Module 4: AI and Your Learning¶
Module 4 of 10
What You Will Be Able to Do¶
- Tell help that builds your knowledge from help that does the thinking for you.
- Have a chatbot quiz you from your own course materials, and check its answer key.
- Say what the early studies of AI and learning show, and why they are early.
The Core Idea¶
An artificial intelligence (AI) chatbot will explain almost anything, at any hour, as patiently as you need. That makes it a strong study partner, and also an easy way to skip the part of studying that does the work. Reading a clear explanation feels like learning; being able to give the explanation yourself, later and without help, is the learning. One question sorts most uses: will this leave me able to do it without the tool?
The same tool, two kinds of help
Help that builds
- it quizzes you from your notes
- you answer first, then it explains
- it points at the step you got wrong
You can do it later without the tool.
Help that does the thinking
- it answers before you try
- it summarizes what you did not read
- it writes what you were meant to work out
It feels like progress; the skill stays with the tool.
Help that builds your knowledge leaves the recalling and the reasoning to you: it quizzes you, asks you to explain, waits for your answer, and then shows you where you went wrong. Help that does the thinking takes those steps for you: it answers before you try, summarizes the reading you skipped, or writes the explanation you were meant to work out. The second kind has its uses (a summary can orient you before you read), but it does not build the skill you will be examined on, or, in medicine, the skill you will one day use with patients.
Have It Quiz You¶
Testing yourself is one of the best-supported ways to learn. In a classic experiment, students who practiced recalling a passage remembered more of it a week later than students who reread it, although rereading did better on a test five minutes later and left students feeling more confident (Roediger and Karpicke, 2006). A review of ten study techniques rated practice testing and spreading your study over time as highly useful, and rereading and highlighting as of low use (Dunlosky et al., 2013).
For medical education, a review recommends frequent, spaced testing that makes you produce the answer, with feedback (Larsen, Butler and Roediger, 2008).
A chatbot makes that kind of testing easy, as long as it works from your course. Give it your lecture slides or notes, and ask it to quiz you one question at a time, to wait for your answer, and then to tell you what you missed. Ask for questions you answer in your own words as well as multiple choice, and come back to the same material on later days. Your course materials come first because your exams are written from them.
Its questions and its answer key need the same checking as anything else it writes. When its answer disagrees with your slides, the slides are your reference: look the point up in your textbook, and ask your instructor if the slides themselves look wrong. The prompt library has tutors built for this. The Lecture Tutor teaches one objective at a time and checks you as it goes, the Daily Review Sheet Builder ends each sheet with questions to test yourself, and the worked example shows a full session with a question tutor, including what to check before you trust it.
What the Early Studies Show¶
What the studies showed as of October 2026:
- High-school math in Turkey. Nearly 1,000 students practiced over four 90-minute sessions, some with a chatbot that worked like ChatGPT, some with a version set up as a tutor that gave hints rather than answers, and some with only their notes and textbook. Students with the chatbot scored 48% higher on the practice problems than those without it, but 17% lower on the closed-book test at the end of each session; most had asked it for answers and copied them. Students with the tutor version scored 127% higher in practice, and their test scores did not differ significantly from those of students who had practiced without it (Bastani et al., 2025).
- University physics at Harvard. In a randomized crossover trial with 194 students, an AI tutor built from the instructors' own step-by-step solutions produced larger learning gains than an active-learning class session, in less time, on a test taken right after each lesson (Kestin et al., 2025).
- Medical students. Two small randomized trials, with 41 students in Finland and 33 in the United States, found no significant difference in what students remembered a month and a week later, and both were too small to rule one out (Gustafsson et al., 2026; Kalam et al., 2025).
- Doctors in practice. At four endoscopy centers in Poland, doctors who regularly used AI that flags polyps during colonoscopy found adenomas, a kind of polyp that can turn into cancer, in 22.4% of the colonoscopies they did without it, down from 28.4% before the AI arrived: a drop of 6 percentage points, which the study observed but cannot show the AI caused (Budzyń et al., 2025).
Medical educators describe three risks (Abdulnour, Gin and Boscardin, 2025): deskilling, losing a skill you had; never-skilling, never building a skill in the first place; and mis-skilling, learning a tool's errors as fact. A 2026 perspective in Nature Medicine says direct evidence of never-skilling in medical training is still absent, and treats it as a risk to plan for (Ke et al., 2026).
The studies are early: there are few of them, most are small, they follow students for minutes to a month, and each tests one course with one tool, while the tools change faster than studies are published. So far the pattern fits decades of research on learning: help that does the work can raise scores while you have it and leave you worse off without it, while in these studies help built to make you think did not do that harm.
Where the Policy Comes In¶
The AI Responsible Use Policy asks that AI not become your primary source of foundational knowledge: the learning materials your school provides and the established literature stay that source (Responsible Use, section B). Module 3 gave the reason in your own interest: those are what an AI explanation can be checked against, and what your exams are written from. This module adds the other half of the reason: a tool that does your thinking can leave you less able to do it when it counts.
For Faculty: Sessions and Assignments¶
The same question applies to the work you set. An assignment graded only on its finished product invites help that does the thinking; one that makes the reasoning visible keeps it the student's, for example a first attempt submitted with the AI-assisted revision, a short oral explanation, a note on what the tool got wrong, or a quiz in class before the AI-assisted homework. Showing students, in a session, how to have a tutor quiz them from your slides gives them a good use to copy, which a rule about what not to do cannot. Module 7: Teaching and Assessment covers drafting materials and assessment items with AI, and Your Syllabus AI Statement helps you set expectations for a course.
Self-Check¶
Two classmates use a chatbot on the same practice questions. One reads its explanations; the other has it quiz them from the notes and answers first. Who is better placed on exam day?
The second. Answering first is retrieval practice, which builds memory that lasts; reading explanations feels productive but leaves the recalling for the exam. The first classmate can turn the reading into practice by trying the same questions again later, without help.
A chatbot's answer key disagrees with your lecture slides. What do you do?
Treat the slides as your reference and the disagreement as something to resolve before you rely on either answer: check your textbook, and ask your instructor if the slides themselves look wrong. Models write fluent, confident errors, and your exams are written from your course.
How can a case write-up let students use AI and keep the reasoning theirs?
Make the reasoning the graded part. Ask for the student's own list of possible diagnoses, with a reason for each, first; then let them use AI to challenge it, and have them hand in both, with a note on what changed and why. Or follow the write-up with a short oral discussion. Either way the student's reasoning, and what the AI added to it, are there for you to see.
Going Deeper¶
Color key: Guide
- For Students Habits for the basic sciences, where the AI works from your materials and you do the recalling.
- Worked Example: Practice Questions from a Lecture A tutoring session from start to finish, and what to check before you trust it.
- Standing Setups A project for a course, so a tutor knows your objectives all term.
Content last reviewed October 15, 2026. Review dates are maintained as described on the About page.