AI in Assessment

While not all motivations are aligned with academic integrity and best practices, GenAI has many useful functions, and many more are likely to come that will help make tasks easier and more efficient. Some reasons why students might use AI include:

• Ability to gather information that is (rightly or wrongly) perceived to be unbiased and accurate
• Brainstorming and idea generation
• Efficient or easy completion of tasks
• “Cognitive offloading”
• Ability to engage conversationally or obtain feedback for tutoring or testing oneself
• Perceived ability to obtain higher grades, where focus is on grades, not learning
• It’s the new thing that everyone else is using

GenAI might be used to generate responses to almost any written assessment (including providing scripts, outlines, and background information for video, audio, or presentation-based assignments). Some strategies that can mitigate the advantages for students to offload these cognitive skills onto other tools include:

• Focusing on learning processes is a strategic way to embed both formative and summative assessment into assignments, and to ensure that assessment is not just about judging students, but also directing and fostering their learning.
  • Focusing more on the rationales and reflections behind choices and tasks, to encourage learning, even when AI is used for some tasks.

• Scaffolding, dividing large projects and assignments into smaller chunks allows opportunities for students to engage with more complex problems and questions, as well as for instructors to provide feedback to help students develop higher-level skills and complete higher-quality final products.
  • When breaking up assignments into smaller components, there will always be some that are easier to generate with AI than others. However, simply dividing a larger assignment into discrete steps can helps make it easier and more advantageous for the student to invest their own thinking and planning into each step, as it can otherwise be quite difficult to create coherence and connection between different tasks.

More general examples and tips can be found on CTL’s Assessment Design page and in Creating a Culture of Learning. Examples of process-based assignments, the specific challenges that might be encountered in a world of AI, and a few strategies for creating effective prompts to further support learning can be found in the Process-based Assessments in a World of AI document.

• Before creating any assessments, consider what is that core skill or set of skills that students must learn and demonstrate with this assessment. Make sure to review the learning outcomes for your course/program and assess if they are still being achieved by any new assessments. This reflective process may reveal that the learning outcomes may need to be revised.
• Developing or Revising Learning Outcomes (click the Assessment Planning button to view the strategies from UoT)

• CTL has many learning outcome resources and is available to provide support with revising course or program learning outcomes. Please contact ctl@uwindsor.ca.

• It is also important to consider program learning outcomes.

It is important to understand the capabilities of AI and be honest about the vulnerabilities your assessments might have. You can use Copilot to check how susceptible your assignment may be to AI. If you put in your assignment and ask Copilot to generate a submission, this may give you an idea of what it is capable of. However, you should be aware that different models have different capabilities, and the results you get will likely depend on the prompts you use. Notably, paid/premium models normally have significantly more capability than free models.

If you are concerned about students using AI for assignments, reference University of Windsor, Bylaw 54 & Bylaw 55 and Sample Course Syllabus Statements on the Use of Generative Artificial Intelligence. 

• Ultimately, whether you incorporate AI into your assessments will be up to you. A valuable approach is to consider a mix of assignment types, where some incorporate having your students use AI in an intentional way and some where students are encouraged to not use AI.
• The AI assessment scale (AIAS) may help you think of the different levels of AI use that can be incorporated into various assessments.
• As mentioned above, always consider your learning outcomes and whether they will still be met with the level of AI integration you are considering.
• You will need to clearly identify your AI policies in your syllabus.

• Effective Assessment Design 
• Designing Assessments That are Less Vulnerable to AI (from Melbourne)
• Assessment Design Strategies for AI (from Waterloo)
• Interactive Oral Assessments (from Melbourne)
• Using AI to Enhance Assessment (from Melbourne)
• Adapting Assessment Strategies (from York University)
• Designing Assessments in the Age of Generative AI (part of Generative Artificial Intelligence in Teaching and Learning at McMaster University press book. It includes a handy workbook.)
• What to do about assessments if we can’t out-design or out-run AI? (presents a very useful `two-lane’ approach)
• Proactive Strategies to Address Unauthorized GenAI Use (click the Academic Integrity button to view the strategies from UoT)

• Teaching Examples (Examples from U of T illustrating different approaches to generative AI, including those that intentionally integrate AI tools into their teaching practice, as well as those that prevent GenAI use. )
• Strategies for Common Assessment Types (specific considerations for different assessments from Waterloo)
• AI Assignment Library (from University of North Dakota)