We can’t outrun generative AI 

Any assessment strategy that relies on outpacing current AI limitations will soon become a solution that is out of date. 

There is no “one size fits all” assessment type that will sidestep the challenges faced by genAI. Instead, we recommend deploying a variety of assessment types to evaluate student understanding. 

Click on a strategy below to jump to more details. 

1. Closed book, in-person exams 

In-person exams that exclude the use of electronic devices will not need to change as students cannot use AI tools to take them. However, students might use generative AI as a study partner to prepare for an in-person exam in the same way they might study with a classmate.

2. Live presentations that include robust Q & A

The delivery of a live presentation is less impacted by genAI. Live presentations require students to have planned and prepared a talk to communicate their key message. In preparation for live presentations, students may use generative tools to brainstorm, summarize resources, or generate outlines. However, the presentation and Q&A itself require students to respond in the moment, drawing on their preparation and prior knowledge. When students are speaking live about course content and taking questions from other students or faculty, they are practicing the kinds of conversations an industry expert may have in the field and by doing so, learning along the way.

Some students may find this form of assessment particularly challenging (e.g., students with disabilities, non-native speakers, those who struggle with public speaking). Consider diversifying the types of assessments you use. 

4. Frequent and varied formative assessments 

Formative assessments are a great way to identify where students are struggling in their learning so you can make adjustments in your instruction.  The best are frequent, low-stakes tasks or assignments, graded for completion or ungraded, so students have little incentive to rely on AI. 

Some examples of formative assessment include: 

• Practice problems (partially solved, or with full solutions provided)

• Short quizzes (with immediate feedback)

• Small group problem solving live in class

• Exit tickets / weekly memos

• Required discussion forum posts before or after class

3. Create multi-part assignments 

Use a pedagogical approach called “scaffolded assessments” to break down assignments into smaller parts that build upon each other, leading to the completion of a larger final product. This approach lowers the stakes of any one deliverable and makes success on the final product more likely.  If you ask students to compose all components in the same shared document, you will be able to review the composition and revision process.  

Larger assignments divided into stages can take many forms.

If you permit or encourage use of generative AI tools at some stages of the process, e.g., brainstorming, copy-editing, and planning, you can add a reflection component in which students reflect on the contribution of genAI to their learning process. This reflection may involve considering why they chose to use genAI at certain points, assessing the positive or negative impact it had on their work and learning. Encouraging students to critically analyze their use of genAI fosters metacognitive skills and helps them develop a deeper understanding of the tool's benefits and limitations. 

1. Heavily weighted high-stakes exams (comprising more than a quarter of the course grade)

While these types of exams and assignments can be convenient to employ, they are the most susceptible to plagiarism and other kinds of academic dishonesty. High-stakes assessments can create significant stress and anxiety for students and shift their focus from fostering a deeper understanding of the subject matter to the grade they will receive. 

We recommend 

1. employing diverse assessment types across the course, 

2. providing students with multiple opportunities for feedback, and 

3. incentivizing the knowledge construction process. 

A balanced assessment approach provides a more comprehensive view of students' progress and encourages a deeper and more meaningful understanding of the subject matter.

2. Exams that focus on recall of basic facts

Including too many simple knowledge recall questions in an online or take-home exam will make your assessment more susceptible to AI generated responses.  

Instructors should move away from including recall questions in exams and instead incorporate them throughout the course as frequent, low-pressure, formative assessments that allow students to check their understanding. In general, employing a range of assessment types during your course will encourage critical thinking, discourage cheating, and allow students to demonstrate more of what they know.

3. Simple, generic essays

ChatGPT and other genAI tools can produce seemingly convincing answers to common essay questions, however we do not advise eliminating essays and final papers from your assessment toolkit. Instead, consider the following strategies for improving essay assignments.

• Use a rubric to clearly define the learning goals and criteria for evaluation.  

• Scaffold the assignment so that students can complete smaller tasks like topic approval, outline review, draft review, final submission, etc.

• Incorporate both traditional essay writing and genAI tools as complementary components of the writing process.

• Require a minimum number of citations of relevant course topics and readings.

• Offer students some autonomy in selecting their essay topics or final paper themes as it aligns with the overall learning goals. Allowing them to choose subjects that align with their personal interests, experiences, or career goals can increase engagement and ownership of the assignment.

• Use a problem-based approach: Frame the essay as a response to a real-world problem or challenge, encouraging students to think critically and develop practical solutions, rather than just summarize concepts, theories, or facts.