 THE HEARING / FIRST-USE PRESSURE TEST / 30 OCTOBER 2026



 Ten fresh perspectives. Concrete
 revisions.
 Five simulated novice students and five simulated faculty used the frozen first version before
 changes.

 Main finding: all ten completed a case and comparison and reached a filed report. They found strong
 source discipline, attractive visuals and recoverable navigation. They also exposed vocabulary friction,
 redundant claim fields, easy answer cues and a gap between changing the board and revising the
 paragraph.

 Each reviewer started without design notes, source code, hidden answers or other reviewers' comments.
 Student roles represented uneven ordinary preparation rather than exceptional performance. Faculty could
 read the facilitator guide only after recording their first understanding and playing. Separate save spaces
 prevented reviewers from overwriting one another.

   ID     Perspective                      Main / comparison                Observed outcome


   S1     Willing, low confidence          Russia / Malawi                  Filed; 72-word account, 39-word comparison


   S2     Skimming, impatient              Malawi / Russia                  Filed after recovering three missing items


   S3     Writing-reluctant pair           Malaysia / Malawi                One shared record filed; 64 / 37 words


   S4     Confident, confirming facts      Serbia / Russia                  Filed; weak current claims remained beside a
                                                                            narrower paragraph


   S5     Slow reading, limited terms      Malawi / Malaysia                Filed; sampled keyboard actions worked


   F1     Comparative politics             Russia / Malaysia                Filed; challenged assessment alignment


   F2     Research methods                 Malawi / Serbia                  Filed; source and causal critiques


   F3     Learning design                  Malaysia / Malawi                Filed; affected-actor scaffold critique


   F4     Inclusive course design          Serbia / Malaysia                Filed; reproduced keyboard orientation issue


   F5     Assessment and game              Malawi / Serbia                  Filed; challenged mastery inferences
          design

 These are AI role simulations and expert walkthroughs. Ten completions are not a human usability sample, a validated
 completion rate, a timing result or evidence of learning gains. The first-use reviews remain unchanged; informed retests
 are separate.




THE HEARING / TEN FIRST-USE AI REVIEWS / REVISED 30 OCTOBER 2026                                                            1
 REVIEWER COMMENTS / BEFORE THE REVISION



 What they said, and the response
 Full comments, action logs and sample explanations are included in First_Use_Reviews in the kit.


   ID     First-use comment or finding                             Response implemented


   S1     "Two anchors" could mean two countries; the              Main case and supporting evidence now have
          hearing asked for explanations without an obvious        different names. A short rival explanation and
          place to record them.                                    evidence request are explicit.


   S2     Applying a narrowing felt like finishing the revision,   Before/after board text, an explicit
          although the paragraph stayed unchanged.                 unchanged-paragraph notice, a focused review
                                                                   action and a final reconciliation check.


   S3     "MY-D" looked like another card; "civil forfeiture"      Country actor keys, card/source explanations, legal
          and party labels required decoding.                      terminology and a searchable glossary. The
                                                                   allegation question has one clearer source focus.


   S4     The complication became another confirming fact;         Prompt asks what becomes less certain.
          old causal claims and attachments survived beside        Changed-step attachments clear; students review all
          new writing.                                             current claims and the selected supporting
                                                                   evidence.


   S5     Three controls per chain step, another revision          One current paragraph, compact optional
          choice, and another limitation phrase felt repetitive.   starters/attachments, no required duplicate
                                                                   limitation phrase or overall revision selector; notes
                                                                   wrap.


   F1     Choosing the cautious answer is easier than              Closer alternatives plus an authored rival and
          producing a rival explanation; too many claim            evidence request. One retained opening and one
          objects.                                                 current written explanation.


   F2     Two card IDs are not necessarily independent             Shared-source notice and source-role prompts; full
          corroboration; source limits should fit the claim.       provenance stays available. Human review still
                                                                   judges the causal link.


   F3     The options field could repeat PH rather than            Explicit affected-person/group prompt and a nearby
          identify another actor whose choices changed.            neutral example; original political evidence is
                                                                   preserved.


   F4     Reopen scrolled to a card but keyboard focus stayed      Requested-card focus, return-to-task controls and
          at the stage heading. Missing-item errors were           direct finishing links to the relevant field.
          generic.


   F5     Completion is submission, not mastery. Plausible         Practice is labelled as practice; revised question
          rivals and human calibration examples are needed.        bank and a five-page guide with weak/strong
                                                                   fictional examples and a human rubric.




