Designing AI-Responsive Assessments and Assignments
Small, actionable strategies
Teaching Innovation and Program Strategy | TIPS Nov 14, 2024
It is not possible to create learning assignments and assessments that are fully AI-Proof. Nor are there proven, compelling technological interventions that currently exist that ‘detect’ the use of AI with a high enough level of certainty that it is responsible to use them as policy for your courses.
We do have a few smaller, actionable strategies for designing assignments and assessments that are more responsive to the reality of teaching and learning in an age of AI. We should keep in mind, though, that AI is always changing, and we are always learning more. Perhaps the transitory nature of this technology can also create a positive outcome for our pedagogy, pushing us to continue to iterate, reflect, and evolve.
We want to affirm that this situation is frustrating, and so many efforts to maneuver this tricky situation feel unsatisfying. When students use AI in a course, it often seems like it is a replacement for thinking, learning, and the struggle that is often necessary for growth. However, if we start from a place of intention and care in our teaching, we will always be in a better position to address negative things when they inevitably arise.
And, very often, when our courses center student learning, engagement, and authentic learning experiences, generative AI will also necessarily become less and less of a factor.
Suggestions for Designing AI-Responsive Assessments and Assignments
- Add time limits for multiple choice quizzes (allowing flexibility for students who need more time, of course); scramble questions and answers; have several correct answers for a question. AI programs are pretty sophisticated, so this will not erase all usage of AI; these are just a few small things to try within the quiz itself.
- Use a "wrapper" in which students can reflect on their scores for the assessment: What did students find tough or confusing? What do they feel confident about? What do they want to learn more about?
- Ask students to identify where they found their answers (cite/take a picture/screenshot from a reading, notes, a lecture, or experience from class as appropriate.)
- Increase the number of quizzes to lower the stakes. Ungraded quizzes are useful for students to monitor their own progress.
- Explore contract grading in which there’s an understanding that if students complete a certain number of assignments (for which they are given lots of feedback) they will automatically receive a certain grade. They can put more effort into those assignments to receive a higher grade. The hope is that this arrangement creates less anxiety about individual assignment grades, and therefore disincentivizes cheating (and, perhaps, discourages the use of AI to replace the processes of thinking and learning.)
- Require a meeting to talk through the answers before a grade is given in situations where an instructor may suspect that AI was used
- More broadly, instructors are setting policies that outline acceptable and non-acceptable uses of AI in their courses. These policies talk about potential academic integrity issues, but can also touch on labor, privacy, intellectual property, and climate issues relating to the use of AI.
TIPS has a great AI resources page that has some more ideas and readings. CLAS also has some guidance on syllabus statements relating to AI.
Here at TIPS, we are happy to chat more with you about developing an AI-responsive (and student-centered!) approach to teaching and learning. You can schedule some time to meet with an instructional designer for more pedagogy-focused questions (including thinking through alternate assignments) or an instructional technologist for help incorporating technology into your course, and/or configuring your quizzes/assignments in Canvas.