked bill, abstentions and a legal clock. They are not just nine reskins of the same recognition question.
- **Repair without speed or ideology scoring.** U72/U138 and Guide p2 preserve first answers and repairs, make hints free and reject points for speed, staying in office or a political preference. Read-only stage review is available without overwriting first decisions (U148).
- **Good causal modesty.** Guide p2, Worksheet p3 and U122 explicitly separate a hypothesis from a finding and cabinet turnover from policy delay and democratic survival. The tradeoff lab supplies costs and conditions rather than declaring a universally best system (U121).
- **Usable print structure.** Both guide pages and all three worksheet pages render without clipping. The worksheet provides generous writing space and follows the on-screen progression. The guide is compact enough to serve as a reference, though its broad letter spacing makes some long lines visually laborious. This is a visual observation, not a measured readability result.

## All nine dossiers: instructional demand and load

Every dossier uses the same incident → twist structure. The family and focus are visible before answering (U99–100), the rule drawer opens after an initial answer, and “Rule for this twist” gives the correct option's feedback (U100). These affordances are appropriate for supported practice; they limit what a correct first choice proves. The only written country explanation, “What does this case add to the general pattern?”, is optional and appears after the stamp (U100).

| Dossier and exact evidence | Learning opportunity | Load / rote risk and a concrete refinement |
|---|---|---|
| **United Kingdom:** “The largest party cannot find a partner”; then “Support without joining the Cabinet” (C38–69) | Separates largest party, cabinet membership, confidence support and support for a particular bill. This directly repairs both automatic-largest-party and automatic-bill-dismissal rules. | The 43-seat cabinet plus 12 outside supporters is useful, but “confidence and supply” is unexplained in the incident. Give that phrase a one-line gloss and ask which commitment survives the transport defeat. Keep both moves and the numbers. |
| **Germany:** 55 against the Chancellor, 47 for a successor; then a failed Article 68 request (C106–136) | Requires a majority *for someone*, then prevents transferring the same successor condition into a different procedure. | Article numbers, denominator, successor requirement, presidential discretion and 21 days accumulate. The basic drawer calls Article 68 “a separate confidence/dissolution mechanism” (C104), but does not supply the 21-day rule. Offer a short procedure excerpt, then require the student to identify what differs from the first move. |
| **South Africa:** section 102(2) President-targeted motion versus section 102(1) Cabinet-excluding-President motion (C172–202) | One of the strongest challenges to title-based reasoning. It also shows that a binding consequence can change without becoming “no consequence.” | Closely similar section numbers invite matching rather than tracing. Keep the legal identifiers, but have the student name the target and consequence in plain language before revealing feedback. The visible parliamentary label means this is not an independent classification test. |
| **United States:** rejected education bill; then House articles with no Senate conviction (C238–270) | Separates policy blockage, accusation and completed removal. It provides a useful check on the misconception that staying in office makes a bill law. | Both answers retain the President, so a superficial “President always stays” strategy can work. Ask what additional stated procedure would change the second answer, without implying immunity. Keep impeachment/conviction explicitly country-specific. |
| **Brazil:** coalition partner leaves; then presidential ministerial dismissal costs legislative votes (C308–338) | Distinguishes appointment authority from command over legislators and shows that coalitions are not uniquely parliamentary. | Both moves again preserve presidential office; several distractors use conspicuous absolutes such as automatic resignation or bypassing Congress. Require an explanation of what *does* change: ability to assemble votes. A valid answer must mention legislative support, not just a fixed term. |
| **Chile:** coordinating minister and ordinary criticism; then a valid constitutional accusation with Senate guilt (C374–407) | Particularly valuable because the second move defeats an overgeneralized “only the President can ever remove a minister” rule. It separates job functions from a confidence-dependent PM. | The distinction between ordinary appointment power and exceptional accusation is demanding but worth preserving. Place the two procedures in a small comparison strip and have the learner specify why the minister leaves in one and not the other. Do not turn this into mere article-number recall. |
| **France:** cohabitation; then Article 49(3) with 48 votes for censure in a 100-member Assembly (C445–475) | The strongest stress test of the Chapter I bill slogan: the vote's object and legal link change. The same number 48 now has a different meaning. | Cohabitation and linked confidence/bill passage are two substantial ideas in one pair. Label the object of the 48 votes (“for the censure motion”), then ask the learner to explain why “48 means the bill loses” fails here. Keep the exception and full threshold reasoning. |
| **Portugal:** restricted presidential dismissal; then 48–40 with 12 abstentions under an all-member rule (C511–541) | Opposes both “unlimited presidential power” and “no dismissal power ever,” then tests the denominator rather than the larger side of votes cast. | The office-power question and arithmetic question can feel disconnected. A short “what changed in move 2?” prompt should say the new issue is passage of the legislative motion. Require deriving 51 from all 100, and later transfer with a chamber that is not 100 seats. |
| **Taiwan:** appointment without investiture; then resignation on day 8 and proposed dissolution on day 16 (C577–607) | Distinguishes entry from survival, required resignation from optional dissolution, and two deadlines. It is an excellent higher-demand case. | The basic drawer says dissolution is within “the constitutional period” (C575), while the twist needs the clock's starting event; the answer-bearing help supplies it. Supply the precise rule as a consultable premise before asking for application. Require a timeline with vote, resignation/request and decision. Preserve both deadlines and conditions. |

Across the extended route there are 18 country choices, after the repeated Chapter I sequence, before final comparison, three dilemmas and four research fields. This is a count of visible tasks, not an observed burden or timing claim. Chunking and short explanations can improve processing while retaining all nine dossiers and their difficulty.

## Eight actionable priorities

| Priority | Action and exact reason | Acceptance evidence |
|---|---|---|
| **1 — High: strengthen the exit task** | U103's “unfamiliar republic” states an elected fixed-term President and confidence-dependent PM; U102's correct option repeats those relations. The family-comparison drawer is immediately available. Keep this teaching item, then add a short unlabelled scenario with different titles/numbers and an explanation of office, procedure, condition and result. | A reader can score the explanation without seeing which option was selected. It must distinguish bill result, office survival and legislative continuation; the task must include a case where a slogan learned earlier would fail. |
| **2 — High: elicit the reason before answer feedback** | U72 makes repair easy, while U100 makes every passport note optional. After the second move, ask one brief “Which fact/procedure changed, and why does it change or preserve the result?” response. Rotate the fuller explanation burden across dossiers if needed; do not remove any cases. Replace some extreme distractors with plausible errors about the target, denominator or procedure. | Completion records distinguish first reasoning from a copied feedback explanation. A learner cannot earn evidence of mechanism simply by choosing the longest balanced option or repeating “not automatic.” |
| **3 — High: make rule help support reasoning** | U100's twist help reproduces the correct option's feedback. Germany's dissolution period and Taiwan's clock anchor particularly need usable premises. Put a short source-rule excerpt in the drawer, without its case verdict; keep direct corrective feedback after the attempt. | The provided rule is sufficient to derive the answer, but does not state the answer. Students identify the governing clause and apply it to the facts. Free help remains available. |
| **4 — Medium: bring essential vocabulary to first use** | The useful definitions are in the notebook (U177); the warm-up still promises no background (U79). Add a brief first-use gloss for confidence, censure, investiture, confidence and supply, caretaker, and dissolution. Add one unscored warm-up prediction about which institution remains after the cabinet leaves. | A first-year can state “government means this executive team” and distinguish it from the legislature before the first scored question, without navigating a long glossary. Keep the notebook for deeper reference. |
| **5 — Medium: turn comparison into an explicit learning move** | The main route clearly fixes later votes (U92), while the optional live lab changes them (U152). Ask learners to explain why these two modes have different purposes. In the lab, invite a prediction for one changed bargain and a one-variable replay, recording the predicted change before revealing it. | The learner separates a political input from a constitutional rule and can describe what was held constant. Main-route coalition negotiation is understood as formation practice rather than a promise to determine the fixed bill result. |
| **6 — Medium: make the public stakes visible without ranking systems** | The home asks “Who stays in office?” (U77), and the status strip shows office/bill outcomes (U82). Add a short narrative consequence of delay or replacement for residents, and a stakeholder prompt in the existing tradeoff response. | A response can name a concrete beneficiary, cost-bearer and condition while acknowledging that surviving office does not deliver relief. No policy preference or regime type earns moral points. |
| **7 — High: scaffold the method jump, retain its rigor** | U122 requires a hypothesis, test, rival and falsifier, but its single worked example is ready to copy. Add a compact planning row: outcome, unit/period, observable measure, comparison, rival and observation that would change the claim. Require one measure and one disconfirming observation in plain language. | “More stability” must be unpacked into cabinet turnover, policy delay or democratic survival. A proposal cannot pass a conceptual review merely by reaching 20/10-word minima or restating the worked example. |
| **8 — High: report evidence at the level it supports** | U138 usefully preserves answers and repairs but combines them with a completion passport. Keep those features; describe points as performance on supported predictions. Add independent explanation and confidence evidence using the check below. Use a brief “what changed my mind?” field for an assessment revision. | Reports distinguish initial supported choices, repairs, explanation quality and novel transfer. They do not infer learning gain from a repaired choice, a word count, a stamp or an optional lab run. |

## Proposed six-item pre/post/novel-transfer check

This is a proposed instrument, not a completed evaluation. All countries and rules below are synthetic. Supply the stated rule text within each item so the assessment tests using relationships, not remembering foreign article numbers. Use open responses, not family-labelled cards or multiple-choice elimination. Retain the game's actual nine-country difficulty; this check is an additional measurement layer.

**Administration.** Use A before play and B immediately after; counterbalance A/B order across learners when feasible. C is a novel application form, preferably collected after a delay such as one week, with no game notebook. Allow the rule text printed inside each item on every form. Use the same accommodations and no speed score. Do not give corrective pretest feedback. Collect every initial answer before any neutral revision probe or feedback, so later items are not taught by earlier feedback. Keep the C form unseen until its administration. If C is used immediately, label it immediate transfer rather than retention. Self-contained C items measure transfer of a reasoning method to a new rule card; they do not establish transfer to reading a full constitution or to real political news.

**Evidence per item.** Capture the answer, one causal sentence, the decisive rule clause and confidence (0 = guessing, 1 = uncertain, 2 = fairly sure, 3 = very sure). After all initial answers, use the same neutral probe, “Check the office, procedure, threshold or condition in the stated rule.” Preserve a second answer and confidence, plus “I changed/kept my answer because…”. Score initial and revised work separately. A revision is evidence of reasoning only when the explanation identifies the controlling difference; a changed letter or copied verdict is insufficient. Confidence is not graded. High-confidence errors, justified corrections and persistent uncertainty are separate useful observations.

Each item has four explicit criteria, one point each, with no point for jargon alone. Report item profiles; keep items 1–5 (institutional reasoning) separate from item 6 (research inference). Do not hide these different objectives in one general learning score.

### 1. Bill outcome versus office survival, including a changed legal link

- **A:** A cabinet has continuing confidence support. An ordinary relief bill needs more yes than no votes; it receives 43 yes and 47 no. No confidence condition is attached and no other removal or dissolution process occurs. What happens to the bill, cabinet and assembly? Explain.
- **B:** An ordinary housing bill receives 38 yes and 42 no under the same rule. The cabinet still has a separate confidence agreement. Give the three outcomes and the controlling reason.
- **C:** In a 70-member chamber, a valid special rule makes a finance bill law unless a no-confidence motion wins a majority of *all* members. The motion receives 34 votes; 36 members do not support it. What happens to the motion, bill and cabinet? Is a new assembly required?
- **Criteria:** (1) A/B bill fails; C motion fails and bill becomes law. (2) Cabinet remains on these facts. (3) Explanation uses the absence of a confidence link in A/B and the special link plus 36-vote threshold in C. (4) Assembly remains; no automatic election is invented. Merely saying “bill defeat never removes governments” fails criterion 3.

### 2. Origin and survival instead of office titles

- **A:** An assembly elects an official called President; that official heads the cabinet. A valid majority no-confidence vote can require this executive to resign. There is no separate popular executive election. Which relationship fits, and why?
- **B:** A legislature selects a Chief President and cabinet, whose survival depends on its confidence. Voters elect the legislature but do not separately elect the executive. Explain the relationship without relying on the title.
- **C:** Voters separately elect a fixed-term President. A minister called Prime Minister coordinates cabinet work, but only the President has ordinary appointment/dismissal authority over that minister; there is no legislative confidence responsibility. The legislature can block bills. Does the second title establish semi-presidentialism?
- **Criteria:** (1) Correct source of executive office. (2) Correct ordinary survival relationship. (3) A/B parliamentary; C presidential in the relationship scheme used by the game. (4) Explicit rejection of title-counting: two titles or the word President do not supply the missing confidence relationship. A label alone earns at most one point.

### 3. Appointment, confidence and the precise target

- **A:** Voters elect a fixed-term President, who appoints a PM without an investiture vote. A later valid confidence motion targets the PM and cabinet and requires their resignation. It passes. No dissolution process is initiated. Who leaves and who remains? Does appointment without investiture change this?
- **B:** A separately elected President appoints a premier after an investiture vote. A valid successful no-confidence motion expressly targets the premier and cabinet, not the President. No other procedure occurs. Give each institution's status and reason.
- **C:** An Assembly-elected President heads the executive. The stated rules distinguish (i) no confidence in the President, requiring executive resignations, and (ii) no confidence in the Cabinet *excluding* the President, requiring the President to reconstitute the Cabinet. Procedure (ii) validly passes. What is required, and which inference from the first procedure would be wrong?
- **Criteria:** (1) A/B PM and cabinet must leave; C Cabinet must be reconstituted. (2) President remains for the correct stated reason, not because all Presidents are immune. (3) Assembly remains absent a separate dissolution process. (4) Entry procedure does not cancel later responsibility in A/B; C must apply the expressly different target/consequence rather than import procedure (i).

### 4. A majority against is not necessarily a majority for; denominators matter

- **A:** In an 84-member chamber, removal requires simultaneously electing a successor by a majority of all members. Forty-eight dislike the incumbent, but only 41 support the proposed successor. Can this attempt replace the incumbent, and what is missing?
- **B:** In a 120-member chamber with that same rule, 70 oppose the incumbent and 59 support the successor. Give the result and threshold.
- **C:** A different 120-member chamber's rule requires electing a successor by a majority of votes *cast for or against*, excluding abstentions. The successor receives 41 for, 39 against and 40 abstentions. Does this constructive procedure succeed?
- **Criteria:** (1) A/B do not remove; C elects the successor and replaces the incumbent. (2) Correct thresholds: A 43, B 61, C 41 of 80 votes cast. (3) Distinguishes opposition to the incumbent from votes for a successor and uses the stated denominator. (4) Does not invent an automatic general election. Applying “51” or an all-member majority indiscriminately fails criterion 2.

### 5. Separate clocks and conditional dissolution

- **A:** The rule requires a premier to resign within ten days of successful no confidence. The President *may* dissolve within ten days of passage of that motion, on the premier's request and after consultation. Passage is day 0; the premier resigns and requests on day 8; the President proposes dissolution on day 16 after consultation. Is it available on these facts?
- **B:** A new rule requires resignation within twelve days, but permits dissolution only within seven days of motion passage, on request and after consultation. Passage is day 0; request/resignation day 5; proposed decision day 10 after consultation. Explain the outcome.
- **C:** Under B's rule, the premier requests dissolution on day 2 and the President proposes it on day 6, but the required consultation has not occurred. Is being inside the window sufficient?
- **Criteria:** (1) Uses motion passage, not resignation, as the clock anchor. (2) A day 16 exceeds day 10; B day 10 exceeds day 7; C day 6 is timely. (3) A/B unavailable on timing; C not yet available because consultation is also required. (4) Distinguishes a conditional power from automatic dissolution: neither cabinet departure nor a request itself ends the legislature or removes the President.

