Using ChatGPT in Teaching

1. Planning a Course

An LLM can help draft a course plan. It can also be given an existing syllabus and asked for possible improvements.

2. Explaining Difficult Concepts in Class

For difficult concepts, one useful approach is to ask for a comparative explanation. Another is to request a maximally detailed, step-by-step explanation.

For example, rather than giving only a vague explanation that a proton-pump inhibitor binds to and inhibits the proton pump, the explanation can follow the entire process: the drug is taken, delivered to parietal cells through the bloodstream, secreted, and activated in an acidic environment.

Analogies and concrete examples can also help. The difference between competitive and noncompetitive antagonists can be compared to dance partners, for instance. Warfarin and aspirin can be used as specific examples.

3. Strengthening Lecture Content

Lecture content can be strengthened by uploading lecture-note PDFs and asking what could be improved. It is also useful to ask what students may be likely to wonder about.

Adding short pieces of interesting background knowledge that may catch students’ attention can help as well. When I include this kind of material throughout a lecture, there are clearly fewer students falling asleep.

4. Creating Questions

A short true-or-false quiz at the end of class can be a useful way to conclude. Questions can be displayed on a screen and answered through mobile devices using online tools such as Kahoot or Mentimeter, or they can simply be asked verbally. Practising retrieval has substantial educational value.

Multiple-choice questions are much harder to write than short-answer questions. When using an LLM to create multiple-choice questions, the following practices are helpful:

• Ask for the correct answer to be placed in option 1.
• Ask for one or two extra distractors, then remove unsuitable ones.
• Do not allow “all of the above” or “none of the above.”
• Request a brief explanation of the answer.

5. Organizing Course Feedback

As LLMs develop, it has become easier to use open-ended responses from course evaluations, and their importance will likely increase. I tend to ask students to write substantial open-ended comments when completing teaching evaluations.

There is also an increasing expectation that instructors provide a post-course improvement plan. An LLM can help generate ideas for what to include in that plan.

6. Preliminary Scoring of Short-Answer Responses

This may become useful as examinations increasingly move to CBT or UBT formats. Responses can be preliminarily scored with an LLM, sorted according to those scores, and then graded by the instructor. This may support more consistent grading.

It is also possible to provide a grading rubric and ask the LLM to perform preliminary scoring according to that rubric.

7. Assignments Using an LLM Chatbot

An assignment that simply asks students to write a report on a topic now has less educational value. A better format may be to provide a prompt and ask students to critique the result.

For an important concept, students can ask for two or three analogies or examples, then write which one is most appropriate, why it is appropriate, and what they think about it.

It can also be useful to add a report section titled, “One point that sparked my curiosity and what I freely investigated about it.” In an actual assignment, items 1, 2, and 3 could be generated by an LLM, whereas item 4 required students to act independently. This may be feasible because the assignment was graded pass/fail.

Ideally, an LLM-integrated assignment would go as far as having students present their work, as in a flipped classroom. However, this is realistically very difficult in classes such as ours, with around 100 students.