On this page
Before you teach a topic, it helps to know which wrong ideas your students will bring. An AI chatbot can draft that list quickly. Then check each item against research and your students' answers, and plan around the ones that hold up.
Why do misconceptions matter for lesson planning?
One large study comes from US middle school physical science. Sadler and colleagues (2013) studied 9,556 students of 181 physical science teachers. On test items with a very popular wrong answer, teachers who could identify that wrong answer had larger classroom gains than teachers who knew only the correct answer. On items where students showed no misconception, the teacher's subject knowledge alone accounted for higher gains.
Can AI predict which misconceptions students will have?
Partly. It depends on the model and the subject.
In a paper accepted at ACL 2025, Lee, Kim and Jo trained a model to judge which of two wrong options students would find more plausible, in computer science questions. Its ranking accuracy was comparable to that of human experts. A general chatbot was not built for this.
A 2026 study by Zengaffinen and colleagues, accepted to Findings of EMNLP 2026, studied how large language models write wrong options for multiple-choice questions. On a maths dataset, the models tended to reason from likely student errors. On a science dataset, they leaned more on answers that were merely similar to the right one.
The same study found that putting the correct solution in the prompt gave a relative improvement of 6.4% in how closely the model's wrong options matched those written by humans.
Two practical tips follow from this research. Tell the AI the correct idea, and be more sceptical of its list for science topics than for maths.
How do you ask AI for a useful misconception list?
Here is a prompt you can adapt. The worked example is Newton's first law in Class 9 physics.
“I teach Class 9 physics in India. The topic is force and Newton's first law. The correct idea is: an object stays at rest or keeps moving at constant velocity unless a net force acts on it. List six wrong ideas students commonly hold about this. For each one, give the wrong idea in a student's own words, the everyday reasoning that makes it feel right, and one multiple-choice question where a student holding it would pick a specific wrong option. Say how sure you are that each is common. Do not invent sources.”
How do you check the AI's list against research?
Look for each item in a research-based source before you plan around it.
For mechanics, the Force Concept Inventory by Hestenes, Wells and Swackhamer (1992) is a multiple-choice test written in everyday language with common-sense wrong options. According to PhysPort's page on the test, the current version has 30 questions and suits high school and introductory college students. We explain it in our guide to the Force Concept Inventory and misconceptions about force.
PhysPort lists 123 research-based assessments for physics teachers.
For other subjects, search for “concept inventory” plus your topic, and read the abstract of anything you find.
Then sort the AI's list into three groups:
Keep: a research source describes the same wrong idea.
Test: it sounds plausible, but you found no source.
Drop: it contradicts the research, or the AI's “correct idea” is itself wrong.
AI drafts contain errors, and our guide to checking AI lesson-plan drafts covers the other places they hide.
How do you check it against your own students?
Your class may differ from the research, so run a short diagnostic before the unit.
Pick three or four items from the Keep and Test groups.
Use the AI's multiple-choice questions, but check each one yourself: one correct answer, and each wrong option tied to one misconception.
Ask students to add one line under their answer: “Why did you choose this?”
Count the answers. If many students pick a wrong option, plan a lesson moment around that idea. If almost no one does, drop it for this class.
Note any new wrong ideas in the “why” lines for next year.
Using evidence of what students think to change your teaching is formative assessment, the idea behind Black and Wiliam's 'Inside the Black Box'.
What should students do after the diagnostic?
Students whose answers show a wrong idea need to meet the right one again. The Fermi Concept Library has a lesson for every concept in physics, chemistry, biology, mathematics and computing, so a student can revisit the exact idea their answer flagged. For practice at home, Fermi's AI tutor notices where a student is stuck and decides whether to explain, ask or show.
Rerun the diagnostic at the end of the unit to see which wrong ideas changed.
