# R11 — E-learning, accessibility and course-design review

**AI simulation / independent cold-read content review.** This is a first encounter with *The Hearing*. It is not a human student trial, an observed usability session, a screen-reader or keyboard audit, or mobile accessibility certification.

## Scope and method

I first read the authorized `Player_Screens.txt` packet, concentrating on entry, Malawi and Serbia, their comparison routes, the glossary and completion. I inspected the one authorized entry screenshot. I established the cold understanding below before consulting `Facilitator_Guide.pdf`. I then read the subsequently authorized `How_to_Play.txt`, which supplies the help dialog omitted from the original packet. I did not inspect game source, previous review reports or the browser, and made no source edits.

Locations below use the frozen packet's text line numbers, help text line numbers, or printed guide pages. The screenshot contains an existing saved hearing; it cannot establish the default appearance of a first visit. Disclosures are expanded in the packet, so repetition in the packet does not establish that all of it is simultaneously visible during play. No supplied answer feedback was available for assessment.

**Cold understanding before the guide:** choose a provisional account; curate two supporting observations and a complication; connect an actor and action/resource to another actor's changed opportunity; draft one bounded explanation; consider two supplied source-use challenges and record decisions; name an alternative and evidence that could distinguish accounts; test the explanation against a second case; preserve a discussion record. This is an evidence-reasoning exercise with known historical outcomes, not an election-winning simulation.

## What already works

- The entry explicitly says choices change analysis rather than history. Country files distinguish offices, dates and measurement units.
- Malawi repeatedly separates reopening a contest from explaining its winner. Serbia distinguishes a campaigning president from a parliamentary candidate and active opposition from equal conditions.
- Optional starters, a concrete library microexample, local people-and-terms sections and a substantial glossary provide useful scaffolding.
- The help dialog gives real pair duties: “The Proposer makes a connection; the Skeptic asks what would weaken it. Swap before the second hearing question” (`How_to_Play.txt:17`). These instructions are present, not missing.
- The player-facing record says “Included” is not correctness, and describes multiple-choice work as practice. The guide makes appropriately restrained claims about participation and learning.

The fixes below preserve the two supporting observations, a meaningful complication, claim-specific source limits, revision, a rival account/evidence request and a substantive second-case comparison. None requires weaker evidence or less analytical work.

## Findings

### R11-01 — Give independent learners a visible model of the quality standard

**Severity: High for unsupported independent use; medium with active facilitation.**

**Evidence and location.** The opening contract is “one 60–85-word explanation using both supporting card IDs and a source limit; a 30–45-word comparison; short phrases naming another explanation and needed evidence” (`Player_Screens.txt:1800`, also `2675`). Stage 3 supplies one sentence of library reasoning and a source limit (`2065`, `2905`). The fuller standard and contrasting weak/strong fictional explanations are in the facilitator guide, pp. 3 and 5. The help microexample (`How_to_Play.txt:13`) is useful but does not show the whole finished explanation or a comparison.

**Risk.** A learner can understand the outputs without knowing how to judge the connection. A statement that names an action, cites two IDs and adds a generic “this does not prove everything” can look compliant while leaving the crucial option change unexplained. A short comparison is particularly easy to turn into two accurate country facts. “Human review” identifies who judges; it does not yet show the learner the standard that person applies.

**Practical fix.** Bring a compact version of the existing guide rubric and fictional calibration example into an expandable “What makes this defensible?” panel before drafting. Add one fictional comparison that explains why a process difference tests a claim. Annotate the observation, changed option, inferred link and claim-specific limit. Keep the historical answer open and require students to write their own paragraphs. A small self-check should ask whether the two cited cards supply two distinct observations, rather than merely two valid IDs.

**Check next.** Observe whether first-time learners can explain the difference between an accurate observation and an accountable connection before they draft. Do not use a filled checklist as evidence of that understanding.

### R11-02 — Malawi's headline question and two of its starting routes target different explanatory steps

**Severity: Medium.**

**Evidence and location.** The driving question is “Why could the declared winner not simply keep the first result in place?” (`1798`). Yet learners may begin with “A joint opposition ticket” or “A different contest and participation” (`1804–1805`), which address the later winner. The local terms section accurately asks “Why was another contest held? Why did it have a different winner?” (`1795–1796`), and the guide, p. 3, separates reopening, implementation and the new winner.

**Risk.** The strong distinction taught in the evidence can become an apparent instruction conflict. A student choosing the joint-ticket route may think the headline demands changing to courts, or may overstate the ticket's role in forcing a new election. Conversely, a court-based explanation may feel incomplete because the known outcome includes an opposition victory.