THE HEARING / TEN FIRST-USE AI REVIEWS / REVISED 30 OCTOBER 2026                                                           2
 IMPLEMENTED CHANGES / RESEARCH RIGOR RETAINED



 Less decoding. More visible
 reasoning.
 The four cases, historical results, precise source caveats and comparison requirement remain.

 Plain language and orientation
 Main case, supporting evidence, evidence board and Finish & save replace ambiguous overlapping labels.
 Actor keys explain the cast. The glossary defines denominator, majority versus largest share, ticket, source
 role and other terms. It is deduplicated, sorted and searchable; case notes are labelled separately.

 One explanation, with a visible revision handoff
 The opening is retained as a first version; the current paragraph is shared across stages. A board revision
 shows before/after wording and sends the learner back to review the paragraph. The app does not rewrite
 student prose. Earlier planning notes and old question rounds remain recoverable rather than
 masquerading as current writing.

 A clearer thinking requirement
 A short rival explanation and a specific evidence request are visible work for a reader to assess. A
 complication can be a rival, source limit or missing step; it need not contradict everything. The main
 paragraph must cite the two currently selected supporting cards. Self-checks ask the student to reconcile
 the board, evidence and paragraph.

 More discriminating practice questions
 Twelve active questions now use closer mistakes about populations, units, timing, source status and
 necessary versus sufficient explanations. Two faculty-role reviewers independently checked the revised
 bank. Their corrections removed an ambiguous subgroup claim and several invented-event distractors. Old
 questions retain their original wording in saved histories.

 Reading and recovery behave consistently
 Reading counts update immediately. Opening the same source through the hearing counts as source
 access. Notes wrap. Reopen focuses the requested card and provides a return route. Finishing checks
 identify the missing item and take the learner to it. A shared evidence request now survives first
 comparison selection and later country changes.

 What we kept and what we deferred
 Reviewers consistently valued the calm collage design, visible known outcomes, limited board, editable
 connections and source limitations. We kept them. We did not add timers, points for changing views,
 invented electoral effects or a machine interpretation grade. More competitive audience dynamics are an
 optional future design question after the classroom pilot, not another required phase now.

 The revisions improve instructions and make reasoning inspectable. They do not guarantee that a selected complication is
 logically effective or that an explanation is well supported. A human must read the report and discuss the claim.




THE HEARING / TEN FIRST-USE AI REVIEWS / REVISED 30 OCTOBER 2026                                                            3
 VERIFICATION / LIMITS / NEXT CLASS



 What was retested
 Initial blind reviews and later informed checks are different forms of evidence.


   Check                        Result and scope


   Student S2 retest            Finished the revised Malawi/Russia route. Verified before/after handoff, unchanged
                                prose, live counts, rival/request fields and self-check recovery.


   Faculty F4 retest            Verified requested-card keyboard focus, next-Tab destination, return to the hearing,
                                actionable finishing links and source access through hearing drawers.


   Student S3 retest            Verified terms, party key, focused legal question and revision handoff. Confirmed the
                                request reset. Final completion was not retested because the QA route became
                                unavailable.


   Content verification         Two informed faculty-role checks reviewed all 12 revised active questions against the
                                original evidence. Corrected ambiguity and grounded the closer alternatives.


   Logic and migration          128 automated state/content assertions across four cases: source fidelity, chosen-card
                                citations, review flags, filing, replacement, escaping, old questions/attempts and
                                evidence-request preservation.


   Document review              The revised guide and this report were rendered and visually checked. Full first-use
                                comments are preserved in the kit.


 A real regression found during retest
 S2 noticed a request disappearing after choosing a comparison. Code inspection confirmed that country
 selection recreated the comparison object without the newly shared request. The revised handler preserves
 it. Event tests cover first selection, later switching and continued editing. A final rendered-browser check
 also confirmed preservation on first selection and a later country switch. The original retest note keeps its
 initial, cautious description.

 What this does not establish
 No human students or external faculty participated. Browser automation time is not learner time. Protocol
 timeouts recovered with reload and Resume; they were not classified as application failures. Sampled
 keyboard behavior is not a full accessibility or physical-phone audit. In-game download completion remains
 unconfirmed in the cloud browser; report generation and preview are checked, with print/copy alternatives
 available.

 The next useful pilot
 Run the 55-minute plan with an actual class. Record phase-finish times, misunderstood terms, whether
 revisions reach the prose, and the quality of the rival/evidence request. Review initial and revised work with
 the five human criteria in the guide. A later new-case task can probe retention; estimating learning effects
 requires a suitable comparison design.




THE HEARING / TEN FIRST-USE AI REVIEWS / REVISED 30 OCTOBER 2026                                                         4
