Make Them Teach It

• Source: Mike Kentz Substack
• Author: Mike Kentz
• Original source: https://mikekentz.substack.com/p/make-them-teach-it
• Published: 2026-07-29
• Source type: essay

Summary

Mike Kentz describes a calculus-class assignment where students had to teach the idea of derivative to a deliberately confused AI student named Devon. The article argues that AI teach-back simulations can make conceptual understanding visible because students have to explain, adapt, and clarify in response to targeted resistance instead of merely producing answers.

Kentz frames the assignment as a concrete extension of his broader transcript-based assessment work. Its distinctive contribution is showing that conversational simulation is not only an English-class or writing exercise: it can also expose conceptual understanding in math when the AI learner pushes back in content-specific ways.

Pull quotes

The student has to teach, not just answer

“They had to explain what a derivative actually means to a classmate named ‘Devon,’ who kept almost getting it and then slipping back.”

Resistance creates evidence

“What I’ve tried to add to the process is targeted resistance, a record, and an evaluation mechanism.”

The transcript becomes the artifact

“The conversation leaves an artifact that a teacher can read and evaluate for understanding afterward.”

Big ideas

• AI simulations need clear boundaries for learning
• Voice AI may make learning support easier to access
• Learning still needs some struggle, even when AI can make things easier

Claims

• In an AI world, assessment should focus on watching students think
• AI chat transcripts can make student thinking visible
• AI-built models can help students show conceptual understanding

Key evidence and examples

• A Connecticut AP Calculus AB teacher used an AI student persona with a specific misconception about derivative as slope.
• Kentz and the teacher intentionally designed the bot’s confusion so students had to clarify meaning rather than recite procedure.
• The article links the assignment to the protégé effect and prior work where students learn more when they believe they are teaching.
• Kentz argues that the important additions are targeted resistance, transcript preservation, and a teacher-readable evaluation artifact.
• The piece explicitly positions this as a STEM example to counter the assumption that AI dialogue assessment belongs only in humanities classrooms.

Education relevance

Very high relevance for authentic assessment, math pedagogy, simulation-based learning, transcript evaluation, and subject-specific ways to make thinking visible in an AI era.

Durability note

The named bot and platform will age, but the durable pattern is likely to persist: asking students to teach a resistant AI learner can surface conceptual understanding in ways that static answers and polished products cannot.

My notes

• Strong narrower case of Kentz’s broader observable-cognition argument because it shows how teach-back plus resistance can work in math, not just writing-heavy subjects.