How to encourage smarter AI use in the classroom

When large‑language‑model chatbots entered the consumer market, many teachers found students able to generate essays or solve problems with a few taps, creating a new source of workload as educators had to detect AI‑generated work and adapt assignments. The lack of a unified policy left instructors uncertain about permissible use, even as organizations such as OpenAI and UNESCO promoted broader adoption.

At Cheshire Academy, a private school in Connecticut serving roughly 400 students in grades 9‑12, administrators chose not to mandate any specific AI tool. Instead, following consultant advice, they provided staff training that covered prompt engineering, model limitations, and the risk of biased or incorrect outputs. Teachers now employ a variety of platforms—general‑purpose chatbots like ChatGPT and Perplexity, and the education‑focused MagicSchool—to draft lesson plans, create rubrics, and brainstorm activities. While some use AI to prepare feedback, concerns about quality, personalization, and privacy have kept that practice limited. French teacher Miriam Przybyla‑Baum, who relies on her own long‑standing materials, discusses AI with students who attempt shortcuts using tools such as Google Translate.

Key takeaways

• Training emphasizes prompt crafting and awareness of model fallibility rather than prescribing specific software.
• A mixed toolkit (ChatGPT, Perplexity, MagicSchool) is used primarily for instructional preparation, not direct student feedback.
• No school‑wide AI mandate; adoption remains voluntary and instructor‑driven.

Why this matters

The school’s strategy of investing in general AI literacy instead of locking into a single product reflects an evidence‑based inference that understanding model behavior—such as propensity for hallucination and bias—helps educators integrate AI responsibly while mitigating misuse. By focusing on preparation tasks where output can be reviewed before classroom use, the approach balances potential efficiency gains with safeguards against overreliance on unverified AI‑generated content.