Teachers designing AI tools often skip specifying who makes the final call.

Science teachers often face artificial intelligence tools that just generate text. But real science teaching requires uncovering student ideas, spotting mistakes, and using teacher judgment. Researchers wanted to know how teachers would design AI learning tools to handle this real teaching work. They studied four app blueprints made by science teachers during a training workshop. The apps included a chemistry prior-knowledge checker, a physics scaffolding tool, a dialogue guide for Kinetic Particle Theory, and an experimental-design diagnostic. The teachers designed AI to prompt students, support their thinking, and reveal useful evidence about learning back to the teacher. This included student misconceptions, ongoing explanations, readiness patterns, and lab performance. The AI acted as a learning guide rather than an answer machine. However, only two proposals made teacher control explicit, and only two included safeguards. This shows that teachers naturally use AI to reveal student thinking, but need help defining protections and authority.

Why it matters

Science teachers need AI tools that respect their professional judgment. If tools are built without clear rules for teacher control and safeguards, they might confuse students or fail to address real learning gaps.

Questions answered