Professional learning · Leaders, Teachers, Faculty, Coaches · 70 minutes
AI Ethics Case Court
Educators put fictional AI uses in grading, proctoring, detection, and monitoring on trial, then issue rulings with conditions that could become real policy.
Schools and colleges are already making consequential decisions with AI: scoring writing, flagging suspected cheating, monitoring student devices, predicting who might drop out. These uses can save time and catch real problems, and they can also harm the students with the least power to push back. In this mock-court session, small courts argue fictional cases using an ethical principles bench card, then rule "allowed," "allowed with conditions," or "not allowed." The conditions they write become a draft of practical guidance for their own campus.
In this packet
- Handout A: Case Files (cut-apart cards)
- Handout B: Ethics Bench Card (organizer)
- Handout C: Ruling and Remedy Form (worksheet)
- Facilitator key (last page)
TCEA ELE indicators
- AI2.2 I can identify potential ethical issues in AI applications and propose solutions.
- AI2.1 I know how to apply ethical principles when using or developing AI for education.
- AI2.3 I am aware of the importance of transparency and accountability in AI systems.
- AI1.3 I am aware of the current limitations and potential biases in AI systems.
Full facilitator guide: mglearn.github.io/eles/activities/ai-ethics-case-court.html
Handout A for AI Ethics Case Court
Case Files
All people, schools, and products are fictional. Backs are the facilitator key: issues a strong court should surface.
Case 1 · AI-assisted essay scoring
At Maple Hollow High, an English teacher with 160 students uses an AI tool to score essays on a rubric and draft comments. She spot-checks about one in ten. Students aren't told AI was involved. Scores go straight into the gradebook.
Case 2 · AI-writing detector
Cedar Ridge Middle School runs every essay through an AI-writing detector. Luisa, who moved from Guatemala two years ago, is flagged at "high likelihood AI." Policy gives an automatic zero unless she can prove she wrote it. She has no drafts saved.
Case 3 · Remote proctoring (higher ed)
Bluebonnet Community College requires webcam proctoring software for online exams. It flags "suspicious behavior" such as looking away or background noise. A student with a tic disorder is flagged on every exam; another student's face isn't consistently detected in dim lighting in a shared apartment.
Case 4 · Device monitoring
Pecan Valley ISD uses AI software that scans students' searches and messages on school accounts for signs of self-harm or violence, alerting campus staff 24/7. It flags a message: "I'm going to shoot my shot at tryouts." A counselor and an officer visit the student's home at night.
Case 5 · Dropout prediction
Riverbend ISD adopts a model that predicts which ninth graders are "at risk" of not graduating, using attendance, grades, discipline, and zip code. Every teacher sees a red, yellow, or green label next to each student's name.
Case 6 · Faculty letters (higher ed)
A professor at Bluebonnet Community College writes 40 recommendation letters a year. She pastes students' names, grades, and personal stories into a consumer AI chatbot to draft them, then edits lightly. Students don't know.
Case 7 · AI tutor chatbot
Lone Oak Elementary pilots a fictional "BrightPath Tutor" chatbot for 4th grade math. It helps students who are stuck. The contract lets the vendor keep student conversations to "improve the product." Families were sent a one-line notice in English only.
Handout B for AI Ethics Case Court
Ethics Bench Card
Judges: use the questions to question both sides. Mark the principle that weighs most in your ruling.
| Principle | Questions the bench asks | What we heard (notes) |
|---|---|---|
| Learning benefit | What does this do for students' learning or wellbeing? What evidence shows it? | |
| Privacy | What data is collected, who sees it, how long is it kept, and could it be used for something else? | |
| Fairness | Could it work less well, or cause more harm, for some groups (language, disability, race, income)? | |
| Transparency | Do students and families know AI is involved and roughly how it works? | |
| Accountability | Who is responsible when it's wrong? Is there a real appeal? | |
| Human oversight | Does a person make the final decision, with time and training to do it well? | |
| Voice | Were the people affected (students, families, faculty) consulted? |
Handout C for AI Ethics Case Court
Ruling and Remedy Form
The bench completes this aloud during deliberation. Conditions must be specific enough that someone could check whether they were followed.
Case and verdict (circle one): ALLOWED · ALLOWED WITH CONDITIONS · NOT ALLOWED
The principle that weighed most in our ruling, and the case detail that tipped it:
Strongest argument from the side that lost:
Conditions (who does what, when). At least three, or explain why no conditions would be enough:
What the affected student can do if they believe the AI got it wrong:
Who might still be left out or harmed under our ruling?
Facilitator only for AI Ethics Case Court
Answer key and notes
Handout A: Case Files
- Case 1 · AI-assisted essay scoring
- Issues: transparency (students not told), accountability (whose score is it?), human oversight (90% unreviewed), fairness (does it score non-standard dialects or unusual structures lower?), learning benefit (faster feedback is real). Strong conditions: disclose use, teacher reviews every final score, AI comments are drafts only.
- Case 2 · AI-writing detector
- Issues: AI-writing detectors are known to be unreliable, and there are published concerns about false positives for non-native English writers. Burden of proof falls on the student. Strong rulings: detector output alone never triggers a penalty; evidence comes from process (drafts, conversation, in-class writing).
- Case 3 · Remote proctoring (higher ed)
- Issues: privacy (video of home), fairness (disability, lighting, skin tone and camera concerns raised by students and advocates), accommodations, access (quiet private space isn't equally available). Conditions: flags reviewed by a human before any action, accommodation path, alternative assessment options.
- Case 4 · Device monitoring
- Issues: student safety is a genuine and serious goal; also privacy, false positives, who receives alerts, the escalation path, and chilling effects on student expression. Conditions: trained human triage before any escalation, clear scope disclosed to families, data retention limits, review of false-alarm patterns.
- Case 5 · Dropout prediction
- Issues: using zip code can encode inequities; labels can shape teacher expectations; can students or families see or contest the label? Benefit: earlier support. Conditions: remove proxy variables, show labels only to support staff, use the prediction to offer help, never to limit opportunities; review accuracy by student group.
- Case 6 · Faculty letters (higher ed)
- Issues: privacy (student records entered into a tool without an institutional agreement), transparency, authenticity (a letter carries the writer's personal judgment), bias in generated praise words. Conditions: use only institution-approved tools, no identifying details, disclose, the professor writes the substantive judgments herself.
- Case 7 · AI tutor chatbot
- Issues: data use by the vendor, age of users, meaningful family notice (language access), whether the tutor gives answers instead of hints, evidence of learning. Conditions: contract limits on data use and deletion, notice in families' languages, a teacher monitors, measure learning, not minutes used.