Whose Terms? — See how power works

Under the Map — How an Occupation Becomes a System

Review, revision and human-pilot guide

This revision was examined through six independent AI review streams: novice users for each game, academic faculty, learning and assessment, adversarial software/usability, and art direction. These are simulated perspectives and executed software checks, not human participants, faculty endorsement or measured learning effects.

What changed after the review

Problem foundRevisionEvidence to review
Changed assumptions could leave an old finding looking current.Earlier findings are retained and marked for recheck; the current completion count excludes them.Change an input after keeping a finding, rerun, then keep a revision.
An unapplied or edited amendment could be described as tested.The current draft must be applied. The report retains the decision-maker, trigger and review mechanism.Amend a trigger, apply and compare the original and revised record.
Several tasks tested recognition of supplied text or clicks.A new court-file source application and a prediction–test–revision dependency experiment preserve the learner’s first and later judgments.Inspect the initial/final markers or predictions and the chosen evidence/reason.
Missing background and overly broad inferences.More novice definitions, verified source wording, wider institutional context and distinctions between design, rights, political authorization and implementation.Open the context panels and ask what further evidence would change a conclusion.
Argument edits and some intermediate citations were lost from history.Explicit explanation snapshots preserve versions; every recorded source version stays attributable in HTML and text reports.Keep two snapshots with different explanations or evidence roles.
Keyboard, phone and recovery risks.Controls retain focus after scene updates; essential tools stay visible on phones; abandoned definitions close; blocked or unreadable saves have a visible recovery route.Use keyboard only, a phone-width viewport and a recovery copy.
The visual system felt too corporate.Conceptual relief-print covers, paper/ink treatments, annotated layers and a political field-journal style.Check legibility and discoverability, not visual preference alone.

Follow-up browser checks · 26 October 2026

Both revised games were opened and exercised in Chrome after the earlier browser-service failures. The check covered cover artwork, selected timeline labels, navigation, an evidence/source return, context, the prediction–comparison flow, class entry, the short reflection and the incomplete-report label. Phone-width layouts were also inspected in a 390-pixel frame: the checked pages fit without horizontal page overflow and essential toolbar controls remained visible. This is a responsive browser check, not a test on a physical phone or a complete accessibility audit.

The check found and corrected selected timeline text that could disappear against its own background, misleading hover movement on disabled controls, and a save order that could mark a newly authored record for recheck immediately. Earlier records and genuinely changed findings remain available for review. The report download created a readable text file; the browser automation’s download-event wait timed out despite the file arriving.

Information, learning and play review · 27 October 2026

Six further independent AI review streams examined novice information needs in each game, academic source scope, geography and data, educational assessment, and gameplay. They mapped all sixteen encounters from question to supplied facts, comparison method, inference limit and the further evidence a wider conclusion would need.

Reviewer findingImplemented change
The wider political question could be confused with one corridor’s operating test.A core two-state definition and a short historical/legal starting position separate current control, proposed arrangements and implementation. Every encounter names what it tests and leaves open.
Sources were available, but the method for evaluating them was uneven.Every encounter now has an assessment aid with included facts, a comparison method, a claim limit and a specific next-evidence question. Help returns after source inspection.
A first interpretation could contradict an included factual record and still look like a validated finding.Known factual overclaims require revision before keeping a current finding. Original attempts remain, while supported uncertainty and interpretive alternatives stay open.
A historical or institutional comparison could be saved without comparing the required evidence.Both Gaza and Homesh records must be inspected; the proposal comparison requires a source-linked contrast or a precise operating gap.
The game’s objective could stop matching the player’s current subtask.The next-move cue follows the player’s progress. The footer distinguishes moving to another lead from keeping this finding.
Spatial percentages could remain an abstract wording exercise.An optional dated table compares population shares with locality counts and preserves a first and revised inference. Categories, coverage and chart scales are explicit.

Game-review verdict: Under the Map works best as a documentary investigation; After the Line has the stronger hypothesis-testing loop. These are investigations and conditional systems tests, not emergent strategy games or political forecasts. Further improvement should deepen evidence-linked discovery and learner reasoning, rather than reward a political answer.

Educational-review verdict: The packet supports its bounded questions. It supplies neither every fact about the conflict nor sufficient matched data to settle every possible political arrangement. Original documents and optional online readings extend the investigation. Required decisions use included evidence; optional readings that need additional source material are labeled accordingly.

Earlier saved findings remain recoverable. A finding that predates a newly required evidence comparison is marked for recheck; its original record is retained. The final reflection remains about fifty words.

The service encounter now unfolds in two tests: the hypothetical readiness test first, then the dated fiscal comparison, which opens after the first run and can also be inspected earlier. The Zanuta account remains an optional, self-contained real-case route. In the new court-file aid, provisional reasoning comes before the final-decision facts; the facts are included once the decision is opened.

What a human pilot should test

  1. Ask students to think aloud without telling them the intended distinction. Record where they cannot explain their next action.
  2. Before the game and after it, give different short cases that separate a physical connection, permission, an institution’s capacity and actual performance. Do not reuse the game’s answer wording.
  3. For the historical game, ask what a source about one legal instrument or household can establish about a different claim.
  4. For the proposal game, ask what should change when one input changes and what should remain unknown. Ask for the mechanism, not a political preference.
  5. Check source use directly: can the student locate the support, explain its limit and name evidence that would alter the finding?
  6. Observe return visits, optional help, recovery, missing context and reading burden. Actual completion time and retention require observation; the hour is a design budget.

Assess reasoning

Look for source scope, a specific causal or institutional mechanism, a justified uncertainty and a reason for retaining or revising a claim. An opposite political conclusion can be well argued; a matching political conclusion can be poorly supported. A correct supplied choice alone is weak evidence of understanding.

Known limits

The new historical exercise is scaffolded near transfer, not independent archival research. The dependency experiment tests explicit assumptions, not real-world probabilities. Wider demographic, political-authorization and implementation conclusions require their own evidence. No claim is made that simulated reviewers establish empathy, learning, retention or accessibility for actual users.