# R4 — Independent comparative-methods review

**Scope.** Fresh political-scientist reviewer specializing in comparative methods. Only game entry: `http://terminal.local:4173/?reviewer=R4`. Malawi main case; Malaysia comparison; class/solo mode. No code, candidate build, prior review, answer file, design note or other reviewer’s work was read. All game interaction used cua_repl and visible controls. Companion PDFs were read only after first-use gamework and responses were recorded.

## First-use understanding, recorded before entering

The landing page presents a fictional public evidence hearing about political competition, where historical outcomes are fixed and my explanation is open to revision. The guiding question is “How did an advantage change what another actor could do?” I should choose a main historical case, read available evidence, select two supporting cards and one complication for a three-space board, and later use a second case to test the reach of my explanation. I expect to build, face a serious challenge, revise, and produce a short class explanation. The board is a limit on evidence I use, not on what I may read. The page explicitly warns that cases concern different offices, dates, and measures and “cannot be ranked on one vote or seat score.” I infer the task is a bounded causal explanation and comparison, rather than predicting or changing the winner.

Before entry I do not yet know what counts as an “advantage,” what deliverable the final hearing requires, or how the six stages relate. The approximate 55-minute claim is visible; I have no observed learner timing or learning evidence.

## Actual gamework and completion

I completed all six stages and filed a substantive record. The app counted my explanation as 72 words and comparison as 43. I saw “Hearing finished · explanation ready for discussion,” then opened the current-record preview. It retained my opening, chain, paragraph, complication, hearing decisions, comparison, practice attempts and source limits. I closed QA tab 70 afterward.

| Stage | Work actually completed |
|---|---|
| 01 Opening | Recorded “A new contest was enforced” as a first idea to test. |
| 02 Evidence | Read all six Malawi full cards and all five source records. Selected MW3/MW5 supporting, MW6 complication. |
| 03 Connection | Wrote actor, action, changed option, optional inference, current paragraph and complication reason. Left optional board reasons/attachments empty. |
| 04 Questions | Deliberately tested the excessive compliance inference, read feedback, reconsidered correctly; answered the official-figures question; applied two reasoned keep decisions; supplied another step and needed evidence. |
| 05 Comparison | Selected Malaysia and “Similar result, different ways to act.” Read all six cards and all six source records; wrote comparison citing MY3/MY5. |
| 06 Finish | Confirmed source-limit/board self-checks, answered optional fictional transfer question, finished record and inspected preview. |

**Recovery:** An initial browser binding error cleared when I created the unique tab. After MW3 selection, a batched next click encountered a detached-control protocol error. One reload/Resume retained all reading progress and MW3. Fresh semantic controls worked thereafter. This is consistent with a stale automation target after rerendering, not established human-use failure. No further gameplay protocol failure occurred.

I did not test exports, imports, printing, copying, main-case switching, pair mode, global Glossary or the “People, groups, and terms” disclosures.

## Overall methods judgment

No critical or high-severity analytic falsehood appeared in my route. The game carefully separates legal reopening, implementation and winner; observed conduct from motives; media content from audience or persuasion; and formal completeness from valid reasoning. My deliberately excessive inference received precise feedback: “Observing that organizations complied does not by itself explain their reasons or show that compliance was inevitable.” Both attempts remained in the preview.

The remaining vulnerabilities concern defining the explained event consistently and making a source limit relevant to that event. Findings below are anticipated student risks inferred from the interface, not observed learner mistakes. **Moderate** means a plausible substantive misunderstanding that facilitation can repair; **Low** means local wording/emphasis.

## Findings and minimal fixes

### 1. Keep the explained outcome stable during comparison — Moderate

**Location / phrases:** Malawi opening: “Why could the declared winner not simply keep the first result in place?” and “Explain the opportunity to vote again separately from the choice of winner.” Stage 05 purpose: “Similar result, different ways to act”; prompt: “Compare the changed option in Malawi and Malaysia.”

