Use ChatGPT to Generate Question Banks in Canvas

This teaching tip demonstrates a way in which you can use ChatGPT to produce a large number of quiz questions that you can then use in your Canvas course. Because the method here uses a very specific sequence of steps using a variety of tools from different locations, it is highly likely that certain steps in this sequence will change, however, the general method of getting AI-generated content into your course can benefit you and your students. Creating banks of questions is laborious work and is one of the tasks that ChatGPT is well suited for.

Here’s the general sequence of steps that I used in creating a one-hundred-question multiple-choice quiz based on what might appear in an introductory university-level Astronomy course.

  1. Use ChatGPT to generate a mix of multiple-choice and true/false questions, and direct the AI to produce the results in the AIKEN format.
  2. Visit the New York Institute of Technology’s free web-based tool https://www.nyit.edu/its/canvas_exam_converter to convert your text file from AIKEN to the Question and Test Interoperability (QTI) format. The QTI format can be imported into Canvas.
  3. Visit your Canvas course home page. Under “Course Status” click on the box labeled “Import Existing Content”.
  4. Select “QTI .zip” as the content type using the drop-down menu.
  5. Select “create new question bank” from the “Default Question bank” dropdown. 
  6. From the Quiz course menu, find your newly created quiz with your questions and add additional details like due dates, availability, etc., and publish it so your students have access.

In my exercise to create this 100-question quiz, most of the work involved wrangling ChatGPT to produce the correct format of the quiz questions. I archived my ChatGPT prompts for generating the quiz at: http://ecampus.uaf.edu/go/astronomy-quiz.

Observations from working with ChatGPT to produce the Astronomy quiz:

  • ChatGPT is not an expert in Astronomy. I noticed one question that was conceptually incorrect. Asking ChatGPT to reexamine its questions sometimes gets it to notice errors. Sometimes I had to tell it a question was unanswerable, or that it was wrong. This fixed the problem, but remember that I could have always edited the question manually.
  • There is an output limit to ChatGPT responses in terms of a maximum number of characters per response. I had it generate one hundred questions by asking for questions numbered 1-10, 11-20, 21-30, etc.
  • There is a copy button next to each of ChatGPTs formatted outputs. I used these buttons to copy and paste the content into a text file saved on my local machine.
  • When I noticed that ChatGPT produced something that I thought needed a revision I asked it in different wording, pointing out what needed to be changed and in what manner.
  • I made several revisions and attempts to get a format that worked with the Text to QTI conversion tool that I was using. Perhaps another tool would have worked better. Perhaps such a tool will appear on a UAF website in the near future.
  • Part of my conversation with ChatGPT involved directing the type of content as well as the format of the content.

Considering the Pedagogy beyond these Procedures

Students need practice and providing them with practice quizzes is a good example of formative assessment. You might consider sharing with them the manner in which you create your own practice quizzes, and encourage them to create their own.

It is important to note, that due to the limitations of ChatGPT, not all of the content will be factually accurate; some might not make sense. It’s your job as an instructor to sift through the content and ensure that it’s ready for your students. Or is it?

It might very well be that your students are at a sufficient level of mastery where they can identify what is accurate, and what is in a state of error. Being able to explain why something is wrong is often better than merely being able to select a correct answer.

CTL AI Workshop

Interested in working with AI in your course? Join CTL for a 2-part workshop on Working with Artificial Intelligence in Higher Education on August 17 & 18 from 10 AM to 12 PM, location to be decided. There will also be an online option for those who would like to virtually attend.

Dan Lasota

Dan LaSota

Instructional Designer
Certified QM Peer Reviewer
Certified QM Training Facilitator

dlasota@alaska.edu

Leave a Reply

Your email address will not be published. Required fields are marked *