AI-Aware Assessment Strategies

If you would like to complete the corresponding activity with this webpage, please download the word document worksheet resource

This guide is designed for instructors who are anywhere in the process of considering the impact of Generative AI (GenAI) on their assessments. Whether you are looking to prevent student misuse of GenAI in your assessments, or are considering integrating GenAI directly into your assessments, this resource will walk you through how to develop assessments which are “AI-aware”.

What is an AI-aware Assessment?

We define being AI-aware as recognizing the complexity and potential impact of GenAI and responding appropriately and in alignment with your own values. What this means for assessment practices is that rather than starting from a point of an outright ban or complete integration of GenAI, your assessment redesign considers what you value as an educator, the assessment goals, and how a student’s use of GenAI in completing the assessment could support or impede those goals.

In supporting instructors in designing AI-aware assessments, the goal of this resource is not to advocate for any specific level of integration, but rather to give instructors a chance to reflect on how GenAI may require new considerations in your assessments, and to consider how you may want to respond. Importantly, this starts from the recognition that GenAI is impacting our classrooms, and in many cases, has revealed vulnerabilities in existing assessment practices. These vulnerabilities require careful reflection to ensure that any assessment redesign efforts are deliberate, purposeful, and focused on supporting learning rather than focusing on restricting or policing GenAI use. Many strategies intended to detect or prevent GenAI use, such as AI detectors, have been shown to be inaccurate. As a result, an AI-aware approach to assessment design can help ensure that assessments remain aligned with learning outcomes while reducing the likelihood of AI misuse.

The following process provides a framework for reviewing your existing assessment and deciding whether, and how, assessment redesign may be warranted.

Step 1: Think of an Assessment

Before beginning, think of an assessment that you think needs to be redesigned in response to GenAI. This assessment may be particularly vulnerable to tasks that GenAI performs well, such as creating polished final drafts or answering questions based on publicly available information, as discussed in more detail below. This can be a signature assessment in your discipline, or one that is unique to your course(s). Based on this assignment, please consider the questions below and if you would like, respond in the chart in the worksheet:

• The assessment’s purpose. What is the assessment trying to achieve and which learning outcomes are students demonstrating through this assessment? For example: The purpose of the research essay is to assess students’ ability to critically evaluate scholarly literature, construct evidence-based arguments, and communicate disciplinary knowledge in writing.
• The assessment’s deliverables. What product(s) are the students submitting? For example:
• The evidence of learning. When students submit the deliverable, what do they need demonstrate to you? For example: students must demonstrate a coherent argument, appropriate use of scholarly sources, critical analysis of evidence, accurate application of course concepts, and adherence to disciplinary writing conventions.

Step 2: Reflect on your Teaching Values in Assessment  

Values are our fundamental beliefs about what is important. Our values motivate action and impact who we are and how we exist in the world. In our assessments, our values determine how and what we assess, and how we assess communicates to students what we value in teaching. Examples might include things like curiosity, teamwork, critical thinking, communication, or reflexivity Take a moment to consider:

• What values does your discipline hold for assessment? (For example: creativity, communication, career readiness)
• What values do you hold for assessment in your course? (For example: flexibility, critical thinking, collaboration)

We start with these considerations, with the goal of recognizing that, while GenAI might impact how and what we assess, our core values should continue to guide assessment practices. This helps ensure that efforts to address potential GenAI misuse do not inadvertently compromise those values.

Step 3: Consider the Affordances and Limitations of GenAI in Assessment

GenAI impacts assessment in a number of ways as a result of the various affordances which such models carry.  

Some of the affordances of GenAI include its ability to automate mundane tasks, generate answers based on large amounts of data, simulate human conversation, and rapidly replicating patterns to create polished final drafts (see for example this scientific report in Nature). Some of these functionalities may be beneficial to your students in supporting their learning. For example, students may use GenAI to further develop their understanding of a subject, or they may offload some formatting tasks to spend more time engaged with the content. It may also help make content more accessible to students. Such strategic use of GenAI tools may benefit student learning and prepare them for use of these tools outside of their courses. Importantly though, with these affordances come opportunities for misuse. For example, the ability of GenAI tools to analyze large amounts of information and create polished final drafts may encourage cognitive offloading, where students may engage less deeply in the thinking and revision processes involved in producing their work. Additionally, because GenAI models are trained on existing data, their outputs may reflect limitations or biases present in those data sources. If incorporated uncritically, these biases can be reproduced or reinforced. As a result, assessment design may benefit from emphasizing process visibility or critical engagement with GenAI produced material may be built into assessments to be aware of the impacts of AI on these assessments

