# F2 — independent first-use faculty review

## Basis and completed path

I reviewed the visible `?reviewer=F2` interface as a social-science research-methods teacher encountering this game for the first time. I inspected no implementation, hidden answers, design history, or other reviews. I read the permitted facilitator guide only after recording my initial understanding and completing the interface path. This is an AI faculty-role walkthrough, not an observed student trial; it establishes neither human completion time nor learning outcomes.

I entered as “F2 sample,” selected Malawi, recorded my own opening question, opened all six full cards and their source records, selected MW3 and MW5 as anchors and MW6 as complication, wrote my own causal connection and account, answered both hearing questions, applied two defended responses, compared Serbia using RS5, completed the optional fictional inference question, and filed. The interface confirmed “Hearing filed · interpretation awaits discussion.” I inspected the actual record preview, including my writing and attempt history. I did not test downloading or printing.

## First understanding and instructions

Before opening help, I recorded this understanding: explain how an advantage constrained another actor; use Malawi and a comparison; select three public exhibits; answer a challenge; write a short account. I was unsure whether “two anchors and one complication” meant countries, evidence cards, or arguments. The interface initially uses “anchor country” and “anchors” for evidence in close proximity. The persistent exhibit and later glossary resolve this, but first-use wording could avoid the collision.

The entry statement that “your choices change the analysis, not history” sets an honest task. Known outcomes remove an unnecessary guessing game. Malawi’s opening usefully separates how another contest became possible from why its winner changed. Incumbent, opposition, and coalition receive immediate definitions. Help supplies a neutral library example, distinguishes an attendance record from a causal explanation, and describes all four major operations. I could proceed without the facilitator guide.

The glossary is helpful, especially “An observation is what a source records. An inference is the explanation you draw from it.” Some language still requires support: denominator, retrospective, annulment, institutional implementation, and majority rule. The glossary’s “Seat and majority” defines majority through legislative seats, whereas Malawi applies a majority of valid presidential votes. That is a small but consequential cross-case ambiguity. “Advantage” itself is not unpacked as clearly as the subsidiary vocabulary.

## Evidence, inference, and causal rigor

The historical evidence presentation is unusually careful. MW0 distinguishes declared 2019 figures from the 2020 report, names valid votes as the denominator, gives turnout separately, and warns that the annulled declaration cannot reconstruct a fair count. MW1 separates monitored content from audience size and beliefs. MW6 exposes simultaneous changes rather than allowing an opposition victory to prove a single mechanism. These are worthwhile research-methods operations.

I could make a defensible narrow account: the order surviving appeal, followed by implementation, opened another contest. My final account explicitly refused to make this sufficient explanation of Chakwera’s victory. The actor/action/changed-option fields support that separation, and the fourth field labels the possible consequence as inference. The Serbia comparison enabled a real contrast between available challenges and an implemented national remedy, without comparing presidential vote shares directly with parliamentary seats.

Source records provide issuers, dates, locators, links, and limits. They disclose the two MEC series’ unexplained discrepancy and the absence of retrieved original judgments. That candor is valuable. Nevertheless, repeated MEC notes devote substantial space to numerical differences even when the card’s analytical role is institutional implementation. For MW3, the commission’s own involvement in the appeal makes its retrospective description a particularly relevant source-role issue. A novice needs help identifying the limitation that matters to the specific claim, rather than copying the most prominent caution.

Two distinct card IDs also need not represent independent corroboration. MW3 and MW5 support different steps but share an administrative source. The guide correctly asks a human assessor to distinguish observations; the interface cannot establish triangulation merely by counting cards. This limitation is disclosed, but a brief “same source, different observation” note would improve source reasoning.

## Choices, writing, and assessment

The integration of writing is strong. The draft begun during connection appeared unchanged during questioning and filing; there was no surprise replacement essay. My 73-word account cited MW3, MW5, and MW6, distinguished opportunity from winner, and stated a retrospective-source limit. My 42-word Serbia comparison cited RS5 and explained whose opportunity differed. A separate limitation phrase and next-evidence request made the final product reviewable.

