# S2 — independent first-use pressure test

## Method and result

I simulated an average 18-year-old who skims, knows little about elections or courts, and wants to finish. Assignment: **Malawi; For class; Solo**. I chose Russia because its continuing president contrasted with Malawi's changed winner. I used only the rendered game at `http://terminal.local:4173/?reviewer=S2`, visible cards, source details, and glossary. No source code, hidden answers, other reviews, design history, or outside facts were consulted.

**Actual outcome: filed.** The interface displayed **“Hearing filed · interpretation awaits discussion.”** My first attempt returned **“3 inclusion items still need review. Your draft can be exported now.”** Recovery worked without help. Completion was nonempty, but final prose retained an inference I had supposedly narrowed elsewhere. No product failure blocked completion. Automated interaction speed does not establish human timing.

## First ten actions and the skimming route

These are task actions, excluding observations and screenshots:

1. Opened the assigned S2 URL.
2. Selected Malawi, leaving the already selected For class and Solo settings.
3. Clicked **Enter the hearing** without opening How to play.
4. Selected **A new contest was enforced**. My simple interpretation was “courts made them vote again.”
5. Clicked **Record my opening view**, leaving confidence Provisional.
6. Clicked **Choose evidence**.
7. Used MW3 as Anchor 1 from its short preview, without opening its full card.
8. Used MW5 as Anchor 2, also from the preview.
9. Used MW6 as Complication because “higher share still ended in defeat” sounded like the awkward fact the task wanted.
10. Clicked **Make a connection** with zero full anchor-file cards or source records opened and all exhibit reasons blank.

**Observed interface facts:** The exhibit showed 3/3, and all sources remained available. The opening was retained separately from an editable working view. No “you have read everything” claim appeared.

**Simulated reaction:** I understood the broad task as explaining an election result, but initially collapsed two questions: why a second election happened and why Chakwera won it. The neat previews let me feel ready before doing close reading. The three-slot structure was easy to follow; “foreground” sounded like “pick the three important cards,” rather than an instruction to examine the rest.

## What happened next

I built the chain using starters: **Courts → Ordered a fresh election and required a majority → Voters had another contest → A different winner in June 2020**. I attached MW3 to the action and MW5 to the changed option, wrote 66 words, and moved on.

The first hearing question usefully confronted my exact conceptual shortcut: **“Opening a contest and winning it are different steps.”** I chose the answer distinguishing rulings, implementation, and voter choices. The alternatives about judges directly appointing Chakwera and other institutions being irrelevant were plainly extreme. I could select the careful answer without reading the records behind it.

I then selected **Narrow a step**, wrote that MW3 did not explain why people chose Chakwera, changed **Possible consequence**, and applied the decision. The interface confirmed the changed board step. I answered the second figures question by choosing its cautious option, recorded a defense about reported percentages, and selected **Narrow the reach of my opening view**. I did not revise the evolving prose: the separate decision felt like I had done the revision.

For Russia, I opened RU1's full card and source details, used RU1, and wrote the comparison below. At filing, the three outstanding items were the complication reason, an opened Malawi card source record, and the short limitation field. Filling the two notes and using **Reopen MW3 & sources** solved these. That link opened the full card and source details together. I also read MW5 and MW6 in full, used the glossary, and returned directly through the numbered navigation. Nothing had been lost.

## Language that caught me

| Exact phrase | Simulated interpretation / uncertainty | Recovery observed |
|---|---|---|
| “required a majority” | I initially read this as “whoever gets the most votes.” | MW3's full text explicitly said more than half of valid votes. |
| “Seat and majority” | The glossary defined majority as more than half **the seats**, so it did not directly resolve Malawi's presidential threshold. | The card, rather than this glossary entry, supplied the relevant meaning. |
| “joint opposition ticket” | I guessed a combined team; I did not initially know president/running-mate ballot arrangements. | Malawi's introduction and MW4 preview made the basic pairing understandable. |
| “denominator” | A maths word I skimmed rather than a usable instruction about who was counted. | Russia's table's “Who answered” column was clearer than the warning sentence. |
| “the two 2020 series” | I did not know which two records were meant when the hearing first mentioned them. | MW3's expanded source details named the report and separate polling-station publication. |
| “What does the narrower step leave open?” | I had to translate this into “what am I still unable to explain?” | “Which connection changes?” plus Possible consequence made the edit manageable. |