**Misunderstanding / impact:** My main explanation concerned reopening a contest. The countries share a broader turnover outcome, but their immediate explained events differ. I could construct a bounded contrast; the comparison did not identify courts’ or alliances’ causal effects. A student could silently switch outcome levels, then conclude that Malaysia’s different pathway disproves the courts’ role in Malawi, or that shared turnover demonstrates a shared cause. Those are analytic errors, not vocabulary problems.

**Minimal fix:** Retain a short target-outcome phrase above the comparison: “You are explaining [step]. Is that same step present, absent, or different here?” Reuse an existing choice or phrase; no extra essay. Add: “A contrast can bound your claim; these selected cases do not isolate a cause.” The guide already makes this limitation; surface it at the decision point.

**Retain:** Comparing actions/options instead of country summaries, the original chain displayed beside Malaysia, measurement warnings and the custom comparison question.

### 2. Distinguish a rival from a complementary later step — Moderate

**Location / phrases:** Stages 04/06: “Another possible explanation or different step,” beside “What evidence would help distinguish your explanation from another one?”

**Misunderstanding / impact:** I entered the joint ticket as a possible explanation of the winner while explaining reopening in my paragraph. Those accounts can both be true. Voter evidence can illuminate the winner without weakening the legal-opening account. A student may believe adding a fact about another step has tested a rival explanation, although it predicts nothing different about the same event. My own evidence request therefore split the research questions rather than adjudicating competing causes.

**Minimal fix:** Keep both legitimate routes, but ask: “What evidence would test the missing step, or favor one competing explanation over another?” Add a short helper: “A different step can complement your explanation. A rival for the same step should predict something different.”

**Retain:** A complication need not contradict everything; supported claims may be kept; human review determines relevance. Do not force artificial disagreement.

### 3. Lead with claim-specific source limits — Moderate

**Location / phrases:** MW3 says “Judges interpreted the constitution to require more than half of valid votes for a winner.” Its source details foreground invalidated results and precise totals. MW-C begins “Official retrospective administration account,” then gives 59.33%/39.93% versus 59.34%/39.92% and “The reason for this difference is unestablished.” The numerical paragraph repeats beside MW3, MW5 and MW6 in preview. MW-D’s more relevant caveat—“Original judgments could not be retrieved”—appeared in MW2’s source details and the final preview.

**Misunderstanding / impact:** The exact-total discrepancy is honestly preserved, but for my legal/implementation claim, source mediation and retrospective self-description matter more. A conscientious student can supply a true peripheral numeric caveat while failing to bound the claimed institutional link. The relevant caveats are disclosed; their placement/emphasis is the issue.

**Minimal fix:** Lead MW3’s details with “This is the commission’s account of the court sequence; the original judgments were not directly inspected.” Lead MW5’s with “This retrospective administrator account describes implementation; it does not independently establish why other institutions complied.” Keep the complete numerical audit accessible in a secondary paragraph/disclosure. Preserve its unexplained status; do not dismiss it as rounding.

**Retain:** Upload/event/publication distinctions, exact source pointers, honest original-judgment limitation and the board’s warning that shared MW-C cards are not automatically independent checks.

### 4. Qualify the higher-share headline — Low

**Location / phrases:** MW6: “A higher share still ended in defeat”; body: “Mutharika’s valid-vote share rose from about 38.6% in the declared 2019 result to 39.9% in 2020.”

**Misunderstanding / impact:** The qualified body is accurate as a comparison of reported shares, but the memorable headline can become a claim that underlying popular support demonstrably rose. The invalidated first result, changed participation and multiple other changes prevent that inference. Existing qualifications strongly mitigate the risk.

**Minimal fix:** “A higher reported share still ended in defeat” and “His reported valid-vote share was higher…” Keep numbers, turnout, multiple-changes warning and source-history caveats.

## What works and undergraduate accessibility

The actor/action/changed-options chain is demanding but concrete. The everyday library example offers a usable anchor without supplying a country answer. All reading remains available, while three board spaces constrain public attention. A complication can expose another explanation, missing step or source limit.