### 6. Public stakes and the limits of a game-generated causal claim

- **A:** A player says, “The President survived criticism, so fixed terms cause faster relief delivery.” The only record shows a surviving President and a blocked relief bill. Evaluate the claim; propose an observable comparison, a rival and a finding that would change it.
- **B:** A player says, “Requiring a successor makes cabinet removal harder, so it always reduces policy delay and protects democracy.” The record only shows a failed successor vote and a pending budget. Evaluate the claim and propose the same four elements.
- **C:** A report observes fewer cabinet replacements after a confidence-rule change, unchanged budget delay, and an electoral-system change at the same time. It concludes that the confidence rule improved democratic survival. What is observed, what is not established, and what further comparison or challenging evidence is needed?
- **Criteria:** (1) Separates cabinet turnover, policy delay/delivery and democratic survival. (2) Rejects the stated causal leap: neither game outcomes nor the co-changing before/after pattern isolate it. (3) Specifies units/periods and an observable outcome measure in a comparison, and gives a plausible rival such as electoral change, fragmentation, majority support or crisis severity. (4) States a concrete observation that would revise the proposed mechanism/claim, rather than “need more evidence.” Multiple defensible hypotheses are acceptable; institutional or ideological preference is not scored.

## What evidence would support learning, and what would not

Compare the initial A/B criterion profiles, explanations and confidence, then examine C separately. A stronger result would be justified explanations on new numbers, changed targets and altered rules, with students identifying why an earlier rule no longer applies. A justified revision after the neutral probe would show monitorable reasoning; it is not equivalent to an initially correct unprompted answer. A correct label with no mechanism, a repair after a supplied verdict, or a phrase lifted from the notebook would support a narrower conclusion.

No pre/post difference alone establishes that the game caused a change: repeated testing, reading the supplied rule cards, instruction between tests and prior knowledge can contribute. A real evaluation should preserve forms and scoring rubrics, score explanations without knowing whether they are pre or post when feasible, review ambiguous scores with a second scorer, and inspect item-level errors rather than report only an overall total. No such evaluation has been conducted in this review.

This independent review is locked. I will not read other reviews and revise it in light of them.

# Independent blind faculty review A — comparative politics

**Reviewer status:** Independent AI expert review in the role of a political-science faculty reviewer; this is not a review by a recruited professor or human study participant.

**Review locked:** 29 October 2026. **Packet examined:** only `/workspace/scratch/a5fb32e24502/blind-round-2/input/`. I did not read the original repository, earlier reports, other reviewers, or other agents' work. I made no changes to the packet and used no subagents.

## Judgment

The frozen packet is substantively strong. All nine countries are defensibly classified under the explicitly stated institutional definition, and all 18 keyed country moves produce defensible constitutional outcomes under their stipulated facts. I found **no high-severity country answer-key error**. The important remaining factual correction is a Portuguese article attribution: the absolute-majority consequence belongs to **Article 195(1)(f)**, not Article 194. The numeric answer remains correct.

The core and optional lab are defensible teaching models because their fictional thresholds, fixed roll calls, and limited purpose are disclosed. They should not be treated as country constitutions or estimates of institutional effects. Recommended changes below preserve all nine countries and 18 moves. The most useful refinements sharpen the distinction between a support agreement and cabinet membership, and make variation in presidential dismissal power more explicit.

Severity meanings: **High** changes the correct constitutional outcome or core classification; **Medium** is a consequential citation, representation, or coverage problem that should be corrected before teaching from the claim; **Low** is a bounded clarification or enrichment. A pedagogic simplification is not counted as a factual error merely because a fuller constitution has additional rules.

## Six actionable priorities

### 1. Medium — verified attribution error: Portugal move 2

**Location:** `cases.json`, Portugal `challenge.scenario` and distractor feedback; also the copies in `Sources.json` and embedded `The_Power_Circuit.html`.

The scenario says to apply “Article 194's requirement” of an absolute majority. Article 194 governs initiation and consideration of no-confidence motions. **Article 195(1)(f)** supplies the absolute-majority requirement and resignation consequence. Replace the attribution with “Articles 194 and 195(1)(f), including Article 195(1)(f)'s absolute-majority requirement,” and correct the feedback. Keep the 48–40–12 arithmetic and keyed answer. This is a real precision error, not an incorrect outcome. Evidence: [S7], Article 195(1)(f), PDF printed page 75; Article 194, printed page 74.

### 2. Medium — coverage refinement: show the other semi-presidential dismissal relationship

**Location:** Taiwan origin/survival/twist drawer; general semi-presidential comparison.

France and Portugal correctly constrain claims of an unrestricted presidential power to dismiss a confidence-supported government. Taiwan is an opportunity to make the contrasting presidential relationship explicit: Additional Article 2 addresses presidential orders removing the premier without the premier's countersignature. The present Taiwan moves test legislative responsibility and dissolution timing, so they do not contradict that rule; they simply leave the contrasting route implicit. Add one brief sentence to its drawer and a comparative question asking how this differs from France's formal Article 8 rule. Name the premier-presidential/president-parliamentary distinction if appropriate for the course, while treating Portugal's exceptional safeguard power carefully rather than forcing it into an unqualified binary. **This is an omission, not an answer-key error.** Evidence: [S6], Article 8; [S7], Article 195(2); [S8], Additional Article 2.

### 3. Medium — graph wording can reverse the intended presidential lesson

**Location:** `power-art.js`, presidential block around line 39.

The box reads `NO CONFIDENCE` followed by `REMOVAL LINK`. A novice can parse that as a label for a no-confidence removal mechanism, although the code and description intend the absence of that mechanism. Use unambiguous text such as “CENSURE DOES NOT / REMOVE THE PRESIDENT,” with a short nearby indication that exceptional removal has separate rules. Do not label the presidency as immune from removal. The actual arrows and state transitions correctly omit an ordinary assembly-to-president confidence dependency. This is a representational ambiguity, not a constitutional error in the engine.

### 4. Low — put the lab's forced opposition assumption beside the result

**Location:** `power-lab.js` `compare()` and `power-ui.js` lab result screen.

The lab defines challenge votes as `100 - confidenceSeats`. This is a transparent fictional assumption, disclosed in the assumptions and enacted as a separate motion. It is not a general rule that every legislator outside a signed agreement votes to remove the cabinet. Add a one-line result-screen reminder: “Here, all legislators outside the continuing agreement vote for the challenge; real legislatures can also abstain or tolerate a minority government.” Keep the independent bill-vote model and matched comparison. Otherwise learners may transfer the arithmetic to the UK minority-government case despite that case's correct explanation. Evidence for the real-world boundary: [S1], confidence need not mean cabinet seat-majority ownership or winning every policy vote.

### 5. Low — label legislative support independently of cabinet coalition membership

**Location:** core `coalitionBuilder()`, PM selector/help, and lab alliance terminology.

The software's `coalition` array functions primarily as a set of legislative confidence supporters; it does not model the distribution of ministerial posts. The UK second move correctly teaches outside support. Use “support agreement” or “supporting parties” as the broad interface term, reserving “cabinet coalition” for parties actually holding posts. A short note can explain that supporters may remain outside the cabinet. This needs no additional country, no reduction in arithmetic, and no new cabinet-allocation mechanic. The 51-seat investiture and selecting a PM from an agreeing party remain acceptable explicitly fictional constraints.

### 6. Low — finish the German Article 68 time-window explanation

**Location:** Germany `challenge` help/feedback.

The keyed continuation is correct on the stated facts: dissolution is discretionary, proposed by the Chancellor, and limited to 21 days. Add that this authority lapses if the Bundestag elects another Chancellor by the required majority before dissolution. The scenario already stipulates no successor at its starting point and satisfaction of the other requirements, so omission of a later intervening event is a defensible simplification. A short addition prevents “21 days” being learned as unconditional authority throughout the period. Evidence: [S2b], Article 68(1).

## Complete audit: nine countries and 18 country moves

“Correct” below means the keyed answer follows under the incident's express assumptions. It does not endorse every broader inference a student might make from the country label.

| Move | Country and mechanism | Finding | Official basis |
|---|---|---|---|
| 01 | UK: largest party lacks support; alternative 58-seat agreement | Correct. Confidence, rather than a largest-party entitlement, governs the stated choice. Incumbent negotiation conventions do not overturn an already specified viable alternative. | S1; S1b |
| 02 | UK: 43-seat cabinet plus 12 outside supporters; ordinary bill defeated | Correct. The expressly continuing confidence-and-supply arrangement can sustain a minority cabinet. Ministerial posts are unnecessary for outside support. | S1 |
| 03 | Germany: 55 against Chancellor but only 47 for successor | Correct. Article 67 needs a majority of all members for the successor; 51 in the toy chamber. | S2a |
| 04 | Germany: failed Article 68 request, Chancellor proposes dissolution | Correct and suitably conditional. It is a separate procedure, not automatic dismissal; add the lapse rule as priority 6. | S2b |
| 05 | South Africa: section 102(2) motion in President | Correct. President, other cabinet members and deputy ministers must resign. The title does not establish separate presidential survival. | S3 |
| 06 | South Africa: section 102(1) motion in Cabinet excluding President | Correct. Reconstitution is required; this motion does not itself require presidential resignation. | S3 |
| 07 | US: House blocks education bill; no exceptional departure | Correct. The bill's defeat does not end the separate presidential term; the President has no power to dissolve Congress. | S4 |
| 08 | US: House impeachment before Senate conviction | Correct. House impeachment alone does not remove; conviction requires two thirds of senators present. | S4 |
| 09 | Brazil: legislative coalition partner withdraws | Correct. Renegotiation is a possible response; loss of legislative support itself is not presidential removal. | S5 |
| 10 | Brazil: presidential dismissal of minister costs legislative support | Correct. Ministerial appointment authority and the votes needed for a bill are different relationships. | S5 |
| 11 | Chile: cabinet coordinator faces ordinary critical resolution | Correct. The President retains ordinary ministerial dismissal authority. For precision, identify the criticizing chamber as the Chamber of Deputies. | S9, Articles 32(7), 52(1)(a) |
| 12 | Chile: valid Senate guilty finding against minister | Correct. Exceptional removal of that accused minister does not establish ordinary cabinet confidence or remove an unaccused President. | S9, Articles 52(2), 53(1) |
| 13 | France: opposing cohesive Assembly majority sustains its PM | Correct. Cohabitation can accommodate the separate presidency and government responsibility. No automatic presidential removal is implied. | S6, Articles 6, 8, 20, 49–50 |
| 14 | France: valid Article 49(3) finance-bill procedure; 48/100 for censure | Correct. The censure threshold is unmet; the specified procedure deems the bill passed. “Validly uses” reasonably brackets procedural prerequisites. | S6, Articles 49(2)–(3), 50 |
| 15 | Portugal: policy dislike, no established institutional-normality problem | Correct. The facts do not establish an unrestricted dismissal power; neither is dismissal categorically impossible. | S7, Article 195(2) |
| 16 | Portugal: 48 yes, 40 no, 12 abstain | Correct outcome, incorrect article attribution. It needs 51 of the fictional 100 members in office; see priority 1. | S7, Articles 194 and 195(1)(f) |
| 17 | Taiwan: appointed without confirmation; valid no confidence passes | Correct. No investiture requirement does not negate later legislative responsibility; resignation is due within ten days. | S8, Additional Article 3 |
| 18 | Taiwan: dissolution proposed on day 16 after vote, day 8 after resignation | Correct. Presidential dissolution's ten-day clock begins at passage of no confidence, not the later resignation. | S8, Additional Article 2 |

South Africa's inclusion is particularly valuable: sections 86–87 establish Assembly origin despite the President leaving Assembly membership, and section 102 establishes confidence dependence. Calling it parliamentary is defensible under the game's declared origin-and-survival relationship test. No officeholder or current party balance is needed for that classification.

## Graph semantics and controlled comparisons

I inspected `power-art.js`, `power-engine.js`, `power-lab.js`, the relevant UI wording, and the two supplied student PDFs. The graph uses separate election, formation/appointment, and confidence edges. The parliamentary confidence arrow terminates at the cabinet, and the semi-presidential confidence arrow terminates at the PM/cabinet, while the President retains a separate electoral path. Cutting confidence leaves assembly and presidential election paths intact. The outgoing/caretaker label prevents the disappearance of administrative government being inferred from resignation. The formal appointing head of state is omitted from the fictional parliamentary graph; that is a defensible abstraction because the named-country drawers supply constitutional variation.

The shared semi-presidential appointment arrow compresses PM appointment and ministerial appointment. It is acceptable for Port Meridian; students should not infer unfettered presidential selection of every minister in France or Portugal. The fiction notice is present directly under the graph. An optional caption distinguishing selection constraints from formal appointment would improve transfer.

Read-only semantic probes of the supplied modules confirmed:

- A 48–52 ordinary bill defeat removes no executive in any of the three rules.
- The separate 58–42 challenge removes the parliamentary or semi-presidential PM/cabinet, while ordinary presidential censure removes neither President nor cabinet. No assembly election is triggered.
- The lab can represent a passing bill with a lost confidence motion (60 bill votes, 42 confidence seats), and a failed bill with cabinet survival (48 bill votes, 58 confidence seats). At 50–50 its expressly all-member threshold leaves the challenge unsuccessful.
- All nine dossiers contain two moves and each move contains exactly one keyed answer.

The fixed 48–52 and 58–42 votes are explicitly described as new roll calls, not derived from the player's initial coalition. The lab's matched table holds the final bill and confidence support constant and explicitly assumes an executive is already in office; this avoids treating a presidential minority launch as satisfying the fictional PM's investiture. These are sound controlled illustrations of a legal consequence. They are not causal experiments showing that changing a constitution leaves political behavior unchanged, and the packet correctly disclaims that stronger interpretation.

## Strengths to retain

1. **High instructional value:** Origin and survival are consistently separated from job titles, and the South Africa pair tests that distinction directly.
2. **High instructional value:** Ordinary bill defeat, binding confidence, censure, and exceptional removal are distinguished across the full case set. France's second move properly complicates the ordinary-bill rule.
3. **High instructional value:** Germany, Portugal and Taiwan require attention to the actual procedure, denominator or clock, rather than recognition of a country label. Preserve these demanding second moves.
4. **High instructional value:** Coalition bargaining is present in presidentialism, while the UK demonstrates outside confidence support. Neither coalition presence nor party count is mistaken for a constitutional family.
5. **High instructional value:** France's Article 8 and Portugal's Article 195(2) avoid a common false generalization about presidential dismissal. Cohabitation preserves both mandates and is not called a change of system.
6. **High instructional value:** Cabinet replacement and dissolution are separate, and the research tasks distinguish cabinet turnover, policy delay and democratic survival. Conditional tradeoffs, rivals and falsifiers make the exercise suitable for analytical comparative-politics teaching.

