Middle school · Grade 7 / Critical thinking / 30 minutes
The Evidence Room
Look closer. Build a defensible conclusion.
Plan your lesson
Teacher notes
Inspect competing sources, weigh their relevance and strength, and explain how well the evidence supports a claim.
The full guide below comes from the artifact’s original README. Its development and validation notes describe the original build.
Download teacher notes (.md)The Evidence Room
Open index.html in a modern browser. No install, internet, account, server, API, or key is needed. Keep index.html, styles.css, cases.js, and app.js together. All materials and statistics are fictional teaching examples, not evidence about actual schools or scientific findings.
1. Learning objective
Seventh-grade students evaluate how strongly evidence supports a claim. They distinguish relevance, evidence strength, and trustworthiness; inspect methods and sample sizes; identify alternative explanations and missing information; compare sources; and justify a qualified conclusion.
The app is an investigation rather than a right-answer quiz. It has no source score, approved ranking, or preferred verdict. A small study can be directly relevant but limited; an expert can be trustworthy yet indirect. Card counts describe students’ placements, not the truth of the claim.
The default school-start case is intentionally ambiguous: a larger comparison study reports a modest reading gain but not a math gain, and tutoring complicates the result; a small local trial is positive but uncontrolled; a district report shows little change in learning; and a vivid anecdote has another plausible explanation. The fertilizer case examines the universal word “always” through dose, soil conditions, and counterexamples.
2. Suggested 30-minute classroom activity
- 0–3 minutes — Read the claim. Ask students to withhold agreement or disagreement. Their job is to judge the support in these documents.
- 3–14 minutes — Inspect sources. In pairs, read all six documents, rate the three dimensions, and explain the ratings using specific details. Discuss why reputation alone does not determine evidence strength.
- 14–19 minutes — Organize the board. Place each source under supports, challenges, or uncertain. Use drag/drop or the labeled placement menus. If a source cuts both ways, students can put it under uncertain and explain why in its notes. Revisit and move sources as thinking changes.
- 19–25 minutes — Make a case. Select a verdict; explain a conclusion; identify and compare two sources; describe a limitation; and answer the reflection question. All active sources need notes and a board placement before submission.
- 25–30 minutes — Debrief. Open the optional prompts below the feedback. Ask which source changed their thinking, how they handled disagreement, and what information would help next.
The AI news brief is initially unsigned. Saving its first annotation reveals its AI origin and the deliberately invented statistic. Students must save a verification plan before submitting their verdict. They can revise ratings afterward. The lesson is to trace claims and check evidence from every source, not to reject AI by default.
Feedback uses source tags and simple wording checks to generate revision prompts. It is not AI assessment, a grade, or a reliable evaluation of semantic reasoning. Review explanations with students. A defensible explanation connects the exact claim to specific results, compares sources, considers alternative explanations, and explains limitations. Length checks encourage a short written explanation but cannot ensure thoughtfulness.
3. Create a completely new Evidence Room
In the app
Expand Teacher settings. Choose an example and select Load Example Case, or expand Edit case configuration (JSON). Replace any or all fields, then select Apply configuration. Invalid configurations leave the current case intact. Loading a case or toggling its AI source resets student work after confirmation when work exists.
The editor is for teacher setup, not a protected area. Students can open teacher notes. Hidden strengths and limitations are omitted from student documents but available in Teacher settings and source code. This is a teaching tool, not a secure assessment.
In the files
Edit cases.js in a text editor. window.EVIDENCE_CASES is an array of configuration objects. The first object is the default case for a fresh browser. To make a new case permanent and portable, replace that first object. An already-open browser may restore its previous local case; use Load Example Case to load the edited file’s case.