**Practical fix.** Ask explicitly which step the current explanation addresses: reopening the contest, carrying it out, or explaining a changed winner. Let the chosen scope appear next to the working question and paragraph. Use an umbrella question that accommodates all three routes, while retaining the warning that evidence for one step does not explain the others. Do not require one 60–85-word paragraph to prove all three steps.

**Check next.** Ask learners from different starting routes what exactly they are explaining. Inspect whether the hearing invites a justified retained account rather than prompting everyone toward the same court narrative.

### R11-03 — The Serbia seat-allocation route loses support in the connection stage

**Severity: Medium.**

**Evidence and location.** A legitimate starting option is “Ballots became a seat majority,” explaining 129 of 250 seats from 46.75% of all ballots cast (`2680`). RS6 introduces proportional allocation, a 3% threshold and minority-list rules (`2854–2856`). The Stage 3 starter sets (`2909–2954`) cover favorable news, presidential campaigning, opposition organization and challenges, but provide no allocation-rule route. The changed-option prompt demands “What could they do before, and what could they do after the action?” (`2932–2933`).

**Risk.** These starters are optional, so the route is possible, but the available scaffolding nudges a learner away from an offered explanation. A stable rule or position also fits a before/after action formulation awkwardly. Students unfamiliar with proportional allocation may recite “threshold” as a cause without explaining what changes in whose parliamentary opportunity. The aggregate results do not by themselves establish the contribution of every rule.

**Practical fix.** Add a rule/resource route to the starter set and adjust the prompt to permit a comparison under a rule as well as a chronological action: what opportunity did the rule make available or restrict, for whom? Supply a small fictional allocation example to teach the unit distinction, without presenting a purported reconstruction of Serbia's actual allocation. Explicitly retain the limit on what the displayed totals establish.

**Check next.** Follow a learner who selects the seat route. Confirm they can complete an action/rule-to-option connection without inventing a history of rule changes or abandoning their initial route solely to fit the starter menus.

### R11-04 — Document-level caveats sometimes substitute for the limit of the observation being used

**Severity: High because source judgment is a central learning goal.**

**Evidence and location.** MW5 is “Organizations carried out the fresh election” (`1988`), describing administration, result copies, posting and security. Its MW-C limit prominently discusses exact candidate totals: “59.33% … 39.93%,” the separate publication's “59.34% and 39.92%,” and unexplained differences (`1997`). That identical caveat appears on other MW-C uses. RS5 concerns challenges through 6 February (`2841`), while its source limit repeats “Recorded ballots do not independently validate free preferences. Media content does not measure audience exposure or persuasion” (`2847`).

**Risk.** The statements are cautious, but the most prominent limit can be remote from the learner's claim. A student using MW5 to explain implementation could paste a decimal discrepancy as the source limit and never consider that an administrator's retrospective report does not independently establish why institutions complied or how each safeguard worked. Repeated full metadata and totals also compete with the distinction the card is meant to teach.

**Practical fix.** Lead each card with a short “For this observation” limit tied to its use. For MW5, distinguish the administrator's account of steps taken from independent verification of their effects or reasons for compliance. For RS5, emphasize the time boundary and that a recorded failed challenge does not establish permanent impossibility. Preserve the full source record, conflicting exact totals and provenance in a clearly linked document-level section; avoid making students reread the same block to find the relevant qualification.

**Check next.** Give learners the same underlying document used for different claims and ask them to select the appropriate limit. A copied true caveat is not sufficient if it leaves their actual inference unbounded.

### R11-05 — The source-reading instruction is stronger than the final readiness signal

**Severity: Medium.**

**Evidence and location.** Stage 2 instructs: “open the full text and Source details of the cards you use” (`1858`, `2733`); second-case directions say the same (`2371`, `3207`). The final inclusion list checks only “At least one main-case source record opened” (`2563`, `3414`). The card/source counters correctly warn “Opening a record does not prove you have read it” (`1911`, `2776`).

**Risk.** The disclaimer is good, but the visible readiness requirement can still become the operational minimum: one opened main-case record may look sufficient even when selected cards rely on additional records and the comparison source has not been considered. “Source records: 0 of 5” versus six cards is reasonable when documents are shared, but novices need that relationship made explicit.

**Practical fix.** Track coverage of the source details relevant to the cards actually used, including the comparison, rather than implying that all records must be opened or that one establishes adequate source work. Show which selected cards share a document. Keep the explicit warning that opening is only an access trace, and ask the learner to confirm that they reviewed the relevant details. Do not treat distinct card IDs as independent corroboration.

**Check next.** Verify actual completion behavior and the report's representation of opened sources. Test a record with selected cards whose distinct source records were not opened; do not infer the present gate from text alone.

