Generative AI in Teaching and Learning: Assessment

One of the best approaches to mitigate concerns that students will use generative AI tools in a way that goes against your policies or expectations of academic honesty can be found in how you develop your own assignments and how they are then assessed.

Clarify your expectations about student use of generative AI tools (from basic to more complex uses)

• Define the universities relevant policies in the context of your class
• Define what “completion” of work entails
• Define what citation of generative AI work involves
• Require a documentation the process of use/engagement
• Indicate your consideration of using ‘checker’ tools early and often (well before submission of work

Consider employing multimodal submissions with students ‘closing the loop’

• Ask students to submit work in Canvas and include a short audio/video clip that clarifies the overall scope of the submission, their process of developing their ideas, etc.
• Include this component of submission in a rubric/evaluation standard

>> Faculty Spotlight: As a recipient of a PAIR student-as-partners grant (Partnership in AI Research), Dr. John Griel (UT School of Law) is redesigning elements of his Law and Religion Clinic to take advantage of generative AI tools in crafting and revising emails to clients, creating and improving contract terms, preparing questions for an oral argument, and increasing productivity in "Bluebooking."

In addition, consider the following suggestions for mitigating concerns of the use of generative AI tools (as outlined by the AI support initiative at the University of Oslo):

• Link the task to concrete work in teaching. For example, you can work on one or more cases in the lesson, and questions can be linked to this specific work. If necessary, work with cases in teaching can be included in the assessment basis (for example portfolio assessment).
• Link to experiences. For example, let the students explain and reflect on process execution, learning activities, problem solving, etc.
• Assignment formulations linked to the syllabus. Design complex task formulations that are specifically linked to the curriculum contribution and that require critical reflection. For example, students may be asked to use a specific syllabus to illustrate an actual/practical problem. Current issues are particularly suitable, as ChatGPT will have little or no content on this.
• Elicit the student's own views. Give tasks that involve analyzing complex issues, evaluating alternative solutions and arguing for their own positions.
• Get the students to create something new. Give tasks without pre-given answers and encourage originality and creativity. For example, students can be tasked with developing and justifying new research questions or justifying arguments, with references to specific literature.
• Visualize thought processes. Include a reflection note in which the students describe how they have worked on the assignment and what they have learned in the process.
• Current issues. Give the students (current) cases/scenarios which they will then use specific syllabus contributions to illustrate. 
• Project work. This work may have submissions/presentations during the work.
• Oral. Forms of oral assessment, such as video reflection, presentation, conversation.
• Iterative feedback. Forms of assessment where students receive feedback along the way and improve their work.
• Assess and/or compare different texts using the syllabus (for example fellow students' work, fictitious student answers, published articles or text produced by ChatGPT).
• Ask students to explain the workflow. If there is reason to believe that students will use ChatGPT to answer the assignment, ask them to document how they proceeded (what they asked, how they assessed and processed what the bot generated, what they added themselves and what they learned in the process)

References