Each case has:
| Field | Format and purpose |
|---|---|
id |
Nonempty case identifier string |
title |
Case-file title |
subject |
Subject label: English, science, social studies, health, etc. |
centralClaim |
Exact claim under investigation |
studentInstructions |
Student task and framing |
includeAI |
true or false; hides cards with ai: true when false |
finalReflection |
Final required open-ended question |
discussionQuestions |
Array of teacher debrief question strings |
evidence |
Array of 2–30 source objects, with at least two non-AI sources |
Every source includes:
| Field | Purpose |
|---|---|
id |
Unique string within the case, e.g. "01" |
type, title, preview |
Card label, document title and short preview |
source, date, author |
Visible source metadata; use “not supplied” when appropriate |
body |
Actual evidence/document text |
method |
Methods, sample, comparison and measurement details |
limitations |
Visible information gaps; reserve additional teaching interpretation for hidden notes |
strengths |
Array of teacher-only strength strings |
hiddenLimitations |
Array of teacher-only limitation strings |
tags |
Array of feedback tags, or [] for generic feedback |
ai |
Optional boolean; mark AI-generated evidence |
aiReveal |
Required string when ai is true; explain origin and what is unsupported |
Supported feedback tags: anecdote, small, study, indirect, selective, unsupported, confound. The first five tailor feedback when that source is selected as strongest. unsupported and confound describe instructional issues for teachers and future extensions; they do not assign scores. All imported text is rendered as text, not executable HTML.
To repurpose the tool for English, use a claim about a character’s motives and excerpts with context. In social studies, compare accounts of an event and gaps in records. In health, compare reports of a habit with controlled evidence. Preserve genuine tensions; do not make the source name an answer key. If a source labeled study does not involve sample-size reasoning, omit that tag and use generic feedback, or adapt renderFeedback() in app.js.
Example source
{
id: "01", type: "Experiment log", title: "What changed?",
source: "Class science notebook", date: "May 12, 2025", author: "Group B",
preview: "Two groups produced different results.",
body: "Describe actual observations here, including exceptions.",
method: "Describe sample, comparison, controls and measurements.",
limitations: "Explain what the document does not report.",
strengths: ["A relevant comparison."],
hiddenLimitations: ["A variable was not controlled."],
tags: ["small", "confound"]
}
Accessibility, storage and classroom use
Native buttons, radio groups, labels, modal focus handling and keyboard-accessible placement menus support keyboard and touch use. Dragging is optional. Layout adapts to tablets and phones and respects reduced-motion preferences. Browser zoom can enlarge the workspace for projection.
Saved annotations, placements, verification plans and verdict drafts stay in local browser storage when available. Nothing is sent to a server. Some browsers restrict storage for local files: keep the tab open in that case. Unsaved source notes are not retained until Save source notes is selected. Use Reset student investigation between groups on a shared device. There is one saved investigation per browser/origin; switching cases replaces it. Use the supplied ZIP to share the app, then extract it before opening index.html.
Behind the activity
The original AI prompt
Generated in ChatGPT using 5.6 High, then run in ChatGPT Codex using GPT-6 Astra Medium. The text is preserved as supplied; it describes the requested build, rather than a guarantee that every requirement was implemented exactly.
Read the full creation prompt
Build a polished interactive educational web app called “The Evidence Room” for middle-school students in seventh grade. GOAL Teach students to evaluate claims using evidence. Students should learn that evaluating information involves multiple dimensions: - Who is making the claim? - What evidence is provided? - How directly does the evidence address the claim? - How strong is the evidence? - What might be missing? - Does another source support or contradict it? This should feel like an investigation or detective case, not a quiz. TECHNICAL REQUIREMENTS - Build a self-contained browser-based app requiring no login, backend, API key, installation, or internet access. - Prefer index.html plus CSS and JavaScript. - No external APIs. - Make it highly usable on a classroom projector, Chromebook, laptop, or tablet. - Use accessible typography and controls. - Keep the interface visually sophisticated enough that middle-school students do not perceive it as an elementary-school game. OPENING EXPERIENCEOpen with: THE EVIDENCE ROOM“Can you decide what the evidence actually shows?” Give students a claim to investigate. For the default case, use a fictional school-related claim that is interesting but nonpolitical, such as: “Starting school one hour later would improve student learning.” Make clear that the task is NOT to agree or disagree initially. The task is:“Decide how strongly the available evidence supports the claim.” THE EVIDENCE DESKDisplay six evidence items