The selected cards visibly affected the hearing: “Prompt selected because your exhibit includes MW3, MW6” and a second prompt for MW3, MW5. The statement “No software has read your prose for meaning” prevents a false impression of personalized semantic feedback. I defended both challenges because my account already respected their bounds. Defending rather than changing merely to satisfy the software was legitimate and recorded.

The questions themselves are the largest pedagogical weakness. Alternatives such as judges directly appointing Chakwera or an official label proving a fully fair process are conspicuous overclaims. The qualified, multi-clause answer is easy to recognize without explaining why a more plausible rival fails. My actual account was never seriously threatened. Better alternatives would distinguish necessary from sufficient conditions, administrative compliance from judicial capacity, or coalition coordination from individual vote transfer.

The final “Inclusion checks” correctly say they cannot judge “quality, truth, originality, or logic.” The record preserves first attempts, responses, chosen evidence, and initial/final views, making it useful for formative discussion. The guide’s participation scale is suitable for substantive engagement, but it is too coarse to assess research-methods mastery by itself. An assessor needs brief examples of strong, developing, and unsupported explanations against the existing evidence/connection/scope/challenge/comparison criteria. Filing should remain an organizational milestone, as it currently is.

## Observed recovery and limits of observation

I deliberately selected the unsupported official-label answer to test recovery. Feedback explained that administrative publication records what was declared rather than validating a fair count. I then selected the bounded comparison answer; feedback changed and “Attempt history · 2” appeared. Both attempts were present in the filed preview. This was a controlled test, not evidence that a novice naturally made that error.

After selecting Serbia, browser observations timed out for both page text and screenshot. One reload returned a saved-hearing entry; resuming preserved my prose, exhibit, decisions, and comparison selection. I cannot attribute the timeout to the app. A separate detached-element error occurred when my automation tried another numbered control after the exhibit rerendered; fresh observation resolved it. Neither event establishes an ordinary student-facing defect.

## Agency, visual clarity, and lesson load

Agency is meaningful as curation: foregrounding evidence changes the exhibit and authored questions; custom accounts, bounded defense, alternative cases, and revision remain possible. The three-slot constraint represents public attention coherently. It provides a reason to choose without pretending that historical results are controllable. However, the experience is principally a structured inquiry workspace. There is little strategic uncertainty or response from a simulated audience beyond source feedback, and choosing an already cautious account can make the hearing feel procedural.

The visible desktop layout is restrained and readable. Strong stage headings, a six-stage navigation, and the persistent exhibit support orientation. The report preview presents the student’s own words prominently. Long source expansions create substantial scrolling, and single-line exhibit-note fields visually clip longer reasons. I did not inspect a phone or narrow viewport and make no responsive-layout claim.

An average 18-year-old with ordinary reading and percentage knowledge could probably operate the controls. Independent causal interpretation is a different demand. Anticipated novice difficulties include treating chronological sequence as explanation, equating higher shares with increased support, and reproducing supplied qualifiers without understanding them. These are predictions, not observed learner errors. The guide’s modeled opening and proposer/skeptic discussion would help. Its 55-minute schedule is a design estimate: six cards, source evaluation, four connection fields, hearing decisions, two pieces of writing, comparison, and debrief impose appreciable switching costs. Protecting comparison and discussion will require selective source-detail reading and explicit checkpoints.

## Priorities and verdict

- **Blocking:** No blocking defect remained on my completion path. Actual 50–60-minute feasibility and educational effects remain untested; this walkthrough cannot supply those results.
- **Major:** Replace caricatured distractors with plausible rival inferences and require one sentence explaining the disputed step. Add claim-specific source-role prompts, including shared-source dependence. Supply a compact assessed example and a realistic novice route that identifies essential versus extension source detail.
- **Minor:** Use “supporting exhibit” alongside “anchor country”; define denominator and majority by the relevant unit; let exhibit reasons wrap; reduce repeated source blocks through a shared-source reference; make the distinction between working-view text and evolving final-account text easier to see.

I would use this in a facilitated first-year methods class as a promising formative exercise. It already supports bounded claims and produces an inspectable reasoning record. I would pilot it before relying on its timing, and strengthen the rival explanations before using performance as evidence of independent causal reasoning.
