Instructors and institutions are trying to set policies and standards on the uses of AI in the classroom. We have seen and heard about issues of academic integrity related to the use and adoption of these tools. Instructors are grappling with the challenges of designing valid and fair assessments. Here are some strategies that you may consider for your own practice.
- Make Assessments More Realistic
You should be realistic with your assessment. The ‘realistic’ part does not refer to your ability to be reasonable or fair. Rather, it refers to the understanding that students will use whatever tool is available to their advantage. Knowing this, design for the use of AI in your assessments. Let’s face it, in their life outside of school there is a fairly good chance that students will use AI to solve problems, generate ideas, and crunch numbers.
- Make Assessments More About the Process Not the Product
Assessments and assignments should emphasize learning over the final product (Hodges and Kirschner, 2024). Having students complete assignments in stages helps to measure and monitor learning and growth. Consider having students submit outlines, drafts, or annotated bibliographies as part of their final submission. Having students demonstrate or highlight their pathway to completion is critical. AI can assist with this, but there may be glaring issues of inconsistency and misalignment.
- Make Assessments Preformative
In the right environment you can have students demonstrate their understanding or knowledge of material and concepts. This might not work for a class of a few hundred students, but in smaller settings presentations or practical skills demonstrations may be a way to remove reliance on AI by students.
- Make Assessments More Course Specific
To help eliminate the use or impact of generative AI tools on assessments, Hodges and Kirschner (2024) suggest tailoring assessments to be ‘more specific, personalized or context-dependent’ (p. 198). This means including prompts directly related to in-class discussions, activities, and uniquely crafted scenarios. These areas of focus are less likely to be picked up with AI as there would be an absence of context.
- Have smaller and more frequent assessments
Having more frequent lower stakes assessments helps to monitor student learning and progress in your course. This may be an issue in online courses. You may consider some icebreaker or anticipatory-type questions at the onset of an online module and unit, and accompany this with a follow-up assessment to see how learning has been modified as a result of the progression through material.
- Make Assessments Personal With Reflection
Generative AI tools are not good with personal tone or expressing emotion (Rudolph, Tan, and Tan, 2023). Reflections produced by an AI tool are more obvious to detect as they don’t often point to specific elements of learning, and are in a more ‘stilted’ tone. Further, you may have a basis for comparison with other examples of writing and vocabulary that students have used in other course contributions.
- Include More Critical Thinking and Analysis in Assessments
Incorporating some form of critical thinking and personal reflection in assignment prompts helps to eliminate the reliance on AI by students. Metacognitive prompts in assignments are a good practice already (Valeyeva and Valeyeva, 2020), and it hard for students to rely on a tool to comment about their own learning processes, critical thinking, and experiences.
- Use Mixed Approaches in Assessments
Look at the types of assessments that are more susceptible to the use of AI. For example, essays and discussion board posts. Mixing assessment approaches helps to ensure validity in responses. An example of this comes from an article many years ago (See Tamir, 1990). This article highlights the use of multiple choice questions with justification. The intention is to give credit for not just the correct answer, but also the justification for why the learner selected a particular answer. The goal of redesigning assessments in the age of AI is to give consideration for the process of learning, not just the outcomes of learning. In a mixed assessment approach students must demonstrate reflection, critical thinking, and rationalization. Not just ‘getting the right answer.’
- Use Peer Review and Collaboration
As a continuation of the above suggestion, you may consider something like a collaborative assessment. Several years ago Gilley and Clarkston (2014) examined the use of collaborative assessments like a multiple choice exam. In this process students took the exam individually, then without receiving feedback took the exam with another student or a small number of students collectively. The follow-up to the initial attempt was ripe with discussion and elaboration among pairs or teams of students. The outcome of this approach was better individualized learning on subsequent assessments. Perhaps using this approach for a lower stakes assessment prior to a more significant assessment might help with stemming the reliance on AI.
- Embrace AI and include it in your assessment design
It is assumed that you are in the business of teaching as part of your role. Policing and playing games is likely not part of your instructional responsibilities. So, why avoid something that you cannot control or police all of the time? Some instructors have made the decision to embrace AI and use it in their assessment design. Smolansky, A., Cram, A., et al. (2023) offer the suggestion of using an adapted essay prompt to incorporate AI into the assessment. Example: “You are given a 5-page essay produced by ChatGPT on [a given topic in your discipline; e.g., Greek mythology, human rights, sustainable energy, sorting algorithms]. You have 7 days to analyze the essay and edit it yourself to improve its quality, making clear references to the original text where applicable.” (p. 378)