as visually distinct documents/cards. Use fictional but realistic examples representing different evidence types, such as: - a student social-media post - a school survey - a local news report - a scientific research summary - an expert interview - a school attendance/data report The evidence should NOT be designed so that one source is obviously “good” and another obviously “bad.” Introduce realistic complications: - a credible source that is only weakly relevant - a highly relevant source with a small sample - correlation that does not establish causation - an anecdote that is emotionally compelling but weak evidence - a study with useful evidence but limitations - a source that selectively presents information SOURCE INSPECTION When a student opens a source, let them inspect: - source - date - author/organization - evidence presented - methodology or sample, where appropriate - possible limitations Students should be able to annotate each source using categories such as: RELEVANCE Low / Medium / High EVIDENCE STRENGTH Weak / Moderate / Strong TRUSTWORTHINESS Low / Medium / High Then require a short justification:“What made you rate it this way?” IMPORTANT:Do not automatically equate prestigious sources with strong evidence.Do not teach “website A is credible, website B is not.”Teach students to evaluate the actual evidence and reasoning. EVIDENCE BOARD Create an investigation board where students can organize evidence into: SUPPORTS THE CLAIM CHALLENGES THE CLAIM UNCERTAIN / NOT ENOUGH INFORMATION Allow them to drag evidence cards among the categories. As evidence accumulates, visually show that conclusions can become: - stronger - weaker - more uncertain FINAL VERDICT At the end, require the student to choose: Strongly supportedSomewhat supportedEvidence is mixedSomewhat unsupportedStrongly unsupportedNot enough evidence Then require: “My conclusion is ______ because the strongest evidence is ______.” and “One important limitation or unanswered question is ______.” Do not tell students that only one verdict is acceptable if multiple conclusions could reasonably be defended. Instead, evaluate the QUALITY OF THEIR REASONING. Provide feedback such as:“You relied heavily on one anecdote.”“You noticed that the largest study supports the claim, but also identified its limitation.”“You classified this evidence as relevant, but did not explain why.” AI-ERA EXTENSION Include one evidence item generated by an AI system. Do not label it initially as “bad.” Its content should sound plausible but include a subtle unsupported statement or invented statistic. After students analyze it, reveal:“This source was generated by AI.” Ask:“What should you verify before relying on it?” The lesson should be that AI output must be evaluated using the same evidence standards—not that AI is automatically unreliable. TEACHER MODE Include a collapsible Teacher Settings section. Make it easy for a teacher to replace the entire case using a configuration object. Teacher-customizable fields should include: - case title - central claim - student instructions - evidence cards - evidence metadata - hidden strengths/limitations of each source - recommended discussion questions - whether the AI-generated evidence card is included - final reflection question Include a button:“Load Example Case” and provide at least one alternate built-in case in a different subject, such as:“Does adding fertilizer always make plants grow faster?” This is important because I want teachers to immediately understand that the SAME TOOL can be repurposed for English, science, social studies, health, and other subjects. TEACHER DEBRIEF After the student submits a verdict, provide optional teacher discussion prompts such as: - Which source changed your thinking most? - Is the most trustworthy source necessarily the most useful? - What evidence would you want before making a decision? - Can two reasonable people examine the same evidence and reach different conclusions? - What does “not enough evidence” mean? POLISH Make the interface visually resemble an investigation workspace: - evidence folders - cards - subtle connecting lines - investigation board Avoid cartoon detective clichés.Make it modern and intellectually serious. Before finishing: - run and test the artifact - make sure drag/drop or alternative accessible controls work reliably - confirm that the default case contains genuine ambiguity - ensure students cannot succeed simply by guessing the “approved” source - improve the UI wherever needed Finally, create a README explaining: 1. the learning objective, 2. a suggested 30-minute classroom activity, 3. how a teacher can create a completely new Evidence Room by editing the configuration.
Your next experiment
Make it your own
Start with Teacher Settings inside the activity. For a larger change, copy the original prompt and change the grade, subject, learning goal, and scenario. Keep the prediction, evidence, explanation, and revision cycle.
Ask the AI to explain the changes and the model’s assumptions. Try the revised activity yourself before teaching with it, including unusual inputs and the classroom devices students will use.
These adaptation suggestions were added for this resource hub; they are not part of the original prompt.