## Prioritized issues and fixes

**1. High — a recorded revision can leave final prose untouched.** Observed: my narrowed consequence said the new vote did not explain the winner, but my final account still said **“This helped Chakwera win instead.”** This is an unreconciled emphasis, rather than proof of a logical contradiction. All checks passed and filing succeeded. The game clearly says prose is not graded; this is not a semantic-grading bug. A skimmer can mistake applying a decision for updating the account. Add an explicit handoff after Apply: **“Your connection changed. Your paragraph still needs your edit.”** Show the paragraph beside the changed sentence or offer a focused return to its editor. Preserve student authorship; do not automatically generate the paragraph or pretend to judge its logic.

**2. High — the source-reading requirement arrives late for a preview-driven player.** I reached filing having answered both challenges without opening a Malawi source record. Filing appropriately stopped me, but the easiest recovery was opening one source and filling short fields. Make the next reading action concrete earlier: “Open one selected card's full text and Source details before building your explanation.” A visible selected-card reading checklist would support this without locking exploration. Keep the source dates, conflicting totals, and uncertainty; reduce the navigation burden rather than the evidence.

**3. Medium — hearing density encourages choosing the careful-sounding answer mechanically.** Two long challenges, separate answer submissions, exhibit decisions, explanations, a revision selector, and the evolving paragraph created the most tiring stretch. Present one challenge at a time with a compact progress indicator. Retain the full proposed claim and source access. Make distractors plausible bounded errors rather than obvious absolutes, so the decision depends more on reading.

**4. Medium — a few definitions miss the task-specific gap.** Add presidential majority versus largest share, define denominator as “the people or votes this percentage is out of,” and identify the two MEC publications directly in the question. This preserves precision while making the cautions usable.

**5. Low — reading counters looked stale.** The Malawi evidence screen still said **“Full cards opened: 0 of 6 Card source records opened: 0 of 6”** while MW3's full text and **“Source details · opened”** were visible and MW5/MW6 were expanded. The later filing check correctly recognized a source record opened. This was confusing feedback, not a completion blocker.

## What worked and actual completion sample

The cream, dark-green, serif-heading layout looked credible and calm. Numbered stages and the persistent exhibit gave reliable orientation. The known outcome, incumbent/opposition definitions, and permission to revise reduced fear of a “wrong” opening. Source details clearly distinguished event dates from upload dates. Reopening evidence and preserving drafts worked well. Narrowing courts' role was meaningful; selecting previews and supplying repeated short notes felt more mechanical.

**Filed Malawi account — interface counted 66 words:**

> The courts made a big difference because they ordered another election. The president could not just keep the first result once there had to be a fresh vote (MW3). Parliament then set a date so voters got another chance (MW5). This helped Chakwera win instead. The awkward part is that Mutharika's share actually went up, so the old president did not just lose all his support.

**Filed comparison — interface counted 39 words:**

> In Malawi, courts let challengers get another election. In Russia, the commission refused to register Nadezhdin (RU1), so voters could not choose him. This suggests institutions can open or close a route, but does not prove he would win.

Separate limitation: **“The 2019 figures were set aside and cannot show a fair count.”** Next evidence: **“What would support for Nadezhdin look like if he could actually campaign?”**

**Verdict:** A skimming novice can finish and learn a real distinction. Recovery from missing requirements works. The main risk is a complete record whose paragraph has not caught up with its revision. A 55-minute lesson depends on deliberate reading and discussion; this simulation cannot establish human lesson timing.
