Designing Assignments That Keep Students Thinking

In my previous post, I argued that faculty don't need to redesign their entire course because of AI. Instead, start with one assignment. Identify the assignment where you're least confident that the final product reflects actual learning and ask a simple question: "What evidence of learning am I actually looking for?" That question naturally leads to another: What kinds of assignments become more valuable when AI is readily available?

I don't think AI-proof assignments exist. Students can use AI with almost any task. They can use it to brainstorm ideas, summarize information, generate outlines, suggest solutions, draft responses, explain concepts, and provide feedback.

The technology will continue to improve. Which is why I don't believe the goal should be designing assignments that prevent AI use. The goal should be designing assignments that keep students thinking.

That's an important distinction.

Much of the conversation around AI in higher education still focuses on what students can produce. Can they write the paper? Can they answer the question? Can they complete the discussion post? Can they generate the presentation? But increasingly, AI can assist with all of those tasks.

If our assignments are primarily measuring information production, then AI will inevitably be able to participate in that process. The more interesting question is: What parts of learning remain uniquely valuable when information generation becomes easier?

When I think about the skills that matter most in both education and the workplace, I keep returning to things like:

• judgment

• evaluation

• decision-making

• problem-solving

• adaptation

• reflection

These are not skills centered on producing information. They are skills centered on making sense of information. And that distinction matters. AI can generate ten possible solutions to a problem. Students can evaluate which solution is most appropriate. AI can provide multiple perspectives on an issue. Students can determine which arguments are most convincing and why. AI can summarize a collection of sources. Students can decide what information is most relevant in a particular context.

In other words, AI can participate in generating possibilities. Students can engage in evaluating those possibilities. That shift reminds me of ideas found throughout the literature on authentic assessment and expertise development.

Experts are not simply individuals who possess more information. They are individuals who can recognize patterns, evaluate alternatives, make decisions under uncertainty, and apply knowledge in unfamiliar situations. Those are exactly the kinds of capacities that become more important when information itself becomes abundant.

This is one reason I find myself designing more assignments that ask students to:

• explain their choices

• defend recommendations

• evaluate competing alternatives

• apply concepts to unfamiliar situations

• troubleshoot problems

• identify weaknesses in proposed solutions

• justify decisions using evidence

Notice what these activities have in common. The focus is not on generating an answer. The focus is on explaining thinking. And that changes the nature of the assignment.

For example, consider a traditional question that asks students to identify the best solution to a problem. AI can often provide a reasonable answer. But what if we instead ask students: Why is this solution preferable to the alternatives?

Now we're examining judgment.

Or imagine asking students to compare two competing AI-generated recommendations and explain which one they would implement. The assignment shifts from information production to evaluation.

Similarly, instead of asking students to summarize a concept, we might ask them to apply that concept to a new scenario, identify limitations, or explain how changing conditions would alter their decision.

The goal isn't to make assignments harder. The goal is to make student thinking more visible.

This distinction is important because I sometimes hear faculty describe these approaches as attempts to "outsmart AI." That's not how I see it. I'm not particularly interested in winning a competition against technology. And given the pace of development, that seems like a losing strategy anyway.

What interests me is preserving the parts of learning that matter most. The reality is that professionals already use tools to support their work. Engineers use software. Writers use editing tools. Researchers use databases. Accountants use calculators and spreadsheets. Learning has never been about refusing to use tools. It's about developing the judgment required to use those tools effectively.

AI should not change that goal. If anything, it may reinforce it. The challenge for educators is not determining whether students used AI. The challenge is determining whether students engaged in the thinking we hoped the assignment would develop.

That is why I believe one of the most productive questions faculty can ask when reviewing an assignment is: "Where does the thinking happen?"

If the answer is unclear, there may be opportunities for improvement. If the answer is obvious, then the assignment is likely doing important work regardless of whether AI is present. Because ultimately, the value of an assignment has never been the artifact itself. The value lies in the thinking that the assignment encourages.

And that leads to what I increasingly believe is the most important design principle in an AI-rich classroom:

Making student thinking visible.

Series 1: AI Is Exposing Existing Problems ✓ Completed

Series 2: What We Do About It

Current Post (2 of 8): Designing Assignments That Keep Students Thinking

Next Up: Making Student Thinking Visible