## Official evidence register

All sources below were independently retrieved or searched on **2026-10-29**. URLs are exact official sources used. The findings above paraphrase only the provisions relevant to this audit. Search excerpts were used where an official HTML fetch timed out; that limitation is stated rather than treated as a successful direct read. No secondary commentary is necessary for the constitutional verdicts.

| Code | Official URL and provision | Supported finding and access |
|---|---|---|
| S1 | https://commonslibrary.parliament.uk/research-briefings/cbp-10915/ — *Commanding confidence*, published 29 June 2026; sections on confidence, appointment, testing and loss | Minority/outside support is possible; an ordinary defeat need not remove government; clear confidence loss conventionally entails resignation or requesting dissolution. Direct official page read. |
| S1b | https://www.parliament.uk/about/how/elections-and-voting/general/hung-parliament/ — formation and largest-party sections | Coalition/minority formation and the absence of a conclusive largest-party entitlement. Direct official page read. |
| S2a | https://www.gesetze-im-internet.de/gg/art_63.html and https://www.gesetze-im-internet.de/gg/art_67.html — Basic Law Articles 63, 67 | Bundestag election and formal presidential appointment; constructive replacement requires a member-majority. Article 63 directly read; Article 67 independently verified in official search results after timeout. |
| S2b | https://www.gesetze-im-internet.de/gg/art_68.html — Article 68; corroboration https://www.bundeswahlleiterin.de/en/service/glossar/n/neuwahl.html — official guidance updated 23 July 2026 | Chancellor's failed confidence request can open discretionary dissolution within 21 days; a majority-elected successor ends the power. Constitutional official search excerpt plus current official election-authority guidance. |
| S3 | https://www.justice.gov.za/constitution/chp05.html — Constitution sections 83, 86–87, 89, 102; backup https://www.justice.gov.za/legislation/constitution/saconstitution-web-eng.pdf | Assembly origin, departure from Assembly membership, distinct exceptional removal, and the two confidence targets/consequences. Chapter page directly read; section 102 confirmed via official search and official PDF excerpt after a later fetch timed out. |
| S4 | https://www.archives.gov/founding-docs/constitution-transcript — Articles I §§2–3, II §§1, 3–4; https://www.archives.gov/electoral-college/provisions — Twelfth Amendment | Separate electoral origin, term and legislative powers; House accusation versus Senate conviction/removal. Original electoral clauses must be read with the amendment, as the packet appropriately recognizes. Both official pages read. |
| S5 | https://legis.senado.gov.br/norma/579494/publicacao/33296461 — Constitution Articles 76–77, 82, 84(I), 85–86 | Separate presidency, ministerial dismissal authority, and distinct responsibility proceedings. Direct official Senate English constitutional text read. The two incidents do not depend on changing inauguration-date details. |
| S6 | https://www.elysee.fr/en/french-presidency/constitution-of-4-october-1958 — Constitution Articles 6, 8, 20, 49–50 | Popular presidency, appointment/resignation relationship, government responsibility, censure threshold and finance-bill procedure. Direct official constitutional text read. |
| S7 | https://www.parlamento.pt/sites/EN/Parliament/Documents/Constitution7th.pdf — Constitution Articles 121, 187, 194, 195; printed pages 73–75 for government rules | Presidential election/appointment, constrained government dismissal, and the correct provision for the all-member no-confidence majority. Official parliamentary PDF text read. |
| S8 | https://english.president.gov.tw/Page/95 — Additional Articles 2–3 | Direct presidency; premier appointment/removal orders, displaced confirmation provision, legislative responsibility, resignation deadline and vote-based dissolution clock. Direct official page read. Do not substitute superseded main Articles 55 and 57. |
| S9 | https://www.bcn.cl/leychile/navegar?idNorma=242302 — Constitution Articles 25–26, 32(7), 52(1)(a), 52(2), 53(1) | Fixed elected presidency, ordinary ministerial dismissal, nonbinding political observations, and exceptional ministerial accusation/removal. Current official consolidated HTML directly read, including relevant clauses; this succeeded even though the packet records earlier access failures. |

## Review boundaries

This review evaluates institutional correctness, legal-mechanism transfer and the specified graph/counterfactual semantics. It is not a report of classroom learning gains or a participant study. I did not verify every constitutional article unrelated to these moves, real electoral outcomes, current officeholders, or optional Lithuania as a tenth playable case. The supplied packet's nine-country/18-move rigor should be retained; correction and targeted clarification are preferable to reducing its scope.

# Locked independent faculty review B: pedagogy and research methods

**Reviewer:** AI professor perspective in comparative politics and research methods. This is a simulated expert review, not a claim that a real professor or student participated. **Date:** 29 October 2026. **Status:** final, locked; no peer reviews consulted.

## Scope and judgment

I read only the supplied files in `blind-round-2/input/`, including all nine country cases and eighteen moves, the source and tradeoff data, the standalone HTML and its supplied JavaScript, and text extracted from the two student PDFs. I did not consult the repository, tests, earlier reports, or other reviewers. This is a content and assessment review, not a runtime test, visual accessibility certification, or observation of learners. Public constitutional and primary research sources were checked independently for consequential content claims.

The game offers a sound guided introduction to executive **institutional origin and survival**: who selects an executive, whose confidence matters, and which vote changes which office. It is strongest where it distinguishes ordinary policy defeat, cabinet replacement, exceptional removal, and legislative dissolution. “Origin” here means the source of office, not the historical emergence of a constitution or regime type.

It provides opportunities to practice country differences, conditional tradeoffs, and research design. Its current completion and points cannot establish that players can explain these relationships independently. Corrective multiple choice with immediate feedback, optional country notes, and word-count gates leave that inference open. The correct next step is an assessed teaching pilot with all nine cases and eighteen moves retained, stronger first-attempt explanations, and a separate transfer instrument. No reduction of cases or constitutional rigor is recommended.

### What the evidence can and cannot establish

| Evidence level | Defensible conclusion |
|---|---|
| **Validated rules/content** | The checked sources support the central origin/survival distinctions and the paired country answers, subject to the Portugal article attribution correction below. These checks validate specified rules, not every contemporary political practice. |
| **Supported design inference** | The matched roll-calls, office diagrams, paired procedural twists, and research prompts make the intended distinctions available for teaching. First answers are preserved in several important records. These are properties of the supplied design. |
| **Unproven learning outcomes** | There are no supplied learner observations, pre/post results, delayed tests, scored explanations, inter-rater results, or comparative treatment data. Increased understanding, retention, far transfer, enjoyment, and the advertised duration remain untested. A completed passport is a completion record. |

## Strengths to preserve

- The field guide opens with the right three outcomes: “who gets office, who keeps office, and who can pass policy” (Student Guide, p. 1). The repeated 48–52 ordinary defeat and 58–42 formal challenge make that distinction inspectable.
- The game explicitly labels its assumptions. Formation says “an explicit majority of all seats must support the PM” **for this fictional formation**, then acknowledges minority governments (`power-ui.js`, line 92). The lab discloses its deterministic vote ledger and satisfaction rule.
- Both moves must be correct for a country stamp (`caseComplete`, line 144). The second moves contain substantive changes of target, procedure, denominator, or clock; they are not merely nine flags on one seat-counting problem.
- The notebook defines minority government, constructive no confidence, cohabitation, and censure carefully. The minimal Elgie classification is named, and institutional family is separated from federalism, electoral proportionality, party count, and democratic quality.
- The guide says “Word checks measure completion, not the quality of reasoning” (p. 2). The research desk warns that the selected countries are not a controlled causal test. No points reward ideology, speed, or staying in office.
- `decisionRecord` preserves a first prediction and appends repairs; `reflectionRecord` preserves the first saved explanation. The report displays many of these records. This is a useful base for formative assessment and should be extended rather than replaced.

## Nine ordered issues and precise teaching changes

### 1. High: score and completion reward successful repair, not demonstrated explanation

**Displayed evidence.** Predictions earn “2 points first try or 1 after repair”; a repaired choice displays “Circuit repaired.” The passport prominently displays “circuit points,” “first predictions correct,” and “connections repaired” (`power-ui.js`, lines 73, 79, 138). `score` computes `correct * 2 + repaired` (`power-engine.js`, line 246). Completion requires eventually correct predictions and response lengths, without evaluating the meaning of the prose (`complete`, line 137).

**Implication.** A learner can reach every correct option by trying alternatives after feedback. With the full nine-country route there are 31 scored prediction IDs: twelve rulebook predictions, eighteen country moves, and one transfer question, for a maximum of 62 points. A record repairing every initially wrong item can still complete. This is appropriate formative practice, but does not show an unaided mechanism explanation. Even a correct *first* choice may follow a free hint or the notebook; it is not a pretest. Randomized option positions reduce position cues, not recognition or answer elimination.

**Change.** Retain repairs and the friendly language. Report three separate quantities: first response accuracy with help exposure recorded; successful repairs; instructor-scored explanation quality. Keep any competition on an identical assigned set. Do not combine optional dossiers into a common comparative grade: the displayed first-correct denominator grows as a player opens more items. Add a four-part rule trace before selected answers: target office → governing procedure → applicable fact/threshold → consequence. Score its reasoning separately. Label the passport “practice completed” until an independent assessment is scored.

### 2. High: the final “transfer” question restates the definition almost verbatim

**Displayed evidence.** The unfamiliar republic has an elected fixed-term president and a PM/cabinet removed by legislative no confidence; the correct answer is “Semi-presidential: elected fixed-term President plus a cabinet responsible to the assembly.” A systems comparison is available immediately above it (`TRANSFER` and `finalScreen`, lines 102–103). The subsequent explanation asks for comparison through origin and survival plus one case.

**Implication.** This is useful recognition in a new place name. It cannot distinguish learning the relationship from matching the supplied definition. It contains no ambiguous title, appointment/confirmation distinction, constrained dismissal power, change of target, or exceptional procedure requiring transfer beyond the taught template.

**Change.** Keep this item as an accessible formative check. Add a separately locked, unfamiliar constitutional excerpt with two executive offices and a changed procedure. Require a prediction, its legal mechanism, and one missing fact that would alter the inference. Include a counterexample: a head called President elected by the assembly and removable through confidence, or a cabinet coordinator who is removable by the president but has no parliamentary confidence relationship. Ask about bill passage, executive office, and the legislature separately. Do not display the family label or answer choices until the first explanation is locked. Use the instrument below for the learning claim.

### 3. High: country stamps do not require country-specific explanations

**Displayed evidence.** The country screen asks multiple choice; the “Passport note (optional)” appears only after the dossier is complete and asks, “What does this case add to the general pattern?” (`countryCase`, line 100). The optional twist-help drawer contains the correct answer’s feedback. The field sheet asks for “one case that complicates a general rule” and the office, procedure, threshold, and result in both moves (Worksheet, p. 2).

**Implication.** A stamped country records eventual correct choices. It does not show whether the learner can explain the contrast that makes Germany different from the UK, Taiwan different from France, or a Chilean minister’s exceptional removal different from routine confidence. Asking for one written country example does not establish breadth across all nine.

**Change.** Preserve all eighteen moves and add a compact first-attempt trace to each: office; procedure; threshold/condition where relevant; direct consequence. This can be short structured fields, not eighteen essays. Require three deeper paired explanations, approximately 50–80 words each: UK–Germany, France–Taiwan, and US–Chile, with a specific rule difference and a political consequence. Rotate or add South Africa–Portugal–Brazil contrasts across assignments. Require each student, including team members, to write an individual final transfer answer. Keep free help in practice, but record whether help or the correct-feedback drawer was opened before the first response. Do not mistake this assisted trace for the independent assessment.

### 4. High: the toy model needs a required bridge back to law and political behavior

**Displayed evidence.** The formation agreement is followed by fixed votes: “Later 48–52 and 58–42 roll-calls are held constant for comparison; members may realign” (line 92). The map explains that same seats and fixed shocks make the contrast visible. In the optional Midnight Session, satisfaction is two points for a preferred priority plus concession tokens; the least satisfied party faces the dispute, with ties assigned to the larger party. The comparison holds final bill and confidence support constant across rulebooks, assuming an executive is in office (`power-lab.js`, `ASSUMPTIONS`, `shockFor`, `compare`).

**Implication.** The fixed exercise deliberately disconnects the bargain from later votes; it isolates a legal consequence. The lab adds choice-sensitive arithmetic but still assumes identical political support under different institutions. It also turns all seats outside the continuing confidence agreement into affirmative challenge votes (`100 - confidenceSeats`), with no abstention or uncertain tolerance. None of these is an estimated law of party behavior. The text acknowledges fictional assumptions, but a player need not explain what those assumptions permit or exclude. “Same crisis, three constitutions” must not be read as an identified real-country effect or an institutionally realistic equilibrium comparison.

**Change.** Make the model-to-country bridge a required short debrief for the assessed route. Require two lab runs differing in exactly one planned input, with the expected change in bill support and confidence support written *before* launch. Then hold that political result constant to compare the three survival rules. Ask: “Which result followed from a constitutional rule, which from our invented bargaining rule, and which might change because real parties respond to institutions?” Require one actual case that invalidates a simplification: UK minority toleration, German constructive confidence, French Article 49(3), Portuguese abstentions, or Taiwan’s absence of initial investiture. Retain the transparent arithmetic, all country moves, and the explicit fictional label.

### 5. High: conditional tradeoff reasoning is supplied before it is elicited

**Displayed evidence.** Each chosen move immediately displays a condition and its benefit/cost, followed by “Who gains, who bears a cost, and under what condition?” Class mode requires ten words (`tradeoffLab`, line 121; `next-dilemma`, line 210). The underlying instruction rightly says to grade mechanism, cost, and condition rather than preference (`tradeoffs.json`).

**Implication.** The written response can simply repeat the displayed effect. The three dilemmas communicate appropriate qualifications, but a ten-word response cannot certify that a student can generate a conditional argument. The no-universally-best framing is a strength; it does not itself show understanding of when the chosen mechanism works.

**Change.** First capture a short benefit–cost–condition explanation before showing the supplied interpretation; preserve the original and the revision. Then change one fact: a previously cohesive majority becomes divided; an additional chamber’s consent is required; an alternative majority disappears; or a president’s dismissal power is constrained. Ask whether the same move remains defensible and why. Assess a relevant mechanism, a stated beneficiary and cost bearer, a binding condition, and a consequence of changing that condition. Accept different policy preferences. Use the shared analytic rubric below; do not award automatic correctness for the selected policy.

### 6. High: the research desk requests components, but not an identification strategy

**Displayed evidence.** The four fields ask for a hypothesis, comparative test, rival explanation, and challenging observation. The example proposes comparing constructive-confidence settings while “considering party fragmentation and majority status.” The boundary statement says the dossiers “illustrate rules; nine selected countries are not a controlled causal test” (line 122). The data’s research starters propose time to replacement, policy delay, electoral timing, and voter blame.

**Implication.** A response listing Germany, the UK, a mechanism, and several confounders can satisfy all word gates without explaining why the comparison separates institutional rules from parties, history, or selection into those rules. Selecting only governments that faced a successful motion also excludes failed challenges and confounds the opportunity for replacement with the legal procedure. Cabinet replacement frequency, time to a successor, policy delay, and democracy survival have different units and risk sets. The simulator’s fixed political inputs demonstrate an implication by construction; they do not solve that observational problem.

**Change.** Add a concise design card with: (a) X and a distinct Y; (b) units, observation window, and eligible cases; (c) indicators and when they are measured; (d) comparison and why selected independently of Y; (e) the most serious rival process and how it would generate the same observations; (f) one feasible design response plus its remaining limitation; and (g) a concrete challenging observation. For a confidence hypothesis, include unsuccessful challenges and governments that could have faced one; define caretaker days separately from policy delay. A UK–Germany comparison can illustrate plausibility but must state its limited causal identification. A within-country rule change, matched episodes, or a broader panel is a proposal to investigate, not automatically a solution: reform adoption and changing party competition may be endogenous. A randomized information vignette could identify an effect of information on respondents’ attribution, not the effect of presidentialism on a country’s democratic survival.

The supplied primary research usefully supports keeping mechanisms conditional. Cheibub, Przeworski, and Saiegh explicitly investigate presidential coalitions and legislative success rather than assuming minority presidents cannot govern [R10]. Tsebelis defines required veto actors and distinguishes policy stability from government and regime stability [R11]. Use these as rival-mechanism readings, not as universal empirical rankings.

### 7. Medium: “first attempt” preservation is uneven across the final evidence

**Code/display evidence.** Prediction and saved reflection firsts are preserved (`power-engine.js`, lines 89–113). The first submitted research proposal preserves all four fields in `researchRecord`, with later versions appended (line 216). However, `caseNotes` and `tradeNotes` are overwritten on input, while tradeoff decisions preserve `firstIndex` but not the first reasoning. The readable report shows current research fields; after revision it prints the first hypothesis and a revision count, leaving the first comparative test, rival, and falsifier in the JSON backup (line 138).

**Implication.** This is substantially better than overwriting all practice answers. Nevertheless, the claimed record of original reasoning is incomplete for country notes and tradeoffs. A grader reading only the report cannot inspect the original full research design. A first *submitted* explanation is also different from an explanation made before feedback. Local backups and starting new records are appropriate for practice but do not establish secure, identity-verified examination records.

**Change.** Preserve every assessed first response as an append-only object with item/version, exact prompt, scenario seed, answer, explanation, timestamp, phase, route, and help exposure; append revisions under the same ID. Show the full first and latest research proposal and all first tradeoff reasoning in the readable export. Keep practice saves convenient. For evaluation, lock the first submission before feedback and archive the assessment export before revisions; do not use the latest repaired value as the baseline or posttest score. This is a measurement requirement, not a recommendation to build high-security exam software.

