Assignments in the Context of A.I.

The proliferation of sophisticated AI tools presents both challenges and opportunities for education. It’s no longer a question of whether students will use AI, but how they will use it, and how educators can adapt. This necessitates a thoughtful re-evaluation of assignment design and assessment methods. Students need to be guided not only on the ethical and effective use of AI but also on understanding its limitations and when to rely on their own cognitive abilities.

For guidance on how to communicate with students about which AI uses are accepted, see our Course Policy Guidance page.

Table of Contents
Assignment Design Considerations
AI-Resistant Assignments
Modelling effective engagement with AI
Assessment in the context of AI
Going Further with AI
Access and Equity

Assignment Design Considerations

The first step in assessing what approach to take to AI for a particular assignment is to get clear about the fundamental purpose of the assignment.

Begin by defining your learning outcomes. What is the skill/ability you want them to learn? Be explicit. Is it critical analysis, problem-solving, creative synthesis, persuasive writing, data interpretation, coding proficiency, or something else? The more clearly defined the skill, the easier it is to design an assignment that targets it directly.

Identify: What do students need to do for themselves in order to develop that skill/ability? Which parts of the process are non-negotiable for student development? For those practices, consider what you can do to ensure that students complete those tasks. Part of this will be communication with students; you can also consider ways to design AI-Resistant Assignments.

Then consider: Where could AI assist or extend the work without undermining the core learning outcome? AI can be a powerful assistant for brainstorming, initial literature searches, grammar/style checks, data visualization, or even generating counterarguments for students to then refute. Where could AI extend the work to allow for more complex or ambitious tasks? AI could help analyze larger datasets than manually feasible, generate code snippets that students then integrate and modify, or create initial drafts of sections that students then critically revise and expand upon.

Where these points are effectively communicated, students have a good idea about which uses of AI will get in the way of, and which develop, their learning.

This flowchart from the University of Michigan is an excellent decision-making tool for educators. It prompts a crucial series of questions:

• Does the assignment assess essential learning? If not, the assignment itself might need rethinking, regardless of AI.
• If yes, can AI pass the assignment? If AI cannot currently replicate the desired skills, no immediate change might be needed (though this can change rapidly).
• If AI can pass, would its use undermine learning and your ability to assess it?
• Yes: This is where strategies like monitoring AI use closely, redesigning the assignment to be more AI-resistant, or shifting the assessment focus become critical.
  No (or it could enhance it): This opens opportunities to modify the assessment to incorporate AI thoughtfully or integrate AI use as a learning outcome itself.

AI-Resistant Assignments

If you have considered your learning outcomes and identified particular skills that students need to practice themselves in order to learn, the first thing to do is to consider how to communicate to your students the importance of practicing a particular skill. Doing this can make it more likely that students become willing collaborators in their own education.

Beyond that, you can also design assignments in a way that reduces the temptation to use AI, makes it less helpful, or shapes the way in which it is used. For example, you can:

• Focus assignments on personal reflection and lived experiences. This can be faked by an AI, but students are more likely to see the interest in the topic and be motivated to do it themselves.
• Put emphasis on metacognitive tasks like personal reflections on the learning process, challenges faced, and ethical considerations.
• Use oral/practical activities like debates, presentations, Socratic seminars, group problem-solving on novel scenarios, and practical skill demonstrations performed live.
• Where a particular stage in the work is crucial to learning, you can have students complete that work during class time. This also makes an opportunity for you to provide support with tricky activities.
• Shift some of the assessment weight from the final product to the process of creating it (e.g., documented research logs, annotated bibliographies, drafts with revisions).

It is possible to design assignments that blend AI use with independent work, but ensure that students are performing the key tasks themselves. If the outcome is “students will be able to synthesize information,” the assignment might ask students to do initial research with the assistance of an AI, and then write their own synthesis in-class or with requirements for documentation of and reflection on the process.

Modelling effective engagement with AI

An effective method for promoting good uses of AI can be to model effective usage in class. In these activities:

1. Thoughtful reflection can be made central.
2. You can work on co-creating effective prompts, and make these available to students.
3. You can supervise, ensure that students are critically engaging, and likely highlight some cautionary tales.
4. Students work through the kind of process you want to promote and see the value in it, making them likelier to work in this way in independent work.
5. AI use is brought into an above-board, accepted space in which students feel more comfortable asking questions about accepted use.

Once students have worked through this kind of process with support, they are better positioned to make judgements and embody good habits in the future.

Assessment in the context of AI

Assessment strategies need to adapt to ensure academic integrity and accurately measure student learning in an AI-influenced context. In some cases, you can assess understanding in settings where an AI cannot do the work for the student. In others, it is appropriate to be assessing the effectiveness of the AI use itself, taking into account that AI use will be present in many professional lives.

