Assessment and Assignment Guidance in the GAI Era

CATL is often asked questions about how to approach assessments in the wake of easy access to generative artificial intelligence (GAI). We hope to crowd-source suggestions and examples from our own instructors so that we can build a repository of work from the UWGB community in Canvas. Please take our GAI Assignment Repository survey if you have any ideas or assignment samples that you are willing to contribute. We will collect results and share them once we have a critical mass. In the meantime, please feel free to use these resources from other institutions or professional organizations.

Deterring the Use of GAI

One of the most frequent concerns expressed by instructors is that students will simply use GAI to complete their assignments. It makes very good sense to type your own assignment prompt into Copilot or another GAI tool to see if it can complete the assessment – and modify the assignment if it can. Although these are not perfect strategies, multiple authors have suggested the following mitigation strategies:

  • Be very clear with students about your GAI policy and talk with them about the reasons for it and for completing their work independently. Carleton College has an interesting site aimed at a student audience about GAI use and evolving understandings of academic integrity that you could use to frame a discussion with your class.
  • Use shorter, more frequent assessments that build on each other to reduce the motivation to use GAI that can come with a small number of high-stakes/point assignments.
  • Scaffold your assignments such that each one builds on another.
  • Create assessments that involve integration of material from class (which GAI does not have access to) with other sources.
  • Require some element of reflection on the learning experience or completing the assignment.
  • Ask students to submit outlines, drafts, or save their document history in Word or Google Docs to show the evolution of a paper or project.
  • Consider the use of oral presentations, including question and answer, as a method for demonstrating understanding.
  • Employ in-class writing assignments such as Minute Papers for face-to-face courses.

The Office of Digital Learning at UN-Reno has created a more detailed document with strategies for re-designing assessments in the GAI era.

Teaching About the Use and Ethics of GAI

Some instructors are seeking strategies for teaching students about GAI and how to use it, as well as about related issues, such as the ethics of use. Instructors from the University of Central Florida collected more than 60 assignments related to GAI, including the teaching of prompt engineering. The New School has a shorter, more direct page of instructions for prompt creation. Co-Intelligence author Dr. Ethan Mollick and collaborator Dr. Lilach Mollick have created an extensive paper that outlines seven complex ways to use GAI in education, and they include sample prompts and a discussion of the potential risks of their ideas. Finally, as just one example of teaching about the ethics of AI use, consider this assignment designed, in part, to teach about cultural bias in GAI.

Creating Assessments that Make Use of GAI

Finally, there are instructors looking for creative assignments that use GAI as an intentional tool in an assessment to facilitate student learning. Numerous universities and organizations have assembled collections of such assignments, including the following:

Remember, we hope to create a collection of examples from UWGB instructors. Please complete our survey today to share your contribution with your colleagues!

Sample Assignments for Different Approaches to GAI Use

In a previous CATL article, we recommended using the traffic light model to guide students on the appropriate use of generative AI (GAI) in assignments and course activities. Assuming you’ve already included a policy on GAI in your syllabus, it’s also important to provide clear instructions in your assignment descriptions. Below are some examples of assignment descriptions, using the traffic light approach and graphic. Instructors will vary on whether they want to use that visual or simply explain in words. If you choose to use the stoplight visuals, please be sure to provide an accompanying description of what that means for your specific assignment. While tailored to specific subjects, these samples share common strategies.

Consider the following general suggestions when designing your assignments:

  • Be clear and specific about GAI use in your syllabi and assignments. Clearly outline when and how GAI can be used for assignments and activities. Avoid ambiguity so students know exactly what’s expected. For example, if brainstorming is allowed but not writing, specify that distinction.
  • Include GAI usage disclaimers in assignment directions. Regularly remind students by adding a GAI disclaimer at the beginning of assignment instructions. This will make them accustomed to looking for guidance on AI use before starting their work.
  • Explain the rational for AI use or nonuse. Help students understand the reasoning behind when GAI can or cannot be used. This can reinforce the learning objectives and clarify the purposes behind your guidelines.
  • Clarify the criteria for evaluating AI collaboration. Specify how assignments will be graded concerning AI use. If students need to acknowledge or cite their AI usage, provide specific instructions on how they should do so.
  • Define which AI tools students can use. Should students stick to Microsoft Copilot (available to them with their UWGB account, so they don’t have to provide personal information to a third party or pay a subscription fee) or can they use others like ChatGPT?
  • Use the TILT framework. Leading with transparent design for assignments and activities helps students clearly understand the purpose, tasks, and assessment criteria. This framework can also help instructors clarify how GAI should be used and assessed in assignments.

Sample Assignment Instructions on AI Use

Red Light Approach: No GAI Use Permitted Assignment Example

The example below is for a writing emphasis course and the assignment purpose is to evaluate students’ own writing. For this assignment, GAI tools are not allowed. The instructor includes an explanation of this description to further clarify the assignment’s purpose.

Yellow Light Approach: GAI Use Permitted for Specific Tasks/Tools Examples

The yellow-light approach can be hard to define depending on what you want students to practice and develop for a given assignment. We’ve provided two samples below that each take a slightly different approach, but all clearly label what tools and for what tasks AI can be used and why.

Green Light Approach: All GAI Use Permitted

Instructors may choose to take a green light approach to AI for all assignments or allow AI use for selected assignments. The example below takes a low-stakes approach, permitting full AI use to encourage experimentation. Even with this method, instructors should provide clear assignment expectations.

Learn More

Explore even more CATL resources related to AI in education.