### 8. Medium: correct one legal attribution and make subtype caveats actionable

**Concrete correction.** The Portugal challenge says “Apply Article 194’s requirement of an absolute majority” and its distractor feedback again attributes the threshold to Article 194 (`cases.json`, lines 540 and 553; mirrored in the combined data). Article 194 governs the motion’s initiative and timing; Article **195(1)(f)** specifies the absolute majority of all members in office that makes passage entail resignation. The answer—48 yes, 40 no, 12 abstentions fails in the fictional 100-member chamber—is correct [R8]. Change the scenario to “Apply Articles 194 and 195(1)(f)” and the feedback to “Article 195(1)(f) requires…” Keep Article 195(2)’s distinct constrained presidential removal rule.

**Category caveat.** The named minimal definition is defensible [R12]. Still, the opening labels and generic rulebook cannot tell the player whether a semi-presidential PM is also removable by the president. That is precisely a central subtype distinction in Elgie [R12]. France’s resignation-linked Article 8, Portugal’s constrained Article 195(2), and Taiwan’s presidential appointment/removal provisions should be juxtaposed as a required “same family, different powers” contrast, without asserting all three share an at-will dismissal rule. The game already avoids that false generalization; make the difference part of the student’s explanation, not only a notebook caveat.

**Source use.** Keep primary articles available, with the exact operative clause and rule version identified. The official Chile text independently supports both ordinary ministerial appointment/removal and exceptional Senate removal [R6]; an old access-failure note need not be the only path students have to that content. The Taiwan timing challenge is sound: the presidential ten-day window runs from passage of no confidence, not the later resignation [R9]. The France Article 49(3) case is also sound under the stated assumption of valid use [R7]. Avoid expanding this to “any bill at any time.”

### 9. High: the full learning claim requires all-nine coverage and independent evaluation

**Displayed evidence.** The default is “Core: 3 rulebooks + 4 countries”; the other five are bonus dossiers. Extended assigns nine. The source and guide explicitly call the 65–85 and 80–100 minute durations “design estimates,” and the optional lab adds 8–12 minutes. Instructor controls may release subsets, so completion follows the released route, not inherently all eighteen moves.

**Implication.** Core can responsibly teach the common distinctions, but it cannot provide evidence that a learner encountered all nine cases or their within-family variation. Team completion also cannot establish every member’s understanding. Nothing supplied establishes that the expanded writing and research work fits the estimated time, or that repaired recognition transfers to new constitutional sources.

**Change.** Retain both paths, but use the full nine-country/eighteen-move assignment whenever assessing the stated breadth. Export the exact assigned and completed item set; mark incomplete coverage explicitly. Use two class sessions plus separate assessment time as an initial plan, then measure completion burden without lowering the content requirement. Rotate operator, analyst, and skeptical reporter, then assess each student individually. Pilot the protocol below. Until results exist, describe the game as designed to support these goals; do not say that it improves them or that simulated AI reviews establish classroom effectiveness.

## Audit of all nine cases and eighteen moves

These are validations of the specified rule applications under their stated hypothetical conditions, not claims that these nine illustrate every institutional arrangement.

| Country | Move 1: mechanism retained | Move 2: real contrast retained | Source judgment |
|---|---|---|---|
| United Kingdom | Confidence-capable coalition may govern although not the largest party. | Outside confidence/supply support can sustain a minority cabinet after an ordinary non-confidence bill defeat. | Supported by official parliamentary explanations [R1]. Ordinary defeat is not the same as clear loss of confidence. |
| Germany | Article 67 needs an all-member majority for a successor, not merely against the incumbent. | Article 68 opens a conditional, nonautomatic dissolution route within 21 days; losing it does not itself dismiss the chancellor. | Supported [R2]. The scenario expressly assumes the other constitutional requirements. |
| South Africa | Section 102(2) targets President and Cabinet, including deputy ministers. | Section 102(1) excludes the President and requires cabinet reconstitution. | Supported [R3]. This is a valuable title-based misclassification counterexample. |
| United States | House rejection of a bill does not terminate the presidency or create dissolution power. | House impeachment alone does not remove; conviction is a separate Senate step. | Supported [R4]. Keep the stipulated absence of other departures. |
| Brazil | Coalition rupture affects legislative support, not routine presidential survival. | Presidential ministerial dismissal does not command the departed party’s legislative votes. | Article 84(I) and the separate constitutional structure support the authority distinction [R5]. Negotiating support is a plausible response, not a guaranteed successful outcome. |
| Chile | A coordinating minister remains subject to presidential ordinary appointment/removal authority. | A valid Senate guilty finding in the distinct accusation procedure removes the accused minister. | Supported by Articles 32(7), 33, and 52–53 [R6]. Do not extend a minister’s removal automatically to the President. |
| France | Opposing presidential and Assembly mandates can coexist through cohabitation. | Valid finance-bill use of Article 49(3) changes the link between censure and passage; 48/100 censure votes fail. | Supported by Articles 8, 49, and 50 [R7]. The procedure and all-member denominator, not ordinary bill arithmetic, do the work. |
| Portugal | Article 195(2) restricts presidential removal even if the President dislikes policy. | 48/100 affirmative votes do not meet the all-member censure threshold despite 40 against and 12 abstentions. | Answers supported [R8]; change the threshold attribution from 194 alone to 195(1)(f). |
| Taiwan | Lack of investiture does not preclude later binding no confidence against the premier. | Dissolution on day 16 is outside the ten-day period measured from passage, even if resignation was on day 8. | Supported by Additional Articles 2–3 [R9]. The use of the Additional Articles rather than superseded main provisions is important. |

## Proposed pre/post and transfer instrument

This is a proposed instrument requiring piloting; it is not a validated scale. It is deliberately separate from passport points and retains the full game assignment.

### Administration and immutable records

1. **T0 before the tutorial:** individual parallel-form assessment, approximately 20–25 minutes initially. No game hints, notebook, feedback, or answer choices for the first explanation. Necessary constitutional excerpts are included so interpretation is not confused with memorizing article numbers. Record prior familiarity; do not infer it from a first game answer.
2. **Practice:** all three rulebooks, all nine countries and eighteen moves, the three tradeoffs, research design, and two planned lab comparisons. Free hints and repairs remain available. Team roles may rotate.
3. **T1 after practice:** an equivalent alternate form with new names, numbers, offices, and ordering; first answers locked before feedback. Counterbalance A→B and B→A across students to help detect form difficulty. Match conceptual demand, source length, and denominators, not just superficial phrasing.
4. **T2 after 7–14 days:** three unfamiliar transfer tasks, individually, with no previously displayed feedback. Record interval and intervening instruction. Feedback and further learning can follow the locked submission.

Store `assessment_id`, `form_version`, `item_id`, exact prompt/excerpt, `phase`, student pseudonym, assigned coverage, `first_answer`, `first_explanation`, confidence if collected, help exposure, and submission time. Firsts cannot be overwritten through the assessment interface; repairs form separate events. Archive the first-attempt export before feedback. Timestamps support sequence, not speed grades. A correct guessed choice earns result credit but no invented explanation credit.

### Eight parallel-form tasks and answer criteria

| Task | Form A example | Parallel Form B | Required evidence/key |
|---|---|---|---|
| 1. Origin and survival | Three unnamed countries: an Assembly-elected President removable by ordinary confidence; a separately elected fixed-term President whose ministers answer to that President; a popularly elected President plus PM/cabinet collectively responsible to the Assembly. Classify and draw/describe each selection and survival link. | Replace titles and institutional names; put the same relationships in another order, including “Chancellor” or “Cabinet Coordinator.” | Respect the stipulated relationships: parliamentary, presidential, semi-presidential under the named minimal definition. A title alone is not the explanation. |
| 2. Two votes, distinct outcomes | In a supplied parliamentary rule, an ordinary bill loses 46–54. Later a valid cabinet no-confidence motion passes 56–44. State bill, cabinet, head-of-state, and assembly outcomes after each. | Ordinary bill loses 44–56; valid no confidence passes 54–46; change party labels. | Bill blocked without automatic dismissal; separate confidence motion requires the cabinet consequence; do not invent head-of-state removal or an automatic election. State the supplied replacement/dissolution conditions. |
| 3. Target and procedure | An excerpt gives one motion against Cabinet excluding President and another against President and Cabinet. Apply each to the same vote. Explain why consequences differ. | Reverse order and use new office names with equivalent clauses. | Correct target and exact consequences; no inference from the title President. This tests the South African lesson without asking students to repeat its country label. |
| 4. Denominator and successor | A constructive rule needs a majority of all 100 members for a successor. Fifty-five oppose the incumbent, but 49 support a replacement. | 120 members; 67 oppose the incumbent, 60 support a replacement, remaining members abstain or oppose. | No constructive replacement yet. A needs 51 successor votes, B needs 61. Explain why an anti-incumbent majority or majority of votes cast is insufficient. |
| 5. Clock and appointment | An unfamiliar dual-executive text permits appointment without investiture; later confidence can remove the PM. A dissolution decision is allowed within ten days of passage. PM resigns on day 8; President acts on day 16. | PM resigns on day 6; President acts on day 14; equivalent text and changed event names. | Appointment does not imply immunity. Both proposed dissolutions are late. Identify the event starting the clock and distinguish resignation deadline from dissolution authority. |
| 6. Conditional tradeoff | A cohesive majority can revise flood funding with an audit concession; an alternative cabinet majority may be possible. Defend a response, name a beneficiary and cost bearer, then reassess when the alternative majority disappears. | Presidential legislature offers a narrower bill; reassess when a required second chamber opposes it. Provide the relevant powers in the prompt. | More than a preference or slogan: mechanism, benefit, cost, binding condition, and changed inference. Opposite preferences can receive equal reasoning credit. |
| 7. Research design | A researcher claims constructive confidence reduces *policy delay* after noticing fewer successful removals in two selected countries. Repair the design. | A researcher claims separate presidential terms improve *democratic survival* after observing fewer cabinet changes in selected cases. | Define distinct X/Y, units/window and indicators; identify outcome mismatch and selection; propose a comparison and rival process; state a feasible design response, remaining limit, and observable challenge. No causal conclusion follows from these observations alone. |
| 8. Unfamiliar source and model limits | Supply a short constitution in which an ordinary bill is not confidence, but a finance-bill procedure deems passage unless an all-member censure majority succeeds. A 100-seat chamber has 48 yes, 40 no, 12 abstentions. Ask bill, cabinet, President, and assembly outcomes; ask whether the toy model predicts future public satisfaction. | Equivalent unfamiliar text; 120 seats, 60 yes, 45 no, 15 abstentions. Replace policy and institutional names. | Both censure motions fail (51/61 required); the procedural bill is deemed passed; no resignation/election from that failed motion. Distinguish this procedure from an ordinary affirmative bill vote. Public satisfaction is not identified by the simulator; give one missing behavioral or contextual factor. |

Tasks 3–5 cover different kinds of legal fine print; they should not be collapsed into one arithmetic score. Tasks 6–8 assess generating and qualifying an explanation, not recognizing a family name. Exact constitutional passages must be checked and fixed in the item version before administration.

### Self-explanation rubric

Use this analytic rubric on the mechanism/application/transfer explanations and adapt its consequence column explicitly for tradeoffs. Report result accuracy separately from explanation quality. Do not score length as understanding.

| Dimension | 0 | 1 | 2 |
|---|---|---|---|
| Relevant rule or mechanism | Missing, wrong relationship, or title-based assertion. | Correct general label with incomplete target/procedure. | Correct origin/survival relationship or causal mechanism, including the relevant office and procedure. |
| Use of scenario evidence | No relevant fact, or wrong denominator/clock. | Relevant fact cited but not connected fully to the rule. | Correctly applies decisive facts, threshold, clock, source clause, or political condition. |
| Consequence and separation | Wrong effect or conflates bill, office, assembly, and democracy. | Main effect correct but omits another affected institution/outcome; tradeoff names only gain or cost. | Correct sequence and distinct outcomes; for tradeoffs names both benefit and cost and who bears them. |
| Boundary or counterfactual | Universal ranking, unsupported certainty, or irrelevant caveat. | Plausible condition without explaining its effect. | Specifies a change or missing condition and explains how it changes or limits the inference. |

An 8/8 answer to task 4A could say: “Forty-nine successor votes do not meet the 51 required from all members. Opposition to the incumbent alone cannot replace the chancellor. If 51 agree on a successor, constructive replacement becomes available; it does not itself require an assembly election.” A correct classification with no reasoning earns its result mark, not eight explanation points. A no-confidence explanation that incorrectly removes the separately elected President loses the relevant rule and consequence credit even if its prose is polished.

Score research design separately, 0–2 on each of six dimensions: defined X and distinct Y; mechanism; units/time window and observable measures; defensible case/period selection and comparison; rival causal process plus design response; concrete challenging evidence plus an acknowledged limit. **0** is absent or materially wrong, **1** is present but underspecified, **2** is specific and internally coherent. Maximum 12. Naming a confounder without showing how it creates the pattern earns at most 1 on that dimension. “Anything could happen” is not challenging evidence; an explicit incompatible pattern under specified comparable conditions can be.

Use at least two raters on a prespecified sample, blind to assessment phase where possible; calibrate with low/middle/high anchor answers and report agreement or disagreement by dimension. Do not treat the proposed cut scores as validated proficiency thresholds. Revise unclear items after piloting, while versioning prompts and retaining original records.

### Delayed/far transfer and evaluation inference

Use three new source-based tasks at T2: (1) a dual executive with constrained presidential dismissal and no initial investiture, requiring a post-motion office map; (2) a changed bill-confidence procedure and an abstention pattern, requiring both a prediction and a one-fact counterfactual; (3) an institutional reform claim requiring a proposed comparison, selection warning, and one observable rival-mechanism test. Include at least one “insufficient information” subquestion so naming the missing power or procedure can be the correct answer. The prompts should be checked independently so uncertainty is genuine rather than an accidental omission.

Report first-attempt accuracy and rubric scores by goal, error type, route/coverage, and help exposure. Include missing assessments and the number of students contributing to each comparison. A pre/post difference is evidence of change during the teaching interval, not by itself a causal effect of the game: exposure, testing, instructor explanation, and form difficulty also change. If estimating an instructional effect, compare game use with the same constitutional content and instructional time delivered without the game, ideally through randomized assignment of feasible groups, with comparable facilitation and individual outcome assessment. Predefine the primary outcome as explanation quality on unfamiliar cases, not passport completion. Delayed transfer, first-attempt reasoning, and research-design quality matter more than repairing familiar multiple choice items.