### R11-06 — “Apply to my board” obscures the concrete revision action

**Severity: Medium.**

**Evidence and location.** After “Consider this reading,” learners choose “Keep it, with a reason,” “Make a step more precise,” or “Change supporting evidence,” followed by “Apply to my board” (`2188–2196`, `3023–3031`). The paragraph explicitly says the app will never write or rewrite it (`2119`, `2959`). The guide, p. 2, explains that applying a board change does not rewrite the paragraph.

**Risk.** The button groups three different jobs under one action. In particular, selecting “Make a step more precise” is not itself a more precise step, and a board contains evidence rather than the whole argument. Learners can confuse recording a revision intention with carrying it out. The paragraph warning helps, but does not fully explain which parts of the connection, evidence board and decision record this action changes.

**Practical fix.** Use action-specific wording and a short resulting-state message: record the reason for retaining; return to the named connection field to revise; or return to evidence selection to replace. Afterward show what actually changed and prompt the student to align the paragraph. Preserve the earlier decision and student's own wording. Do not generate the analytical revision for them.

**Check next.** Actual interaction must establish the behavior of each branch, where focus goes, whether an unfulfilled intention can count as recorded work, and what a reader of the exported record can distinguish.

### R11-07 — Requiredness needs one consistent notation across the whole task

**Severity: Medium.**

**Evidence and location.** The connection clearly says its first three steps are required and starters are optional (`2063`). Supporting-board reasons are explicitly “optional phrase” (`2262`, `2272`), but the complication reason says only “short phrase” (`2278`), while the final checks expect “A short complication reason recorded” (`2559`). The rival field is explicitly required (`2236`), whereas the evidence-request field says only “short phrase” (`2238`) despite its final readiness check (`2579`). “What does it show? An advantage / An action / A limit” appears on support slots without a stated optional/required status (`2258–2261`).

**Risk.** A learner has to infer whether the absence of “optional” means required, and may discover unfinished work only at the end. The three board labels also overlap: a recorded action can create an inferred advantage and have a source limit. They can look like exclusive factual classifications rather than choices about how evidence is being used.

**Practical fix.** Apply the same visible “Required” and “Optional” labels to all fields, tied to actual intended readiness rules. Explain board labels as the role the observation plays in this explanation, including whether a best-fit choice or multiple roles are allowed. Include every required operation in the initial task contract without adding more prose responses.

**Check next.** Verify which fields actually gate finishing, whether errors identify the affected field clearly, and whether labels/required state are exposed correctly to assistive technology. The packet cannot establish those behaviors.

### R11-08 — Good help exists; key cues and repeated destination labels still need contextual access

**Severity: Medium accessibility risk; no conformance failure established.**

**Evidence and location.** The screenshot shows “How to play” and “Glossary” in the entry header. The help contains the Proposer/Skeptic roles and swap timing (`How_to_Play.txt:17`). Those role cues do not appear in the supplied hearing-stage text, which still reads as the same individual workflow. The packet repeatedly uses “Go there,” “Go to this item,” “Read the full card” and “Open original source”; by contrast, “Reopen MW1 & sources” gives a useful specific destination. The glossary defines “Political option” as an available action rather than a menu choice (`Player_Screens.txt:3485`), but “advantage” and “actor” already drive the entry question (`10`).

**Risk.** A learner who does not proactively open help can miss the role swap or the technical meaning of a familiar word. Repeated generic control names may also be hard to navigate using a link/control list, speech commands or memory, depending on actual accessible names and grouping. The text alone does not establish that programmatic context is absent.

**Practical fix.** Reuse the existing help text at the point of need: a small pair-mode cue before question 2 and brief linked definitions for the core question's terms. Keep full help and glossary available throughout. Prefer explicit destinations such as “Go to the complication reason” and “Open MW-C original report,” while retaining a compact visual presentation if necessary.

**Check next.** Check help availability on every stage, keyboard opening/closing, focus return, heading and dialog names, control names in isolation and in context, and text at magnification/narrow widths. One entry screenshot cannot certify any of these.

### R11-09 — Finishing, saving, handing in and closing reflection need a clearer end-state contract

**Severity: Medium.**

**Evidence and location.** Entry offers “For class · discussion and a short explanation” versus “Just for fun · explore and keep your evidence board” (`13–14`), and says progress saves on this device (`20`). The finish stage offers “Finish this hearing record” (`2583`), then explains HTML versus JSON (`2586`) and offers several export choices. Only after those controls comes “Close the hearing aloud or in writing” with a 40-second briefing or rehearsal note (`2593–2594`). The initial final-work list does not mention that closing activity (`1800`).

