Why use cognitive science to check an AI lesson
The four principles are attention, active engagement, error feedback and consolidation. Using them to check an AI lesson has one advantage: it swaps the vague question of 'is this tool any good' for four questions you can answer one at a time.
What follows goes through each in turn, showing where AI helps and where it does harm. The same tool, used well, hits the principle; used badly, it gets in the way of learning instead, and the difference often lies only in the lesson design.
Attention: where AI helps and where it harms
Attention is the way into learning: learning happens wherever the attention goes. AI can make material pitched at a student's level, cutting the distraction that comes from work being too hard or too easy, and an instant response helps keep them engaged. The other way round, a flashy interface, too much animation and too many alert sounds pull attention away from the content. A busier tool is not necessarily a more effective one.
The check: does this AI feature carry the student's attention towards the content to be learned, or towards the tool itself?
Active engagement: giving the answer kills engagement
Learning depends on the student actively recalling and thinking; passively receiving works far less well. This is where AI is most easily misused: the student asks, the AI hands over a full answer, the thinking is skipped over entirely, and the work looks done while nothing has actually been learned.
A better use is to have the AI offer a hint rather than the answer, to require the student to try first before the AI responds, or to have students judge whether the AI's output is right or wrong. Judging the AI's answer is itself a form of active engagement. When you design the lesson, hold one line: leave the thinking part to the student.
Error feedback: where AI does its best work
Timely, specific, individual feedback on errors is one of the most effective factors for improving learning, and it is also the hardest thing to do in a traditional classroom, since one teacher can hardly mark for thirty students at once in real time. This is where AI is most valuable: it can point out, there and then, what went wrong and why, so the student can fix it while the memory is still warm.
But the quality of the feedback depends on whether the tool is aligned with the local curriculum and marking criteria. Vague praise or a wrong correction is worse than no feedback at all. Before you adopt it, check whether its feedback is grounded and whether it can point to a specific step.
Consolidation: working against forgetting
What has been learned gets forgotten, and consolidation relies on spaced review and varied practice. AI can generate varied questions close to a student's real mistakes based on their weak spots and schedule the review at the right time, work that takes the most time to do by hand. The misuse is to turn consolidation into a mass of mechanical drilling, where students do a lot but think little, and it wears down their motivation instead.
The check: is this practice helping students make their understanding firm, or just piling up quantity?
A design question you can apply
Condense the four principles into a single question a teacher can use on the spot: does this AI use help students focus better, take a more active part, get better feedback and remember more firmly, or does it just save the teacher some effort? If all four hold up, it is a use worth going ahead with; if it only saves effort with no gain to learning, then however convenient, it should be held back. Judging by learning rather than by convenience is the most solid line to draw as AI comes into the classroom.