## Independently checked primary/official references

- **R1. UK Parliament / House of Commons Library:** [What is a hung Parliament?](https://www.parliament.uk/about/how/elections-and-voting/general/hung-parliament/) and [Commanding confidence](https://commonslibrary.parliament.uk/research-briefings/cbp-10915/). Confidence, minority arrangements, ordinary defeats, and resignation/dissolution conventions.
- **R2. German Federal Ministry of Justice:** [Basic Law, Articles 63, 67–69](https://www.gesetze-im-internet.de/englisch_gg/englisch_gg.html). Successor majority and distinct confidence/dissolution procedure.
- **R3. South African Department of Justice:** [Constitution Chapter 5, sections 86–87 and 102](https://www.justice.gov.za/constitution/chp05.html). Election of President and the two distinct confidence targets.
- **R4. US National Archives:** [Constitution, Articles I–II](https://www.archives.gov/founding-docs/constitution-transcript). Separate presidential term, legislative process, House accusation, Senate trial/conviction and removal.
- **R5. Brazilian Federal Senate:** [Constitution, Article 84(I) and presidential provisions](https://legis.senado.leg.br/norma/579494/publicacao/33296461). Presidential ministerial appointment/dismissal and separate institutions.
- **R6. Chilean National Congress Library:** [Constitution, Articles 32(7), 33, 52–53](https://www.bcn.cl/leychile/navegar?idNorma=242302). Presidential ministerial authority, coordination, constitutional accusation, and removal following Senate guilt.
- **R7. Élysée:** [Constitution of 4 October 1958, Articles 8, 49–50](https://www.elysee.fr/en/french-presidency/constitution-of-4-october-1958). Appointment/resignation and the Article 49(3) finance-bill mechanism.
- **R8. Portuguese Parliament:** [Constitution, Articles 194–195](https://www.parlamento.pt/sites/EN/Parliament/Documents/Constitution7th.pdf), PDF pages 75–76 in one-based navigation. The all-member no-confidence threshold is Article 195(1)(f); constrained presidential removal is 195(2). [Official Portuguese text](https://www.parlamento.pt/Legislacao/Paginas/ConstituicaorepublicaPortuguesa.aspx) independently agrees.
- **R9. Taiwan Office of the President:** [Additional Articles 2–3](https://english.president.gov.tw/Page/95). Appointment, no confidence, resignation and dissolution clock; supersession of old Articles 55 and 57.
- **R10. Cheibub, Przeworski & Saiegh (2004):** [Government Coalitions and Legislative Success Under Presidentialism and Parliamentarism](https://doi.org/10.1017/S0007123404000195), *British Journal of Political Science* 34(4), 565–587. Primary empirical study; not a finding generated by this game.
- **R11. Tsebelis (1995):** [Decision Making in Political Systems: Veto Players in Presidentialism, Parliamentarism, Multicameralism and Multipartyism](https://doi.org/10.1017/S0007123400007225), *British Journal of Political Science* 25(3), 289–325; [author-hosted full paper](https://sites.lsa.umich.edu/tsebelis/wp-content/uploads/sites/246/2015/03/decision_making_in_political_systems_1995_bjps.pdf). Required consent actors, policy differences, and distinct stability outcomes.
- **R12. Elgie (2011):** [Semi-Presidentialism: Sub-Types and Democratic Performance](https://academic.oup.com/book/2211). Minimal relationship definition and premier-presidential versus president-parliamentary subtypes. [Elgie & McMenamin (2011)](https://journals.sagepub.com/doi/10.1111/j.1467-9248.2010.00870.x) is a primary example of specifying and testing conditions associated with cohabitation.

**Lock statement:** This review was completed independently from the supplied round-two input and the public sources above. No peer review was read before or after drafting. Recommendations are faculty design judgments; no learning gains or real participant reactions are asserted.

# Independent blind teaching and interaction review

**Status:** LOCKED independent review, 29 October 2026. No peer review, earlier feedback, or original repository was consulted. No product files were changed.

**Judgment:** The game has a strong conceptual spine: hold a political situation constant, separate executive origin from survival, then test the generalization against country-specific rules. The nine countries and their eighteen moves contribute real instructional variation and should all remain required. The main weakness is that operating the interface is better scaffolded than constructing and testing an explanation. Completing the current prediction sequence demonstrates successful recognition with available help; it does not by itself demonstrate independent causal reasoning.

## Method and limits

I reviewed only the frozen input packet: the standalone `The_Power_Circuit.html`, its corresponding UI, engine, art, lab, CSS and JSON sources, and the student guide and worksheet. I traced the full nine-country path, the stage gates, answer-specific feedback, written-response gates, final stations and report. I checked the standalone bundle's embedded data: it uses the three family-specific tradeoff scenarios from `tradeoffs.json`, rather than the fallback scenarios in the UI source. I extracted both student PDFs and visually inspected the rendered first worksheet page.

This is a **source-informed novice cognitive walkthrough**, not actual playtesting. I modeled a learner with no comparative-politics vocabulary and reasonable general reading/numeracy. The predicted misunderstandings below are hypotheses. A local browser attempt could not launch because its Chromium executable was absent. I therefore read the UI's rendered-string templates and state logic; I did not verify browser layout, scrolling, focus behavior, screen-reader output, touch use, time on task or learning gains. Country rules were evaluated as teaching content supplied by the packet, not independently revalidated against current law. The 80–100-minute full-route duration is a design estimate, not observed evidence.

Primary evidence locators: `power-ui.js` functions `home`, `tutorial`, `map`, `formationTask`, `policyTask`, `crisisTask`, `resolutionTask`, `reflectionTask`, `countryCase`, `finalScreen`, `researchDesk`, `recordBody`, `labScreen`; `power-art.js` functions `circuit` and `country`; `power-engine.js` decision records, stage reducers and `score`; all nine entries in `cases.json`.

## What is already worth preserving

- The same 48–52 bill defeat and 58–42 formal challenge isolate the institutional contrast. The distinction between policy success, cabinet survival and legislature survival is explicit.
- The country set challenges useful shortcuts: President is not a classification; coalitions are not exclusively parliamentary; appointment does not determine survival; an exceptional removal procedure is not ordinary confidence.
- No countdown, dragging or sound is required. Hints are free. Points do not reward ideology, office retention or speed. First answers and repairs remain visible.
- The diagram labels election, formation/appointment and confidence separately. It includes an outgoing cabinet and an unchanged assembly; this is more instructional than a generic picture of government buildings.
- The research task distinguishes a hypothesis from a finding and explicitly names rival explanations and limits of the selected cases. Preserve that rigor.

## Modeled novice walkthrough

| Stage | What the learner can see or do | Likely learning issue / implication |
|---|---|---|
| Home | “Same political crisis. Three sets of rules. Who stays in office?”; default Core with four countries; Extended with nine; named outcomes include a passport and research idea. | The central question is clear. A novice cannot yet interpret mandate, confidence, cabinet or investiture. For the required nine-country product, make the complete 9/18 path the normal route and manage its load through resumable sessions. |
| Warm-up | “Place the mandate token” then “Cut the confidence wire”; both actions have one prescribed button. A modal states “Cabinet out. Assembly still here.” | This demonstrates a result but does not establish whether the learner distinguishes the offices. “Mandate token” risks becoming an unexplained interface metaphor. A tiny prediction before the cut would expose the target misconception. |
| First formation | Choose the origin answer, negotiate parties, count seats and six fictional commitment slots, choose a PM, install the cabinet. | Several new concepts arrive together. The two offered commitments for each party have the same mechanical effect. That can suggest consequential political strategy without supplying evidence that the choice matters. Explain the role of the bargain as illustrating formation, then separate it visibly from the controlled votes. |
| Ordinary bill | The screen supplies 48 yes / 52 no and says no confidence condition is attached. Correct prediction gives the consequence before “Run the vote and watch the board.” | The distinction is well stated. Because the correct option and feedback reveal the outcome before the vote, the later board change mainly confirms text. Learners can move on without inspecting which institutional link changed or did not change. |
| Formal challenge | A new 58–42 roll-call; either valid no confidence or ordinary censure; a target-specific correct answer. | “Read the motion” is exactly the right habit. However, extreme distractors such as every seat becoming vacant permit elimination without a model of removal. Ask for the target and rule, not just the most qualified sentence. |
| Resolve and explain | Replace the cabinet using the same assembly, or continue the separate presidency; then a prose explanation. | Replacement concretizes continuity. The resolution distractor “Delete every assembly seat” is too easy to eliminate. Class-mode word counts ensure text exists, not that the explanation uses an event, rule and consequence. |
| Next two circuits | Repeat the five-stage structure under presidential then semi-presidential rules; prior circuits can be reviewed. | Repetition creates practice, but the important contrast is separated across long sequences. A short persistent comparison record would help learners compare rather than remember whole screens. Fade support in the later circuits while retaining the full task. |
| Optional Midnight Session | Accessible from the first route map; constitution, allies, priority, six slots, three tokens, preferences, satisfaction, shock and bill-vote arithmetic. | It adds meaningful agency, but introduces many dependencies before novices have stabilized the two central relationships. The route text says “then” live bargaining, while the entry is already prominent. Recommend it after the three circuits, with a guided first run and the full calculation available. |
| Countries | A nine-card grid, family labels, two moves each, rules in a drawer; a correct first move unlocks the twist; after any first-move decision the rules drawer opens. | The task is navigable and the source help is appropriate. But two correct options can earn a stamp without the learner stating what changed between the moves. Optional passport notes are not collected into a working evidence structure. |
| Final comparison | An unfamiliar republic essentially restates the semi-presidential definition, followed by one comparison response. | This is useful transfer practice, but a weak independent mastery test: the relationship is supplied almost verbatim and other families are not independently tested here. Add a short rule-based justification or a contrasting unfamiliar case. |
| Tradeoffs and research | Three family-specific dilemmas, then four fields: hypothesis, test, rival, challenging evidence. One possible research example is inside optional help. | The conceptual move from identifying legal consequences to designing comparative research is substantial. An evidence bank and a worked-to-independent bridge are needed. The current word minima cannot assess whether variables or falsifiers are meaningful. |
| Passport | Prediction points, first answers, repairs, written work, boundaries and export instructions. | A useful auditable record. Treat it as a portfolio for human review. A stamp or repaired answer should not be presented as proof of retained understanding. |

## Eight prioritized changes

### 1. Make a small, persistent reasoning record the spine of the game — highest priority

Use the same compact record throughout: **office → origin → removal procedure → threshold/conditions → consequence**. During the first circuit, work one row aloud in plain text: “The assembly's support enables this cabinet to enter office. A successful formal no-confidence vote removes this cabinet. It does not remove the assembly.” Define cabinet, assembly, confidence and mandate beside their first use, with a short concrete example.

Let the learner complete the next row; in the last circuit, ask them to complete it before expanding the explanation. Keep all substantive decisions and explanations. After every circuit, show the three-system record with the completed column alongside the next blank column. This reduces memory demands while increasing the need to explain a relationship.

**Success evidence:** In a later unfamiliar case, the learner can identify the office and procedure without relying on a title or family label.

### 2. Turn prediction, action and feedback into one coherent learning event — highest priority

Currently a correct multiple-choice answer reveals the outcome, then unlocks a separate application button. Instead, have learners predict two or three outputs—bill result, executive offices affected, assembly/election result—then **commit once**. Reveal a brief event receipt beside the diagram: what was voted on, the applicable rule, what changed and what stayed.

For an error, connect the feedback to the precise false inference: “You treated a rejected bill as a cabinet-removal procedure. This bill has no confidence condition.” Highlight the relevant link and invite a repair plus a brief reason. Keep first attempts, repairs and unlimited free help. Use plausible, parallel distractors; replace choices such as “Delete every assembly seat” with competing legal continuations that require the target rule to distinguish.

**Success evidence:** A repaired answer includes a corrected rule, rather than merely a newly selected option.

### 3. Clarify the purpose of agency and sequence the strategy lab after the model

Place a persistent label on the main circuits: “Controlled comparison: your agreement forms the cabinet; these later roll-calls are fixed.” When the majority bargain is replaced by 48–52 or 58–42, explicitly show **new roll-call, not coalition-seat arithmetic**. The small existing caveat is correct but bears too much explanatory work.

Introduce the lab as the next application after all three circuits. For a first run, use one visible example configuration, ask the learner to predict the effect of changing one input, then allow free configuration. The complete lab, replay and exact arithmetic remain available. Separate confidence seats and bill votes in the ledger with words, labels and grouping, not color alone. Explain why the main circuit's two offered policy promises are equivalent there, or remove that cosmetic subchoice while retaining the coalition decision.

**Success evidence:** Learners can explain both why an ally helps a bill and why losing that ally has different office consequences across systems.

### 4. Keep all nine countries and eighteen moves; organize their comparison explicitly

Use three resumable sets, with no skipped dossier: **formation and support** (UK, Brazil, Taiwan), **targets and exceptional procedures** (South Africa, US, Chile), **conditions, thresholds and continuation** (Germany, France, Portugal). Other orderings are possible; the point is to give the learner a comparison question for each set rather than merely nine destinations.

Each dossier should preserve Move 1 and Move 2 in a compact paired view. Before the second prediction ask “Which fact changed?”—targeted office, procedure, denominator, condition or clock. After both moves, save the case's office/procedure/threshold/result into the evidence record. At the end of each three-country set, require one short comparison using two cases. Do not add nine long essays. The existing optional passport note can become a brief interpretation linked to the paired facts.

Make the full 9/18 route canonical; retain pause/resume and session checkpoints to manage workload. A four-country finish or instructor release that bypasses cases would not satisfy this scope.

**Success evidence:** Final responses cite a specific country rule to limit a generalization, with all eighteen moves retained in the record.

### 5. Make country visuals explain the changing rule

Keep the existing live circuit as the primary visual language. Extend it with small exact explanatory graphics rather than more generic political collages:

- **Germany:** a successor-support meter at 47/51 for Article 67, then a separate conditional Article 68 route. Do not visually merge the procedures.
- **South Africa:** target outlines contrasting President + Cabinet with Cabinet excluding President.
- **France:** an Article 49(3) decision flow showing that the censure threshold controls the bill consequence; contrast it with the earlier ordinary bill.
- **Portugal:** a 100-member bar with 48 yes, 40 no and 12 abstentions; mark 51 as the threshold so the denominator is visible.
- **Taiwan:** one timeline anchored at the no-confidence vote: day 0, resignation on day 8, day 10 limit, proposed dissolution on day 16. It directly teaches why resignation does not restart the clock.

The figures should update for the twist and include equivalent text. Retain flags for recognition. A map can orient the nine countries and support a purposeful filter such as “Compare countries with a President”; location alone should not imply institutional causes. If no comparison task uses a map, it is optional orientation art. Use the final screen's repeated collage space for the learner's accumulated evidence instead. No invented democracy, stability or effectiveness scores are needed.

**Success evidence:** The learner can point to the visual feature that establishes a threshold, target or time origin, not simply identify the country.

### 6. Bridge country reasoning into the final research task

Before the four research fields, let the learner select two saved observations and distinguish **what the rules allow** from **what empirical outcomes might follow**. Work one example of that transition: Germany requires agreement on a successor; this suggests a hypothesis about cabinet replacement under specified party conditions, not a finding that Germany is more stable.

Offer a small planning frame for X, Y, mechanism, comparison and rival. Provide one completed example and one partially completed example, then require an independent proposal. Link a candidate indicator to the chosen outcome: days to successor installation is different from bill delay and democratic survival. Ask what observation would contradict the proposed mechanism. Keep all four research fields and all three tradeoff dilemmas.

**Success evidence:** The final claim has an observable outcome, a mechanism, a plausible comparison, and evidence that could actually challenge it.

### 7. Assess explanations separately from completion and points

Retain the transparent two-point first prediction / one-point repair scheme if desired, and retain its non-ideological framing. Present three distinct records: completion, prediction history and human-assessed explanation quality. Free hints and rules are instructional supports; a correct response after consulting them is legitimate practice, but its status differs from a later unaided transfer check.

Add a short transfer check after the country sets, with country/family labels concealed initially, then let students consult sources for revision. Score the explanation with the rubric below, not a new automated semantic grader. Give instructors or peers a concrete prompt: identify the claimed rule, evidence and one missing condition. Preserve the original explanation and the revision.

**Success evidence:** A learner who guesses or copies a qualified option can be distinguished from one who explains and transfers the mechanism. Word counts remain completion checks only.

### 8. Give the same instructional cues across screen sizes and access modes

The source already contains focus management, modal trapping, a skip link, semantic buttons, SVG descriptions, status messages, text labels and reduced-motion CSS. These are useful provisions, not verified accessibility compliance.

On mobile the task precedes the board, and “watch the board” may require scrolling past the question. Put the event receipt and a text “changed / unchanged” summary immediately after the committed prediction, plus a clear jump to the diagram and back. Supply an accessible structured relationship list beside every instructional diagram; the country's mechanism plates are currently `aria-hidden`, so any additional teaching content in them must also appear in text. Prefer a focused read-aloud control for the current task or result over reading an entire long main region. Test 200% zoom, narrow screens, keyboard and a screen reader with actual users. Use plain labels before specialist terms and allow explanations to be composed in the worksheet or through an appropriate classroom accommodation without reducing conceptual criteria.

**Success evidence:** A learner can complete the same prediction–inspection–explanation loop without depending on side-by-side layout, color, motion, sound or fine text in an SVG.

## A manageable redesign

This does not require a new game engine or fewer cases. Reuse the existing stage machine, saved first attempts, country JSON, diagram components and export record. Add four bounded pieces: (1) the reusable office/rule/consequence record, (2) a prediction-and-result receipt component, (3) a paired country-move/evidence view with the five targeted mechanism graphics above, and (4) a final research planner and human rubric. Stage the display of existing controls and explanations rather than deleting content.

