How to encourage smarter AI use in the classroom
Aug 24, 2026, 7:20 AM · MIT Technology Review
A school’s traffic-light rules offer practical governance by making acceptable AI assistance explicit for each assignment rather than pretending prohibition is enforceable.
Why it matters
Cheshire Academy, a Connecticut private school with about 400 students, labels assignments green, yellow, or red. Green permits AI, red forbids it, and yellow permits specified tools or uses.
Teachers use general chatbots and MagicSchool mainly for plans, materials, rubrics, and administrative work. Concerns about quality, personalization, and privacy have held back AI-generated student feedback.
From the desk
The scheme treats AI use as a pedagogical variable rather than one moral category. Translation practice, brainstorming, and final assessment test different skills and should permit different assistance.
Signal Desk's read: assignment-level disclosure is more durable than blanket bans or blanket adoption. Take-home enforcement is weak, while unrestricted use can replace the practice an assignment was designed to produce.
The most useful exercises make students critique model edits and identify where assistance removes voice. That builds judgment instead of rewarding submission of generated text.
This is one small-school case, not proof of effectiveness. The source provides no measured learning outcomes or comparison group, and schools still need privacy, accessibility, records, and misconduct rules.
Context
Widely available chatbots increased teacher workload by forcing educators to redesign assignments and distinguish help from substitution. Cheshire Academy trained staff on general techniques and limitations rather than mandating one product.
A student council is also intended to lead discussion about healthy use.
Who feels it
- Teachers
- Clear labels reduce ambiguity but require explicit learning objectives.
- Students
- Advance notice makes permitted assistance legible and critique teachable.
- School leaders
- A classroom convention still needs institutional privacy and conduct policy.
What to watch
- Comparative learning-outcome evidence
- Consistency of yellow rules across classes
- Privacy safeguards for student feedback