**Risk.** The file-format distinction is clear and useful. The remaining uncertainty is procedural: what does Finish actually do, which artifact is intended for a reviewer, whether anything is submitted, and whether the closing reflection is required work or an optional debrief? A fun-mode user promised exploration may read the same formal checklist as mandatory; the supplied text does not establish whether mode changes behavior.

**Practical fix.** Match the Finish button to its actual action, state the visible resulting status, and provide one short mode-appropriate next step. For class, tell learners which format is suitable for reading and to follow the teacher's actual submission destination; keep JSON identified as the editable backup. Label the briefing/rehearsal as required or optional consistently and mention it in the opening contract when required. Keep the alternative of writing rather than speaking.

**Check next.** Test finish with both complete and incomplete records, each mode, and save/import/resume round trips. Inspect HTML, print and plain-text exports for the evidence, limits, comparison and revision record. Confirm whether the app submits anything before writing definitive submission language.

### R11-10 — Preserve the explicit completion boundary and publish the participation policy when it is used

**Severity: Low; conditional course-delivery improvement, not a claim that software grading is present.**

**Evidence and location.** The player says inclusion checks “cannot judge the quality, truth, originality, or logic” and “Included” never means correctness (`2555`, `3406`). The guide, pp. 3 and 5, separately defines a 100/70/0 participation scale and says these are participation judgments, not analytical mastery. Its p. 4 correctly distinguishes pilot observations from learning gains. The optional Vale question is immediate source inference practice, explicitly not a grade of the historical explanation (`2546–2553`).

**Risk.** These safeguards should be retained. If a teacher uses the suggested numerical scale without sharing the distinction with students, “100” can still be understood as mastery while strict filing lengths look like grade criteria. A later improved paragraph or a successful immediate Vale response cannot establish learning caused by the activity.

**Practical fix.** When the participation scale is used, present a brief student-facing version before work starts, with the human quality criteria from R11-01. Keep readiness, substantive participation and analytical quality separately named. Maintain all reasoning requirements while handling course accommodations explicitly; do not turn length, confidence or practice retries into an automatic score. Keep any learning evaluation separate from the record-completion UI.

**Check next.** In a human pilot, ask learners what “finished” and “100 participation” mean. Review first attempts and retained/narrowed/replaced decisions. Use a delayed new application and an appropriate comparison design before making claims about retention, transfer or learning effects.

## Participation, completion and learning: what the evidence can mean

| Construct | Appropriate evidence | What it cannot establish |
| --- | --- | --- |
| Completion / record readiness | Required items exist; lengths and IDs satisfy filing rules; confirmations are recorded | That observations are accurate, links are sound, writing is original or the source limit addresses the claim |
| Participation | A human sees a substantive attempt at the evidence, connection, challenge, alternative and comparison operations | Analytical mastery; a preferred conclusion; a learning effect |
| Analytical performance | A human applies the substantive rubric to the explanation, comparison and evidence request | That the activity caused the performance or that it will persist |
| Learning / durable transfer | Baseline and later performance on new material, with suitable design and delayed evidence when relevant | Supported by clicks, finishing, retry success, one immediate multiple-choice item or this AI review |

## Coherent next actions

1. **Align the student task contract and quality standard.** Add the compact fictional calibration and rubric; make Malawi's chosen explanatory step explicit; keep the Serbia rule route scaffolded; standardize required labels. Preserve the existing word-bounded written work and comparison.
2. **Make source work claim-specific.** Put the relevant limitation first, retain full provenance and conflicting totals, and distinguish source-detail access from actual reading or independent corroboration. Track source coverage for the evidence used without claiming semantic validation.
3. **Make revision and closing actions explicit.** Clarify what each hearing decision changes, surface existing role help at the relevant moment, use meaningful destination labels, and specify the actual finish/save/hand-in behavior.
4. **Then run real interaction and classroom checks.** Include solo and pair learners, learners with varied prior political knowledge, and users of relevant assistive technology. Verify keyboard/focus, dialogs, field labels, errors, link names, table relationships, magnification/reflow, persistence and readable exports. Separately inspect whether learners produce distinct observations, genuine action/options links, relevant limits and comparisons that test a claim. Do not substitute this content review for those checks.

**Overall judgment:** the instructional purpose is coherent, with unusually explicit boundaries around measurement and software checks. The strongest remaining content improvements concern making the quality standard available to the student, maintaining alignment across legitimate explanation routes, and attaching each source limit to the claim it actually qualifies. The permitted materials support these risk judgments; they do not demonstrate actual learner failures or accessibility conformance.