Use resumable checkpoints after the three circuits and after each three-country set. The next session should open with one short recall prompt and the evidence record, then continue the same full route. Preserve the three complete rulebooks, all nine dossiers/eighteen moves, the three tradeoffs and the four-part research task. Calibrate time estimates only after observing learners; do not promise a shorter completion time from this review.

## Learning outcomes and a compact human rubric

Use this alongside, not inside, the prediction-point total. Each row earns 0–2; the maximum is 10. Treat any substitution of bill defeat for removal, title for relationship, or automatic election for a conditional procedure as requiring revision even if the total is otherwise high.

| Outcome | 0 — not yet demonstrated | 1 — partial | 2 — demonstrated |
|---|---|---|---|
| Trace origin and survival | Uses titles or conflates offices. | Names a relationship but misses one office or condition. | Correctly traces executive origin and continuing survival, distinguishing President, PM/cabinet and legislature where relevant. |
| Apply a legal mechanism | Gives an outcome without the rule, or uses the wrong procedure. | Identifies the procedure but misses target, threshold or timing. | Uses the stated target, threshold/conditions and consequence, and distinguishes office survival from policy passage. |
| Compare and qualify | Ranks systems universally or repeats a family label. | States a general contrast with a vague example. | Uses two specific country observations to explain both a pattern and an exception or limit. |
| Explain a conditional tradeoff | Gives a preference only. | Names a benefit and cost without a mechanism or condition. | Explains who benefits, who bears a cost, and a condition under which the proposed benefit may fail. |
| Design a testable investigation | Presents a slogan or a game result as a causal finding. | Specifies some variables or evidence but weakly separates outcome, mechanism and rival. | Defines X and a distinct observable Y, a mechanism, comparison/indicators, a plausible rival and an observation that would challenge the claim. |

For validation, observe novices explaining their next move, then use a brief unfamiliar-case check and a delayed return to one threshold or target distinction. Watch for whether they inspect the board, distinguish the two counters, explain a repair, and reuse case evidence. Those observations would test the design hypotheses in this review; the source packet alone cannot establish them.

# Independent blind review: graphic and program design

Locked 2026-10-29. Reviewed only the frozen files in `blind-round-2/input/`. No repository, history, previous reports, or peer reviews were consulted. No product files were changed.

Evidence: source inspection plus isolated Node module/VM probes against the frozen engine, lab, and UI. **No browser interaction, screenshot review, screen-reader test, print rendering, or visual device test was performed.** Responsive findings below are consequences of the CSS/markup, not claims from a rendered UI. The standalone HTML's engine, art, and lab script bodies match their supplied source files.

The foundation is strong: the dark switchboard gives institutional relationships a coherent visual language; election, appointment, and confidence are separately labeled; outgoing cabinets are distinct from surviving assemblies; the lab explicitly separates policy votes from confidence support. The deterministic lab validator replays evidence rather than trusting reported totals. Nine countries each retain an incident and a twist: 9 cases / 18 country questions. Keep that scope and the explicit fictional-model boundaries.

## 1. P1 — A record authored within the permitted field limits can create a backup that cannot be restored

**Pointers:** `power-ui.js:18–54`, `:183–184` (`download`, `backup`), `:216`, `:231` (proposal revisions and restore file limit); `power-engine.js:23`, `:105–110`, `:249–332`.

`save()` persists without a record-size check, and backup downloads the entire record. Restore rejects files above 600,000 bytes; `readSave()` also rejects compact serialization above 600,000 characters. Yet allowed field lengths and revision counts can exceed that threshold through ordinary authoring. The engine also copies full reflection text into its journal, amplifying growth. Pretty-printed backup bytes can exceed the file limit before the compact character limit is reached.

**Probe:** 16 permitted 12,000-character final explanations, plus 16 permitted research proposals with four 6,000-character fields each, produced an engine that passed `E.validState()`, an 809,713-character wrapper, and an 813,873-byte downloaded representation. `PowerCircuit.validateSave()` rejected that wrapper. This is a boundary-case probe, not a claim about typical student response lengths.

**Improve:** define one UTF-8 byte budget for serialized backups, align save/export/import limits, and ensure legal authoring cannot silently cross it. Avoid duplicating full explanation bodies in journal details. If pruning is necessary, preserve the first submission, current submission, explicit archive counts, and the declared recent-history window. Surface size/recovery problems before promising a successful backup. Add one meaningful round-trip test at the largest supported record size.

## 2. P2 — Imported lab predictions can falsely claim a correct answer

**Pointers:** `power-ui.js:51–53`, `:152–162` (`labScreen`, `labRecord`, `labAction`); `power-lab.js:56–60`, `:89–103`.

The canonical lab state is carefully replay-validated, but the outer `labGuesses` validator checks only answer enumeration, boolean type, timestamp, and revision count. It does not recompute correctness or connect each run key to retained lab evidence. The UI trusts the imported `correct` flag both to reveal the comparison and to describe the first prediction in the report.

**Probe:** a canonical completed parliamentary run plus a first answer `{answer:'president-out', correct:true, ...}` passed `PowerCircuit.validateSave()`. This answer cannot be correct under any of the lab's three rulebooks. By comparison, tampering with the lab's actual result total was rejected by `L.validState()`.

**Improve:** derive correctness from each retained run's canonical configuration and comparison, and reject guesses without corresponding retained evidence. Use the same pure answer checker in `labAction()` and import validation. This is audit consistency for an educational record, not a claim that a local game should provide exam security.

## 3. P2 — Core save validation does not enforce the bargaining model it presents

**Pointers:** `power-engine.js:115–150`, `:183–197`, `:268–305`; `power-ui.js:20–54`, `:90–95`, `:201–207`.

The UI enforces six commitment slots, but the engine's formation/replacement transitions and validator do not. The outer save validator validates individual bargain strings without cross-checking their costs against coalition membership. It also allows a formed core government to omit formation evidence entirely. An accepted backup can therefore display an agreement exceeding its own stated rules or lose the original formation record used in review.

**Probe:** forming all four parties (42+31+17+10 seats, costs 2+3+2+1 = **8 slots**, against the displayed limit of 6) passed both `E.validState()` and `PowerCircuit.validateSave()`; the wrapper was on the policy scene. A formed Civic+Greens government with `formation:null` also passed.

**Improve:** centralize commitment costs and limits, enforce them in formation/replacement transitions and import, and require/cross-check original formation evidence for formed records. Keep a separate explicit rule for the Presidential case: legislative allies are optional for holding the presidency, while the fictional bargaining-slot constraint still applies to selected agreements.

## 4. P2 — Starting a replay removes completed lab evidence from readable exports

**Pointers:** `power-ui.js:137–143` (`recordBody`, `reportScreen`), `:161` (`labRecord`); `power-lab.js:63`, `:73`, `:77`.

`labRecord()` returns an empty string whenever the current run lacks `result`. `REPLAY` deliberately resets that result while preserving `firstRun` and bounded `history`. Thus a learner who completes a run and starts the encouraged replay gets a readable/printed passport with no Midnight Session section, although completed evidence still exists in the backup. The pending replay also receives no status entry.

**Improve:** export the preserved first completed run, recent completed history, and the current run's status separately. Build all of them from one evidence-summary function and keep unrevealed predictions marked pending. A readable report should retain the experiment that motivated the student's comparison.

## 5. P2 — On small screens, actions and the evidence they change are separated; diagram text also shrinks sharply

**Pointers:** `power.css:10–15` (`.scene`, `.lab-layout`, `.live-circuit-art`, SVG text classes, mobile ordering); `power-ui.js:82`, `:91–95`, `:152–162`; `power-art.js:22–52`.

At widths ≤760px, CSS puts the entire task before the board, and the entire lab task before its ledger. Coalition building, hints, and choices can make these long sections. A successful action re-renders or moves to the top of the next scene, while the changed circuit is below the task. This weakens the promised “run the vote and watch the board” connection on phones. The CSS also unconditionally hides `.mobile-move` at the end of the revised rules.

The SVG remains a fixed 360-unit drawing. At a 320px viewport, a roughly 260px inner board width scales 12-unit office detail text to about 9px, and the 7.4-unit bottom slogan to roughly 5px. Country plate figure labels start at 7.7 units and are even smaller in multi-column cards. These are calculated sizes, not rendered measurements.

**Improve:** keep a compact, persistent or inline evidence strip adjacent to the current action: bill tally, confidence tally, cabinet state, President state, assembly state. Place a small circuit before the prediction, with the full circuit expandable. Move essential text out of SVG into readable HTML; remove tiny slogans or make them redundant decoration. Verify at 320/390px and 200% zoom with keyboard focus preserved and no toolbar/diagram overflow.

## 6. P2 — The live circuit does not visually carry party identity or the cohabitation state

**Pointers:** `power-ui.js:82` (`board` passes party data and separately prints the cohabitation notice); `power-art.js:22–52` (`circuit`); `power-engine.js:62–83` (`snapshot`).

`board()` supplies `presidentParty`, `pmParty`, and cabinet status to the artwork, but `circuit()` does not use those party fields. Every President, assembly, and cabinet keeps its institutional color and generic office label. A Civic PM beside the Workers President therefore looks identical to a Workers PM beside that President; only a separate text box identifies cohabitation. This misses a valuable consequence of the learner's leadership choice.

**Improve:** preserve stable institutional colors, then add explicit text badges such as “Workers President” and “Civic PM,” and a labeled “different political camps” annotation when the engine's cohabitation condition holds. Use party chips consistently with the coalition builder, without relying on color alone. Update the SVG description or adjacent HTML state summary with the same information.

## 7. P2 — The central matched comparison remains three blocks of prose rather than a shared visual experiment

**Pointers:** `power-ui.js:152–162` (`labScreen`, `.lab-comparison`), `:101` (`comparisonTable`), `:138` (`recordBody`); `power.css:11`, `:13` (`.lab-comparison`).

The lab's strongest teaching mechanism is “same political inputs, different legal consequences,” but its reveal uses three vertically stacked text cards, each repeating the vote totals. The systems table is mostly prose as well. Learners must mentally align numbers and office outcomes across sections.

**Improve:** add a compact comparison matrix with one shared input row and three aligned institutional rows: bill passed/blocked, PM/cabinet in/out, elected President in/stays/not present, assembly stays, automatic election no. Show bill and motion counts as clearly labeled 0–100 bars with a visible **51-vote** threshold and a table/text equivalent. Use the actual `result.comparison` values or the explicitly fictional 48–52 / 58–42 exercise values. Do not convert nine selected countries into empirical performance scores, rankings, or causal charts.

## 8. P2 — Round explanation revisions are supported in the engine but unreachable from the finished-round interface

**Pointers:** `power-ui.js:91`, `:96`, `:145–148`, `:199`, `:207`; `power-engine.js:105–110`, `:200–210`.

`reflectionTask()` says a saved explanation can be saved again as a revision and contains a “Save revision” branch. But `round()` routes every completed round to `reviewTask()`, whose reflection view is read-only. Re-entering the finished round likewise selects that read-only route. A learner who saved a blank optional explanation cannot later add reasoning through the visible flow, although reflection revisions and their export are implemented.

**Improve:** add a distinct “Revise explanation” action in completed-round review. Present the latest explanation, append a revision through `REFLECT`, and retain the first explanation and prediction record. This supports reflection after the country cases without reopening graded choices or changing the matched exercise.

## Geographic and flag opportunity — useful, but lower priority than the issues above

**Pointers:** `power-ui.js:99–100`, `:164–177`; `power-art.js:53–65`.

The nine flags already have country-specific accessible names, and dossier codes/titles give the cards identity. Avoid adding decorative flags as the main visual upgrade. A small optional world overview on the dossier screen could instead answer “where are these nine selected cases?” and provide another route into the same cards. Use authoritative, offline-bundled geographic geometry and verified coordinates, labeled point locations, and a complete text list. Explain that these are selected teaching cases; do not color the whole world by system family, imply regional representativeness, invent comparative country data, or treat the dossier drawings as maps. Preserve the 9 cases / 18 questions and the current constitutional distinctions.

## Suggested implementation order and verification

1. Repair backup round-tripping and semantic validation, using bounded state probes.
2. Preserve completed replay evidence in readable exports and make round explanation revision reachable.
3. Put live evidence beside mobile actions; add party identity and a shared comparison graphic using existing state.
4. Only then consider a geographic overview or more ornamental illustration.

Verify one complete offline journey, restore/export at boundary sizes, a replay-in-progress export, a corrected lab prediction import, and the smallest supported viewport. Include keyboard/modal focus and a rendered print check. The present review does not claim those UI checks have already passed or failed.

# Fresh blind student-perspective acceptance review

**Decision: not yet accepted for the promised entry-to-exit learning record.** Two reachable interface paths make the timing of that record less reliable than its labels imply. The full expedition contains all **nine countries and eighteen moves**; these findings do not call for removing or shortening them.

This is a source-informed learner walkthrough, not observation of real students or a live browser test. I read only the supplied acceptance-input materials: the player interface and teaching/learning strings, country and tradeoff materials, and the Student Guide and Student Worksheet. I checked the bundled HTML's player data and interface against the supplied sources. I did not inspect power-engine.js, audit constitutional accuracy, edit the product, consult other reviews, or recruit participants.

## Remaining findings

### 1. High: an entry record can still be called a complete pre-play baseline after guided practice

**Learner path:** start a new record, leave the first entry question through **Tools + help → Report / print**, use **Route map → Parliamentary → Enter this rulebook**, answer the executive-origin prediction and see its feedback, then return through **Report / print → Resume entry check**. The entry check can also remain offered while using the optional strategy lab before forming the first parliamentary cabinet.

The report offers Resume entry while the tutorial is still at its initial step and the parliamentary cabinet has not been formed. Leaving for the map, round or lab does not seal the entry baseline; only navigating to the tutorial seals it in the navigation helper. Consequently, a student can receive explanatory practice feedback and then complete responses that the report labels as a complete entry baseline. The learner is following available navigation, rather than editing a save or bypassing a control.

**Evidence:** power-ui.js:66, 70, 77, 198–201, 278–281, 290. The Student Guide, p. 1, describes the entry checks as occurring before the tutorial; the report says a complete entry record is a pre-play baseline and that initial items precede model feedback.

**Why it matters:** “Entry” can mean different exposure histories without the student realizing it. Retaining the first answer is insufficient if the first answer was collected after teaching. Report timestamps do not supply a learner-facing warning or exposure classification.

**Acceptance change:** seal the entry status when the learner first enters teaching or guided practice through any available route. Preserve already locked entry responses as partial, and clearly explain that remaining entry responses cannot be added after practice begins. A preview or report visit alone need not seal it.

### 2. High: the exit check is offered before the advertised final learning activities

After completing the circuits, assigned countries and three cross-country comparisons, a learner can open Report / print and select **Continue the six-question exit check** before doing the final systems comparison, three tradeoffs, or research proposal. Both the checkpoint button and the start check use the earlier prerequisites. The ordinary end-of-research route does launch the exit check after those activities, but it is not the only visible route.

**Evidence:** power-ui.js:105 and 132–137 describe three final writing stations followed by an exit check; 218–222 implement that normal sequence. The alternative report path is in 278 and 290. The Student Guide, p. 1, likewise lists comparison, tradeoffs and research before the six exit checks.

**Why it matters:** exit item 6 asks for a measurable comparison, rival explanation and challenging evidence (power-learning.js:122–129), while the substantial research-design teaching appears at the later Research Desk (power-ui.js:124–129, 137; power-teaching.js:194–203). Students can therefore take an “exit” before receiving teaching directly relevant to that question. The six first responses are then immutable; later completion of the research work cannot produce a new first exit response.

**Acceptance change:** offer the exit check only after the required final writing stations and submitted research proposal are complete. If an earlier checkpoint is intentionally supported, label it as an earlier checkpoint and retain a distinct genuine exit administration. Do not silently reuse its first responses as the post-expedition exit.