GenAI also has limitations though. Due to GenAI’s predictive nature, it is not capable of true understanding of material, and may not be able to reason through complex topics. Importantly, GenAI tools are limited to making predictions based on the data they have been trained on, which now includes most publicly available information on the internet, as well as specific data which individual developers have accessed to feed to their models such as licensed material like books and news articles. In this way, while GenAI can produce final drafts, showing process-oriented steps and consistency between ideas and assignments can be more limited. Additionally, due to limitations in GenAI training data, engagement with real world, or course-specific examples and their nuances produce more superficial results. This is because GenAI tools are not able to draw on the specifics of these examples in their responses, limiting their accuracy and ability to reflect the case. Lastly, due to the fact that GenAI cannot truly engage with the learning process, nor is it capable of understanding the students’ own learning journey, metacognitive evaluations of students’ engagement with material are not able to be replaced by GenAI tools. This means that assignments which ask students to engage deeply with the reasons behind their decisions, as well as their own growth from term to term is not a task which is easily replaced by GenAI

With this context in mind, it is important to consider in your assessments where and how student use of GenAI may be valuable, and where you may want to redesign to prevent misuse. Keeping your values at the front of this redesign is important to ensure your assessments remain pedagogically valuable and useful.

Step 4: Consider the Do’s, Don’ts, and Don’t Knows of Assessment Redesign

With the assessment and values that you considered above in mind, consider the following questions from a revised version of the Do’s, Don’ts, and Don’t Knows adapted from the University of Cape Town and Stellenbosch University.

Do clarify the assessment’s purpose and constraints: 

• What do I need evidence of?
• What knowledge and skills must students demonstrate independently?
• Where might AI support learning without undermining outcomes?
• What tradeoffs am I making in my assessment design? (e.g., validity over feasibility)

Do consider assessment authenticity and process-focused design: 

• How can I assess process-based thinking? 
• How can my assessment be broken into steps or components?
• Where can I develop students’ AI literacy and competencies?
• How does a student’s access to tools impact their ability to complete the assessment?

Do consider how communication and collaboration can impact your assessment: 

• Who can you collaborate with in testing out assessment designs before rolling out to your course (e.g., peers or students)?
• How will you communicate the purpose and conditions for requiring or banning AI use?

Based on these questions, how might you adapt your assessment? As a few examples, consider some options below:

• A pre-submission proposal step to show process visibility
• Requiring specific references to course material to show engagement with content/skills which AI may not have access
• Building in a metacognitive element for students to justify decisions made in the assessment
• Incorporating nuanced real-world examples without a single correct answer to engage higher order thinking and application in relevant cases

If you would like a consult about your assessment or to discuss the possibility of a retreat for your department, feel free to reach out to us at ctl@uwo.ca.

Additional Resources

Adendorff, H., Cilliers, F., Huang, C.-W., Lester, S., Strydom, S., & Walji, S. (2026). Do’s, don’ts and don’t knows: Responding to AI in assessment in universities: A practical guide for lecturers. Stellenbosch University & University of Cape Town. https://files.su.ac.za/public/centre-learning-technologies/documents/2026-02/dos-donts-and-dont-knows-responding-ai-assessment-universities-practical-guide-lecturers_0.pdf

Boles, S. (2025, July 16). Does AI understand? Harvard Gazette. https://news.harvard.edu/gazette/story/2025/07/does-ai-understand/

Bousfield, D. (2025). Assessment and AI. Centre for Teaching and Learning, Western University. https://teaching.uwo.ca/genai/posts/2025/article_five.html

Califano, G., & Vecchio, R. (2026). Is generative AI WEIRD? Considerations on the use of silicon samples in sensory and consumer research. Food Quality and Preference, 145, Article 106024. https://doi.org/10.1016/j.foodqual.2026.106024

Chapman University AI Hub. (n.d.). Bias in AI. Chapman University. https://www.chapman.edu/ai/bias-in-ai.aspx

Pazur, B. (2025, May 16). Generative AI: Everything to know about the tech behind chatbots like ChatGPT. CNET republished in Yahoo Tech. https://tech.yahoo.com/ai/articles/generative-ai-know-tech-behind-160003895.html

Recker, M. (2024, December 15). Benefits of GenAI in business operations. InterVision Systems. https://intervision.com/blog-benefits-of-genai-in-business-operations/

Riemer, K., & Peter, S. (n.d.). AI doesn’t really ‘learn’ – and knowing why will help you use it more responsibly. Sydney Executive Plus. https://plus.sydney.edu.au/ai-doesnt-really-learn-and-knowing-why-will-help-you-use-it-more-responsibly

Suazo, I., Santiesteban Velázquez, M., Saracostti, M., & Chaple Gil, A. M. (2026). Between cognitive offloading and critical autonomy: A systematic review of the epistemic implications of generative AI in higher education. Frontiers in Education. https://doi.org/10.3389/feduc.2026.1878665

Tishcoff, R., Agoe, E., Isik, M., & MacFarlane, A. (2024, November 20). Using generative AI to make learning more accessible: Insights from Ontario PSE students and staff. Higher Education Quality Council of Ontario. https://heqco.ca/pub/using-generative-ai-to-make-learning-more-accessible-insights-from-ontario-pse-students-and-staff/

Wu, S., Liu, Y., Ruan, M. et al. Human-generative AI collaboration enhances task performance but undermines human’s intrinsic motivation. Nature Scientific Reports 15, 15105 (2025). https://doi.org/10.1038/s41598-025-98385-2