What Counts as Evidence of Learning?

AI is forcing a question that higher education has been able to avoid for a long time: What actually counts as evidence that learning happened? Because for years, many assignments have focused heavily on the final product:

• The paper.

• The discussion post.

• The answer.

• The presentation.

But now AI can help generate many of those products quickly and convincingly. And that changes the conversation. Not because learning no longer matters. But because polished output is no longer reliable evidence of understanding on its own. A student can now produce something that looks academically complete… without fully engaging in the thinking process that was supposed to build the learning.

So, the shift I keep coming back to is this: From: “Did the student produce the answer?” To: “What evidence shows me the student actually understood the work?”

That doesn’t mean abandoning projects, papers, or writing. But it does mean we may need to look for different kinds of evidence alongside the final product.

Things like:

• Justification and decision-making

• Explaining why choices were made

• Reflection on the process

• Application to new or unpredictable situations

• Troubleshooting and revision

• Connections between concepts

• Demonstration of reasoning

In my own technical courses, I see this clearly during labs and troubleshooting activities. A student can sometimes arrive at the “right answer” quickly, but the deeper evidence of learning often appears in the process itself: how they diagnose a problem, how they explain their reasoning, how they respond when something doesn’t work, and how they adapt when conditions change.

That’s much harder to outsource. And honestly, those are often the skills that matter most outside the classroom anyway. AI is pushing us to reconsider something important:
Maybe learning was never best measured by the final artifact alone.

Maybe the real evidence has always been in the thinking behind it. And if that’s true, then the future of assessment may be less about proving students worked independently… and more about making learning visible.