### 3. Medium: new-record route and session instructions disagree across surfaces

The home screen offers a new **Instructor short route: 4 countries; all 9 remain available**, while the Student Guide introduces the full nine-country route and says that **older saved Core records** may retain a four-country assignment. A new learner can choose the short route without understanding whether it is an authorized current assignment or only legacy compatibility. The full-route interface asks learners to plan three resumable sessions; the Guide, p. 1, calls a two-session plan a starting estimate.

**Evidence:** power-ui.js:79, 99, 101, 139–140, 265; Student Guide, pp. 1–2. The UI does identify Core versus Extended in the report, so this is an assignment-expectation mismatch, not a claim that the five remaining dossiers are absent.

**Acceptance change:** state one consistent assignment rule in setup and the Guide. If new short assignments are intentional, describe them explicitly and distinguish their completion from the complete nine-country/eighteen-move expedition. Keep all nine cases and both moves in each. Use the same session estimate everywhere and continue to describe it as unpiloted.

### 4. Medium: the screen asks less of the tradeoff explanation than the worksheet does

The screen's writing label asks **“Who gains, who bears a cost, and under what condition?”** That can be satisfied literally by listing a beneficiary, cost bearer and a condition. It does not directly ask the student to say when the choice fails or how changing the condition would change their judgment. Those extra reasoning steps are explicit in the Student Guide, p. 3, and Worksheet, p. 6; the tradeoff data also asks under what condition the move would fail, but that full prompt is not rendered by tradeoffLab().

**Evidence:** power-ui.js:106 and 123; tradeoffs.json, each reflection_prompt; Student Guide, p. 3; Student Worksheet, p. 6. The human rubric expects a changed condition and an explanation of its effect (power-teaching.js:207–211).

**Likely novice response:** “Residents gain, supporters lose part of their proposal, and the parties must cooperate.” This names the requested elements without explaining why changed cooperation would alter the student's choice.

**Acceptance change:** add the missing explicit instruction to each screen: name a condition under which the move could fail, then explain how changing it would change the student's judgment. This improves the explanation demand without adding another activity or grading a political preference.

## What the learner is asked to explain

| Point in the route | Explanation requested | What the resulting evidence can support |
| --- | --- | --- |
| Three fictional circuits | Executive origin; ordinary bill defeat versus formal challenge; who leaves, stays and whether an election follows | Supported application of the model; first predictions and repairs are guided practice |
| Nine country pairs / eighteen moves | Office, procedure, controlling threshold or clock, consequence and what changed between moves | A teach-back after answers, rule traces and debriefs are available; it is supported explanation |
| Three saved cross-country notes | UK/Brazil support relationships; target and procedure across South Africa/US/Chile; conditions across Germany/France/Portugal/Taiwan | Ability to compare specific rules rather than repeat a family label |
| Final comparison and tradeoffs | Origin and survival across families, a country complication, beneficiary, cost bearer and a conditional judgment | A written argument requiring human review; the counterfactual instruction needs the alignment above |
| Research proposal | Distinct explanatory variable and outcome, mechanism, units/periods/indicators, comparison, rival and challenging observation | A proposed investigation, not a causal finding from game runs |
| Six exit responses | Apply supplied unfamiliar rules; distinguish policy, offices and legislature; calculate a threshold, trace a clock and challenge a causal claim | An open-response transfer record requiring human assessment; timing needs findings 1–2 fixed |

## Check and revision controls

The question screen tells students that locking a first response preserves it, asks for an explanation and decisive clause or uncertainty, and labels confidence and reported help as ungraded. It permits honest uncertainty at entry. Item-specific human criteria stay hidden until all six exit first responses are collected; return criteria remain hidden until all six return first responses are collected. These are comprehensible distinctions between completion and understanding.

The report supplies **Revise this explanation** after all six first responses in that form exist. Revisions retain the first response and ask why the student changed or kept the reasoning. The Guide states a maximum of three later revisions per item. The on-screen revision form should repeat that limit and show the remaining allowance; currently the button simply disappears after the third revision (power-ui.js:279–281). It should also state plainly that entry model criteria are deferred until the exit form is complete, since “lock all six before feedback” does not by itself explain that longer wait.

The optional return check can be taken immediately, but both the screen and Guide explicitly say that immediate use does not demonstrate retention. The Guide and report also distinguish missing entry data from zero and say that points, stamps and word counts do not automatically assess understanding. These limitations should remain. No claim about engagement, timing feasibility, scoring reliability or actual learning gain can be accepted from this source walkthrough alone.

# Fresh blind AI faculty-style acceptance review

Locked review date: 2026-10-29. Verdict: **ACCEPT for the requested political-content and assessment-honesty source review. No concrete blocker found.** No correction to a keyed country outcome is required.

## Scope and independence

I am an AI reviewer applying comparative-politics and assessment criteria, not a human faculty member or constitutional-law authority. This is a source review, not a human expert endorsement, classroom evaluation, psychometric validation, or browser/runtime acceptance test. It establishes no real learning, engagement, timing, or retention result.

I read only the supplied `acceptance-input/` snapshot and primary official/academic sources needed to verify its claims. I did not consult repository history, earlier reviews, peer findings, or other agents. I made no input edits, installed no browser or automation tooling, and did not run the application. Diagram findings concern the meaning of the authored SVG/HTML and surrounding text, not rendered usability or accessibility.

Reviewed content includes `cases.json`, `tradeoffs.json`, the teaching, learning, art, engine and UI source, the delivered `The_Power_Circuit.html`, and the extracted text of the Student Guide and Worksheet. I also inspected the assessment and country-key text of the Instructor Guide embedded inside the delivered HTML. The delivered country content matches the JSON; its additional source records support the tradeoff material. Embedded teaching/learning/engine/art/lab scripts match the supplied source. UI differences found concern packaged image/PDF assets.

## Nine cases and eighteen moves

All eighteen country choices have a defensible keyed outcome on the supplied facts. The conditions below are essential to those judgments; they are not optional details.

| Case and source location | Move 1 | Move 2 | Acceptance reasoning |
|---|---|---|---|
| UK — `cases.json:29` | Coral can govern with Green's stated confidence agreement, despite Amber being the largest party. | The 43-seat cabinet can retain office with 12 outside confidence-and-supply supporters after the transport defeat. | The second bill is expressly not a confidence matter and the agreement continues. Outside supporters need not enter Cabinet. The text correctly permits a replacement government without an automatic election. [UK] |
| Germany — `cases.json:114` | 47 successor votes cannot meet 51 of all 100, notwithstanding 55 opponents of the Chancellor. | A valid failed Article 68 confidence request and Chancellor's proposal make discretionary presidential dissolution available within 21 days. | Article 67 requires agreement on the successor. The second move explicitly assumes the other constitutional requirements; it does not present any lost policy vote as sufficient. The successor-election lapse is included. [DE] |
| South Africa — `cases.json:197` | A valid section 102(2) majority motion requires the President, other Cabinet members and deputy ministers to resign. | A section 102(1) motion excluding the President requires Cabinet reconstitution by the President. | These are different targets and consequences. Neither motion itself dissolves the Assembly. Assembly election and confidence dependence justify the stated parliamentary relationship despite the presidential title. [ZA] |
| United States — `cases.json:280` | House rejection blocks the bill and does not end the presidency. | House impeachment without Senate conviction does not remove the President. | The exclusions of other departures are explicit. Senate conviction requires two thirds present; impeachment and ordinary disagreement are not substituted for conviction. The text does not give the President a dissolution power. [US] |
| Brazil — `cases.json:367` | Coalition departure does not itself end the presidency. | Ordinary ministerial dismissal can coexist with loss of the dismissed minister's party's bill support. | Article 84(I) appointment/removal authority is distinguished from control of congressional votes. Exceptional responsibility/criminal proceedings remain separate. No general authority to bypass Congress is inferred from an electoral mandate. [BR] |
| Chile — `cases.json:450` | The President retains ordinary ministerial appointment/removal power after an Article 52(1)(a) critical agreement. | A valid Senate guilty finding following constitutional accusation removes the accused minister. | Article 32(7) and Article 33 support the coordinator premise. Articles 52–53 distinguish ordinary scrutiny from accusation/judgment; a minister's conviction uses a majority of senators in office. The President is not the accused. [CL] |
| France — `cases.json:538` | An Amber President and Coral-supported PM can coexist under cohabitation. | On the valid Article 49(3) finance-bill premise, 48 censure votes fail against the 51-member threshold; the bill is considered passed. | Appointment does not override Assembly responsibility. Only votes for censure count; failure to take an ordinary affirmative bill vote is not decisive under 49(3). Article 8 is properly distinguished from an unrestricted dismissal power. [FR] |
| Portugal — `cases.json:621` | Policy dislike alone does not establish unrestricted presidential Government dismissal. | 48 for, 40 against and 12 abstentions do not reach 51 of all 100. | Article 195(1)(f), not a votes-cast majority, controls the resignation consequence. Article 195(2) separately requires institutional necessity and prior Council of State consultation. The text preserves that constrained power rather than denying it. [PT] |
| Taiwan — `cases.json:704` | Appointment without investiture does not cancel binding later no confidence; the premier must tender resignation within ten days. | Proposed dissolution on day 16 is outside the ten-day period from the motion's passage. | Additional Articles 2–3 supply the operative rules. Resignation/request does not restart the clock. Presidential survival, required consultation, and the martial-law/emergency-decree bar are stated. The formal premier-removal contrast with France is supported. [TW] |

The country families are used as selection/survival relationships under an explicitly identified minimal definition of semi-presidentialism. The classification note recognizes competing definitions; neither titles nor party count nor democratic quality stand in for a relationship test. Elgie's publisher abstract supports the elected-President plus legislature-responsible PM/cabinet definition: https://academic.oup.com/book/2211.

## Diagrams and teaching criteria

`power-art.js` marks the general circuits as Port Meridian's fictional rulebooks. Its election, appointment and confidence wires carry different meanings. Cutting cabinet confidence leaves the elected assembly and, where applicable, separately elected President in place. Caretaker status is presented as part of the fictional model, not as an identical rule for all countries.

The mechanism graphics in `power-teaching.js:96` preserve the decisive distinctions: Germany counts successor support; Portugal retains abstainers in the all-member denominator; France labels the 48 as censure votes; South Africa contrasts the two targets; Taiwan starts both displayed periods at the motion. None changes the keyed result through an unlabeled reversal of the vote's object or clock.

The matched circuits disclose their fixed 48–52 and 58–42 shocks. The Midnight Session discloses the assumption that every seat outside continuing support votes for the challenge and that the same support is held constant across rulebooks. These are coherent controlled exercises, not observations of actual institutional behavior. Country cases, minority governments, abstentions, and conditional dissolution supply meaningful boundaries rather than being erased by the model.

The explanation rubric (`power-teaching.js:206`) separately checks office/procedure, decisive facts, consequence, and a boundary/counterfactual. The research rubric distinguishes outcomes, mechanisms, eligible units, selection/comparison, rivals/design response, and challenging evidence. The embedded Instructor Guide tells scorers not to double-count the 0–4 item criteria and 0–8 explanation profile, to withhold rule/consequence credit for removing the wrong office, and to pilot scoring before setting proficiency cutoffs. The tradeoff choices are conditional possibilities and do not grade a political preference.

## Eighteen open-response rule cards

I judged these as stated fictional rules, not as attempts to reproduce the constitutions of named countries. In particular, Return 4's votes-cast rule and Exit 5's twelve-day/seven-day split are deliberate premise changes, not country errors.

| Construct | Entry A | Exit B | Return C |
|---|---|---|---|
| Bill versus office | 43–47 ordinary bill fails; continuing cabinet and assembly remain. | 38–42 ordinary bill fails; continuing agreement preserves cabinet. | 34 of 70 censure votes fail to reach 36; the expressly linked finance bill becomes law. |
| Origin versus title | Assembly-elected President with confidence dependence: parliamentary relationship. | Legislature-selected Chief President with confidence dependence: same relationship. | Separately elected President and a non-confidence-responsible coordinating PM: presidential relationship despite the second title. |
| Actual target | PM/cabinet resign despite no initial investiture; President and assembly remain. | Later confidence responsibility controls despite investiture; President and assembly remain. | Procedure 2 requires Cabinet reconstitution and excludes presidential resignation. |
| Successor and denominator | 43 of 84 required; 41 fails regardless of 48 opponents. | 61 of 120 required; 59 fails regardless of 70 opponents. | 80 votes for/against; 41 is a majority and elects the successor. |
| Clock and other conditions | Day 16 is outside the ten-day period from passage. | Day 10 exceeds the seven-day dissolution period despite timely resignation under a separate twelve-day rule. | Day 6 meets timing and a request exists, but missing consultation still prevents dissolution. |
| Inference | Office survival does not establish faster relief delivery. | A successor requirement does not prove universally less policy delay or democratic protection. | Reduced replacements, unchanged delay, and a simultaneous electoral reform do not establish improved democratic survival. |

The item-specific criteria in `power-learning.js:31`, `:81`, and `:131` match these implications. They ask for applied clauses and consequences, not just labels. They do not import an unstated real-country deadline into a fictional card.

## Completion and demonstrated understanding

The distinction is honest in the reviewed source. `power-ui.js:139` uses COMPLETE for a portfolio workflow: required supported predictions, country traces, comparisons, tradeoff/research entries and six exit responses. `caseComplete` at `power-ui.js:264` adds an office/rule/consequence trace, but only enforces presence and word minima. It does not inspect meaning. `power-learning.js:345` counts six recorded first responses, and its summary expressly rejects interpreting completion/confidence/help as semantic scores or empirical gain.

The interface and reports say written reasoning awaits human assessment, points describe guided practice, and portfolio completion is not proof of learning. The Student Guide makes the same distinction. First responses and later revisions are kept separately; answer criteria are withheld until the required first-response sequence is complete. Missing/partial entry work is not silently reconstructed as a baseline. Return checks are optional, timestamped, and described as immediate transfer if taken immediately. The forms are described as proposed and unequated, with scoring reliability requiring a pilot.

Consequently a passport can be complete while its explanations are wrong or weak. That is compatible with the wording actually presented. Acceptance here does not certify the explanations as correct, the forms as equally difficult, an assessment as independent/proctored, or a causal learning effect.

## Primary official verification URLs

These sources were read for the case implications above. No mandatory factual correction emerged. South Africa's Justice HTML request timed out; the official government chapter and matching official Justice search excerpts supplied the pertinent clauses.

- [UK] https://commonslibrary.parliament.uk/research-briefings/cbp-10915/ and https://www.parliament.uk/about/how/elections-and-voting/general/hung-parliament/
- [DE] https://www.gesetze-im-internet.de/englisch_gg/englisch_gg.html and current German Article 68: https://www.gesetze-im-internet.de/gg/art_68.html
- [ZA] https://www.gov.za/documents/constitution/constitution-republic-south-africa-1996-chapter-5-president-and-national and https://www.justice.gov.za/constitution/chp05.html
- [US] https://www.archives.gov/founding-docs/constitution-transcript and https://www.archives.gov/electoral-college/provisions
- [BR] https://legis.senado.gov.br/norma/579494/publicacao/33296461
- [CL] https://www.bcn.cl/leychile/navegar?idNorma=242302
- [FR] https://www.elysee.fr/en/french-presidency/constitution-of-4-october-1958
- [PT] https://www.parlamento.pt/sites/EN/Parliament/Documents/Constitution7th.pdf
- [TW] https://english.president.gov.tw/Page/95

## Locked snapshot identity

SHA-256:

- `The_Power_Circuit.html`: `757ed319253e228291c47d33bc00be07ad8ac81df733215adf1001cd595dd253`
- `cases.json`: `ec8fc3a6c190d6187d34679e41ec965004370c1375b72ce3d56ef35047ff37b3`
- `power-learning.js`: `3d966c9b223bc2c7a142c0c186cc5c6b7d11b6b2cfb6cb2b9b1bbf7a31694b0f`
- `power-teaching.js`: `2f32071c6a09a64752202bd1c5010c0ef0c6cda57fd42fb1536d61dcf2c3a0bb`

This report is locked as the fresh review. It will not be revised after peer findings are available.