The specific source distinctions are unusually valuable: MW1 concerns monitored airtime, not audiences or beliefs; MW0’s invalidated declaration is not a corrected count; MY1 explains local district contests; Malaysia’s PR/PH memberships and Warisan exclusion are explicit; MY6 is clearly post-election and cannot explain an earlier victory. The board exposes source dependence. Final inclusion checks expressly avoid judging truth or logic, and the preview preserves attempts instead of equating reconsideration with mastery.

An able undergraduate could plausibly complete the sequence with normal facilitation. I reached substantive filing before consulting PDFs; both word limits accommodated a bounded response. This is an expert usability forecast, not learner evidence. Country names, acronyms and “complication” may be unfamiliar wording; they are different from collapsing successive outcomes, inferring motives from compliance, or treating aggregate shares as individual voter transfer. I did not inspect glossary content and do not diagnose missing definitions.

## Companion PDFs and scope

After recording gamework, I read the five-page Facilitator Guide and four-page Source Register and visually inspected page 3 of each. They support the cautious interpretation above. Guide p. 2 calls the cases a primer rather than a controlled test; p. 3 says a result is not a mechanism and repeated table/card facts remain one observation. Its Malawi note distinguishes constitutional interpretation from amendment and rejects aggregate-share voter transfer. Page 4 appropriately requires observed classroom work before learning claims; p. 5 separates participation from mastery. Register entries match the visible source-role, sequence and discrepancy qualifications.

I would retain the bounded writing and use explicit closing discussion linking these cases to the assigned reading. Brief instructor prompts—“What exactly are you explaining?” and “Would that observation favor one account, or explain a later step?”—preserve rigor without adding an essay.

This is one AI expert-role path. I independently verified no external historical original and read no assigned political article. I have no valid learner timing, retention or learning-gain evidence. Visible timestamps are record metadata, not controlled timing evidence. No source/build was edited and no external content was written. Browser QA is complete and closed.

## Preserved responses actually submitted

**Connection**

- Actor: “The courts, followed by Parliament and the election commission.”
- Action: “The courts set aside the 2019 election and upheld a fresh vote on appeal (MW3); other institutions dated and administered the new contest (MW5).”
- Changed option: “Opposition candidates and voters gained a new presidential contest; the declared incumbent could no longer rely on the first declaration as a final result. Before implementation, a legal order alone did not supply a new ballot.”
- Possible effect: “It explains the opportunity to vote again, not which candidate won that new contest.”

**Main explanation (game count: 72 words)**

> Malawi's courts set aside the declared 2019 election and upheld a fresh contest on appeal (MW3). Parliament and election administrators then supplied a date, voting arrangements and posted local results (MW5), allowing challengers and voters to contest the presidency again. This explains reopening, not Chakwera's victory: alliances, turnout and the majority rule also changed (MW6). The administrator's retrospective report documents implementation but cannot independently establish why institutions complied or how voters decided.

**Complication reason**

> Reopening the contest does not explain the winner; multiple changed conditions prevent attributing the result to one factor.

**Hearing decisions**

> MW3 and MW5 support reopening; my paragraph leaves compliance motives and the winner unexplained.

> MW6 compares reported shares with changed turnout and rules; it cannot isolate the coalition effect or identify individual voters.

**Another possible explanation or different step**

> The joint opposition ticket may help explain Chakwera’s victory, a different step from reopening the contest.

**Needed evidence**

> For reopening: contemporaneous institutional deliberations and independent observation of compliance. For the winner: voter panel or credible constituency evidence on coalition support and participation.

**Comparison (game count: 43 words)**

> Malaysia's PH chose Mahathir and reorganized cooperation while PAS stayed outside (MY3, MY5). Challengers could contest through a different alliance, whereas Malawi's challengers gained a court-ordered repeat ballot. This bounds judicial reopening to Malawi's pathway; turnover alone does not identify a shared cause.