Understanding checks

An important method here is to incorporate methods that assess how well students understand the material they submit, whether or not they use AI at an earlier stage.

• Oral Defense: A brief conversation can quickly reveal whether a student truly understands the concepts, or simply submitted a polished AI output.
• In-Class Discussion and Activities: Design activities where students must apply the knowledge or skills from an assignment in a new context, without AI assistance.
• Quizzes on Submitted Work: A short quiz focusing on the key concepts and methodologies of a recently submitted assignment can also gauge individual understanding.

Critical engagement with AI

Where you are requiring the use of AI in an assignment, you might assess the student’s effective and critical engagement with AI output, their reflection on the process, and the documented steps.

• Prompt Engineering: Teach students the skill of writing effective prompts to elicit desired outputs from AI (see Introduction to Prompting). Require students to document an iterative process of prompt refinement, and submit the final successful prompt and the AI-generated result, along with a process log.
• Output Evaluation: Don’t just look at the final product. Ask students to critique the AI’s output. What were its strengths and weaknesses? Were there biases in the AI-generated content? How did they verify the information provided by AI?
• Reflection on Process: Require students to document how they used AI. What tools did they use? What prompts did they employ? What iterations did they go through? What did they learn about effective prompting or the AI’s capabilities? This metacognitive reflection is a valuable learning experience in itself.

Make clear attribution a requirement. For further guidance on this, visit our Course Policy Guidance page.

Going Further with AI

Instead of solely focusing on preventing AI misuse, educators can design assignments that intentionally leverage AI to achieve learning outcomes more effectively or to teach students how to use these tools productively and ethically.

AI can represent a powerful tool for student support and tutoring. For more on this, see Applications for Student Learning.

In addition, it is possible to design assignments that will help students to develop the skills they will need to work effectively with AI in their education and professional lives.

Example assignments that leverage AI capabilities:

• Role-play a Conversation: Students ask the AI to adopt a detailed persona, perhaps of a patient or client, and role-play a conversation with them.
• Prototype Three Ideas for a Project: Students use AI to quickly sketch out or model different solutions to a problem, allowing for rapid iteration and comparison before committing to one path. Students carefully evaluate the options and document the process.
• Analyze a Dataset: AI tools can perform complex data analysis, identify patterns, or create visualizations from large datasets. The student’s role then shifts to interpreting these results, drawing conclusions, and assessing the limitations of the AI’s analysis.
• AI-Assisted Content Creation and Curation: Students use AI to generate initial drafts of content (e.g., blog posts, marketing copy, code segments) which they then must critically edit, fact-check, refine, and add their own unique insights to, documenting the process.
• Structured Prompting for Specialist Tasks: Students are tasked with building a tool that will generate reports with a predetermined structure. Taking the example of the Sift Toolbox created by Mike Caulfield, which instructs an LLM to behave as a sophisticated contextualization tool and write structured reports on its findings, students consider the components they want to see in their reports, and evaluate the results.
• Building Simulations: Students ‘vibe code’ simple interactive games or simulations that embody a problem or challenge (more on what vibe coding is). For an example, in this video Mitchell B. Weiss of Harvard describes how he used ChatGPT to build an html game that challenges students to solve a real-world traffic problem; Ethan Mollick at the Wharton School used AI to build a simulator that explains the central limit theorem. Each of these is within the abilities of students to execute themselves with the coding help of an AI.

Some inspiring and provocative ideas can be found in this curated collection of AI-engaged assignments from Harvard’s AI Pedagogy Project.

By thoughtfully redesigning assignments and assessments, educators can harness AI’s potential to deepen learning, foster critical thinking, and prepare students for a future where collaborating with intelligent systems will be the norm. The focus shifts from preventing AI use to guiding its effective and ethical application.

Access and Equity

In any cases where you are designing assignments that allow or encourage the use of AI, it is important to be mindful of accessibility.

• Will all students have the same access to the necessary tools, without paying a fee?
• Will some students gain an advantage because of economic factors or training?
• Will students feel coerced to use AI, because they will not do as well without it?

Think about how you will present the assignment, and provide guidance to students, such that there is a level playing field.

Some students will have legitimate concerns about using AI (see General Considerations for Using AI in Teaching and Learning for a brief discussion of some of the areas of debate around AI). If students have not had the opportunity to ‘opt in’ (for example, by signing up for a class that signaled that it would involve AI use), it is important to offer ways for students to participate without using AI themselves. This can be achieved by offering multiple options for assignments, or asking students to work in groups or pairs.

For more guidance on how to use these tools, visit:

General Considerations for Using AI in Teaching and Learning
Course Policy Guidance
Introduction to Prompting
Assignments in the Context of AI
Labor-Saving Tips for Educators