# Student-perspective fix verification

**Result: both high-priority timing paths from the blind review are closed for newly created records.** One compatibility caveat remains for records saved before the timing safeguards existed. The original blind acceptance report is unchanged.

This verification reads only the current power-ui.js at `/workspace/sites/yoav-duman-recover/resource-sources/power-circuit/power-ui.js`. It is a source-reachability review, not a live UI test, constitutional audit or student study. No product files were edited. Student PDF session-copy alignment is outside this verification.

Reviewed source SHA-256: `c0c7a90be566b472bb7683ca91275e238f7d2399cd271bea2af269baafaaeeee`.

## Timing fixes

| Original path | Current source behavior | Verification |
| --- | --- | --- |
| Entry → Report → Route map → guided origin feedback → Report → Resume entry | navigate() calls beginInstruction() on the map and each teaching/practice screen (line 70). beginInstruction() records instructionStartedAt and seals an offered entry check, preserving any prior responses as partial (265). The resume button requires no instructionStartedAt (294), and startLearning('entry') independently rejects after instruction starts (282). | Closed for new records before the learner can reach the guided feedback. |
| Entry → Report → optional strategy lab → Resume entry | The lab is included in the same navigation seal, as are tutorial, round, country, case and final screens (70). Notebook, How to play and Instructor access also invoke beginInstruction() (196, 199). | Closed for new records. Report/backup access by itself leaves the untouched baseline available. |
| Circuits + countries + cross-country notes → Report → exit, before final comparison/tradeoffs/research | Both the report's exit button and startLearning() use practiceComplete() (282, 294). That function requires the circuits and predictions, assigned countries, required comparisons, final transfer, saved final explanation, all tradeoff decisions and mode-appropriate reasoning, and a submitted research record matching the current research fields (139). | Closed. Merely reaching the country/synthesis endpoint no longer opens exit. |
| Normal completion of Research Desk → exit | finish-game records the research submission and then calls startLearning('exit') (219–223), which applies the same practiceComplete() prerequisite. | The ordinary completion route now agrees with the report route. |

The current entry-close message explains that remaining entry responses cannot be added once teaching begins. Entry criteria are explicitly described as deferred until the entire exit form is complete (283). The report now withholds the general reasoning rubric until exit completion as well as withholding item criteria (285).

## Other reviewed changes

- **Nine-country scope:** new setup offers only the complete nine-country/eighteen-move expedition (79); createGame() releases all country cases (66). The setup no longer invites a new learner to select the four-country route. Existing/instructor route handling remains separate.
- **Tradeoff explanation:** the writing label now explicitly asks when the move could fail and how changing that condition would change the student's judgment (123). This resolves the previously weaker screen instruction.
- **Revision allowance:** the revision view now states the maximum of three and dynamically shows how many revisions remain (283). First responses and revision explanations remain distinct in the report (285).

These are source confirmations of the requested controls and copy. They do not establish actual learning, classroom timing or usability with students.

## Remaining compatibility caveat: older records have unverified timing

readSave() accepts learning records without instructionStartedAt; the validation checks that timestamp only when present (20–56, 249–251). The Resume saved record action renders the accepted record directly rather than passing its current screen through beginInstruction() (197). No timing migration or “legacy timing unverified” report qualification is visible in the reviewed source.

A genuine older save can therefore still contain the earlier problem state: entryStatus is offered, tutorial is at its initial step, guided practice has already occurred, and instructionStartedAt is absent. If that save resumes at the report, learningControls() can offer Resume entry and startLearning('entry') accepts it (282, 294). Similarly, an exit form collected early by the prior interface remains stored and labeled Exit explanation even if the final research work happens later; the new prerequisite prevents a new early launch but does not reclassify previously collected responses.

**Recommended handling:** preserve these historical responses, but mark their administration timing as unverified. Where saved practice evidence proves that instruction already occurred, close remaining entry questions on import/resume. Do not silently recertify an older complete entry or exit form as correctly timed. A report qualification is preferable to reconstructing exposure dates that were never recorded.

This caveat does not reopen either verified fresh-record path. It limits claims about restored records created under the earlier interface.

# Student verification of legacy timing handling

**Result: the legacy timing-integrity caveat and the restored-entry-screen case are addressed. No remaining defect was identified in these bounded source changes.** The blind acceptance report and first fix verification remain unchanged.

This is a bounded source review of the current `/workspace/sites/yoav-duman-recover/resource-sources/power-circuit/power-ui.js`, not a live browser test, broad re-review or student study. No product files were edited.

Reviewed source SHA-256: `19f81a12f880df126274ab4a358e28f9011456ea148ea249190676af0f1fcb35`.

## Confirmed changes

| Bounded change | Source result |
| --- | --- |
| Identify older learning records | readSave() assigns learningTimingUnverified to an unversioned learning record before returning it. It does not manufacture a historical instruction date. |
| Stop completing an exposed older entry check | When an older offered entry has instructionEvidence(), normalization seals it through Q.skipEntry. Evidence includes tutorial progress, guided decisions, a formed circuit executive, or a launched/replayed lab. Entry controls and startLearning() also reject instruction evidence even without a timestamp. |
| State the limit honestly | learningRecord() prominently says “Imported timing unverified” and tells readers to treat entry/exit placement as unknown rather than a verified pre-play baseline or post-expedition exit. Its complete-entry sentence also retains the timing qualification. This resolves the risk of silently recertifying old responses. |
| Distinguish fresh records | createGame() initializes learningTimingVersion: 2 and learningTimingUnverified: false. Existing instruction-start sealing remains in navigation. |
| Check the new administration order on import | Version-2 validation rejects an exit record without a submitted research record or whose first stored exit response predates the first submitted proposal. It rejects a return response when the exit form is incomplete. startLearning() still requires completed practice and, for return, a completed exit. |
| Remove a stale report entry control after instructional help | How to play, Instructor and Notebook actions now seal, save and rerender the underlying screen before opening their modal. The report's entry-button condition therefore reflects the newly closed state. |
| Avoid rendering an entry form that normalization has closed | Immediately after legacy timing normalization, readSave() changes a saved learning/entry scene with status skipped or partial to the report and clears learningRevision (line 56). It leaves the saved first responses, revisions and drafts intact. |

The earlier source-confirmed fresh-record entry and exit safeguards remain present. This check does not establish that older response timing was valid; the explicit unknown classification is the appropriate outcome.

## Restored-entry path verified

An authentic save from the earlier interface could be paused on an entry question after this previously available path: Entry → Report → guided practice → Report → Resume entry. Such a record can have screen `learning`, learningForm `entry`, entryStatus `offered`, no timing version, and a guided decision in its record.

On import, the current normalization closes that entry check where practice evidence exists. The following condition now redirects any learning/entry scene with status skipped or partial to the report before validation returns the restored record. Both Resume and Restore therefore render the report rather than the obsolete entry question. Clearing learningRevision also avoids retaining a revision-screen state during recovery.

**Learner result:** earlier responses remain available with the unverified-timing warning, and the returning learner is no longer invited to lock an answer to a closed entry check. The ordinary report navigation provides a way to continue the expedition.

No additional review scope is implied. The nine-country/eighteen-move content was not changed or broadly re-reviewed here, and this source check does not establish live usability or learning outcomes.

# Post-implementation content verification — faculty reviewer A

**Date:** 29 October 2026. **Status:** AI expert content verification, not a human faculty participant or a classroom validation study.

This is a new implementation-stage check, separate from the locked blind review. I inspected `/workspace/sites/yoav-duman-recover/resource-sources/power-circuit/power-learning.js` and the new scaffolds in `cases.json`. I did not edit those files or the locked review. No new web retrieval was needed: the assessment introduces synthetic rules rather than new real-country constitutional claims.

## Verdict

**No serious content blocker found.** All 18 assessment scenarios have internally correct procedural outcomes and matching human-review criteria. The four numerical majority thresholds, the three dissolution-window applications, the targets of confidence procedures, and the research-inference caveats are correct under the rules explicitly supplied to the learner.

All 15 procedural items start with “Rule,” “Rules,” or “Rules and facts.” Those stated rules control their answers. In particular, the seven-day dissolution windows and the votes-cast successor rule are synthetic variations; they should not be corrected to resemble a named constitution. The assessment does not name a country in those questions.

Two low-severity refinements are worth considering. Neither changes a procedural answer:

1. **Label the return research report explicitly hypothetical.** `return-6` begins “A report observes…” without itself saying that the report is invented. In context it is clearly an assessment vignette, but a detached screenshot could be read as a claim about empirical findings. “A hypothetical report observes…” would remove that ambiguity without needing a citation or changing the task.
2. **Use the research criteria as conceptual review guidance.** In `entry-6`, the relevant distinction is office survival versus measured relief delivery or policy delay. A correct response need not additionally mention democratic survival when the supplied claim does not concern it. The current first criterion's longer list is useful background, but should not become a keyword checklist that penalizes a sound explanation. The module appropriately provides no automatic semantic grade.

## Item-by-item check

| Item | Independent application of the stated rule | Prompt and criteria judgment |
|---|---|---|
| entry-1 | 43 yes is below 47 no, so the bill fails. Continuing confidence and the absence of another procedure leave cabinet and assembly in place. | Correctly requires separate answers for bill, cabinet and assembly. Criteria do not conflate bill defeat with confidence loss. |
| entry-2 | Assembly selection plus continuing legislative confidence is the parliamentary origin/survival relationship, despite the title President. | Correct; an accurate description can earn recognition without requiring the label. No assumption of direct presidential election is imported. |
| entry-3 | The valid motion expressly removes PM/cabinet. It does not remove the separately elected fixed-term President, and no dissolution process occurs. | Correct. Appointment without investiture does not defeat the later responsibility expressly created by the stated rules. |
| entry-4 | A majority of all 84 is 43. The proposed successor's 41 is insufficient, even though 48 oppose the incumbent. | Correct denominator, threshold and distinction between opposition and successor agreement. No election follows from the failed attempt. |
| entry-5 | Both clocks start from passage, day 0. Resignation/request on day 8 satisfies the resignation period but does not restart the dissolution clock. Day 16 exceeds day 10. | Correctly concludes timing bars dissolution even though request and consultation conditions are supplied. A request alone ends neither legislature nor presidency. |
| entry-6 | A fictional President's survival and a blocked bill do not measure faster relief delivery or establish its cause. | Correctly asks for observable comparison, rival and challenging evidence. It rejects an unsupported inference without asserting that fixed terms can never affect outcomes. See the minor rubric note above. |
| exit-1 | 38 yes is below 42 no. The expressly continuing confidence agreement sustains the cabinet, and no dissolution/removal procedure occurs. | Correct and consistent with entry-1's mechanism, with different figures and policy. |
| exit-2 | The legislature selects and can require resignation of the executive. That is the parliamentary relationship, regardless of Chief President as a title. | Correct origin/survival test. The title supplies no additional mandate. |
| exit-3 | A successful motion targets premier and cabinet and requires their resignations. President and assembly remain under the stated exclusions. | Correct. Prior investiture is not immunity against subsequent confidence responsibility. |
| exit-4 | A majority of all 120 is 61. The successor's 59 fails; 70 votes against the incumbent cannot substitute for an agreed successor majority. | Correct threshold and procedural target. The existing chamber remains. |
| exit-5 | The premier's day-5 resignation is within twelve days. Dissolution has a separate seven-day period from passage; a decision on day 10 is late. | Correct. It does not borrow ten days from a country example, confuse the two deadlines, or restart time from resignation. |
| exit-6 | A successor requirement can motivate a hypothesis about replacement difficulty. A single game record with a pending budget cannot establish universally less delay or improved democratic survival. | Correct separation of outcomes and rejection of the universal causal inference. Comparison, outcome measurement, rivals and falsifier are appropriately required. |
| return-1 | A majority of all 70 is 36. Only 34 support no confidence, so it fails; the special stated rule makes the finance bill law. Cabinet and assembly remain. | Correct. The 36 non-supporters are not affirmative votes on an ordinary bill. The synthetic rule expressly supplies the special bill consequence, so no real-country promulgation or legislative-stage assumption is needed. |
| return-2 | The President has separate electoral origin. Neither minister nor cabinet depends on legislative no confidence, and ordinary ministerial dismissal belongs to the President. | Correct presidential relationship under these facts. A coordinating Prime Minister title alone does not establish semi-presidentialism, and the legislature retains its stated bill-blocking power. |
| return-3 | Procedure 2 excludes the President and requires that President to reconstitute the Cabinet. It does not import procedure 1's resignation result. | Correct target and consequence. It does not require every minister to resign automatically or make all presidential offices immune. Assembly remains. |
| return-4 | Only 41 for plus 39 against enter the denominator: 80. Its majority is 41, so the successor is elected and replaces the incumbent. Forty abstentions do not enter this stated denominator. | Correct transfer test. Using 61 of all 120 would contradict the supplied alternative rule. The existing chamber remains. |
| return-5 | Day 6 is inside the seven-day period, and a request exists. Required consultation has not happened, so dissolution is not yet available. | Correctly treats time as necessary but insufficient. The consultation could still occur within the remaining window; the criteria appropriately say “not yet available,” not permanently barred. A request alone does not dissolve the assembly. |
| return-6 | The vignette supplies fewer replacements, unchanged budget delay, a simultaneous electoral-system change, and no democratic-survival measure. It therefore does not isolate a confidence-rule effect on democratic survival. | Correct. Unchanged observed budget delay is not treated as proof of zero causal effect; comparison and rivals are still required. Label the report hypothetical as recommended above. |

## Catalog and arithmetic verification

A read-only evaluation of the module confirmed three forms with six unique items each: **18 items and 72 human-review criteria**. Every procedural scenario states its own rule. Independent strict-majority calculations gave:

| Application | Relevant denominator | Required affirmative votes |
|---|---:|---:|
| entry-4 constructive replacement | 84 total members | 43 |
| exit-4 constructive replacement | 120 total members | 61 |
| return-1 special linked confidence procedure | 70 total members | 36 |
| return-4 successor election | 80 votes for or against | 41 |

These checks validate the content arithmetic, not semantic scoring. The module preserves the distinction: its summary states that completion and self-reported confidence/help are not semantic scores or evidence of empirical learning gain. It also flags a partial or missing entry baseline rather than inventing a paired baseline. This content check does not establish psychometric equivalence among forms, learning gains, retention effects or classroom effectiveness.

## Country scaffold check: institutional facts versus empirical outcomes

I independently reread all nine countries' `orientation`, `mechanism` and `debrief` objects after implementation. Each has the required four/four/three plain-text fields. **No empirical country-performance claim is introduced by those 27 objects.**

- `orientation` supplies geographic region, institutional cast and the kind of hypothetical dispute. It contains no delivery rates, comparative efficiency estimates, real election results or current-officeholder assertions.
- `mechanism` states the target office, legal procedure, denominator or threshold and the institutional consequence. These are normative constitutional relationships and conditional applications, rather than observations of national outcomes.
- `debrief.rule` and `debrief.exception` distinguish legal mechanisms. They do not infer effectiveness, democratic quality or policy success from a constitutional family.
- `debrief.counterfactual` poses hypothetical one-fact changes. Toy chamber numbers are explicitly fictional where used; none is presented as an actual national legislature or election return. The Brazilian question asks what further procedural facts would be needed rather than claiming a real removal took place.
- The top-level scope explicitly says institutional rules are the subject and all scenarios are fictional applications of real constitutional mechanisms. Existing source IDs accompany the country material; Taiwan additionally includes the French constitutional source for its formal dismissal-rule contrast.

The scaffold content is safe to present as an institutional teaching aid. The ordinary caution is one of display context: retain the real-rules/hypothetical-events notice when rendering country incidents, and do not relabel counterfactual results as observed country outcomes. There is no need to add empirical citations for invented results or strip out the nine-country/18-move distinctions.

## Verification boundary

This note assesses the content of scenarios, prompts and human-review criteria, plus the empirical/nonempirical distinction in country scaffolds. It does not claim to verify UI release timing, hiding of criteria before form lock, accessibility, export behavior, scoring by human instructors or any classroom learning effect. No build was run and no implementation file was edited during this verification.