Are Your Assessments Still Measuring Student Learning in the Age of AI?

For years, colleges and universities have relied on papers, projects, presentations, exams, discussions, and other student products as evidence of learning. Generative AI now complicates that relationship. A completed product may still be excellent, but the product alone may reveal less about how a student reasoned, made decisions, evaluated evidence, revised an idea, or transferred knowledge to a new situation.

That is why the most useful question is not simply, “Did a student use AI?” The more important institutional question is:

This shift matters because higher education is already moving in this direction. EDUCAUSE reported in 2026 that artificial intelligence is reshaping assessment practices across higher education, and its current teaching and learning work increasingly emphasizes authentic, process based demonstrations of learning. Recent EDUCAUSE commentary also argues that evidence of learning becomes more visible when students explain reasoning, revise understanding, justify decisions, and apply knowledge in new contexts.

A polished product is not the same thing as visible learning

A strong final submission can demonstrate achievement, but it may not show the full process that produced it. In an AI enabled environment, that distinction becomes more important.

Consider two students who submit equally polished policy analyses. One student independently evaluated competing evidence, made difficult tradeoffs, used AI appropriately for support, and revised the recommendation after identifying a weakness. Another student supplied a broad prompt to an AI system and made minimal changes to the result.

If the assessment captures only the final paper, those very different learning processes can look nearly identical.

Six questions institutions should ask about assessment

Higher Education Partners uses six dimensions when examining an assessment portfolio. These dimensions are designed to help academic leaders and faculty identify where evidence is strong, where it may be obscured, and where redesign could add educational value.

Why AI detection is not enough

Detection answers a different question. Even if a tool could determine with certainty that AI contributed to a submission, that would not tell an institution whether the assessment itself provides useful evidence of learning.

A stronger assessment strategy begins with validity. What learning matters? What evidence would demonstrate it? What parts of the student’s thinking should be observable? Where can AI support learning without replacing the very cognitive work the assessment is intended to measure?

That approach also creates more durable assessment design. AI tools will continue to change. An assessment built around meaningful decisions, contextual application, explanation, reflection, and defensible judgment is less dependent on the capabilities of any single technology.

The portfolio matters as much as the individual assignment

One assignment rarely needs to demonstrate every dimension of student learning. The larger question is whether the assessment portfolio provides a sufficiently rich body of evidence across a course or program.

A program may have strong content alignment but weak process visibility. Another may ask students to perform sophisticated analysis while relying heavily on generic written products that can now be generated with little student reasoning. A third may have several authentic assessments but no clear progression from foundational knowledge to independent professional judgment.

Looking across the portfolio makes these patterns easier to see.

What stronger evidence can look like

Assessment redesign does not have to mean adding more work for faculty or creating elaborate new assignments. In many cases, relatively small changes can make learning more visible.

• Ask students to identify and justify a consequential decision.
• Require them to explain why one alternative was rejected.
• Capture a meaningful checkpoint, draft, decision trail, or revision.
• Introduce a constraint or changed condition and ask students to adapt.
• Ask what evidence would cause them to change their conclusion.
• Require reflection on assumptions, errors, uncertainty, or changed thinking.
• Use multiple forms of evidence when one final artifact cannot support the inference being made.

These strategies shift attention from the surface characteristics of a submission toward the reasoning and judgment that the institution actually values.

Where should an institution start?

A practical first step is not a university wide redesign effort. It is a structured review of a representative assessment portfolio.

1. Select representative assessments. Focus on major assignments that carry meaningful weight in a course or program.
2. Map them to intended learning. Clarify what each assessment is supposed to demonstrate.
3. Examine the evidence. Review learning evidence, cognitive demand, AI resilience, process visibility, contextual judgment, and alignment.
4. Identify portfolio patterns. Look for repeated strengths, gaps, redundancies, and areas where evidence has become difficult to interpret.
5. Prioritize redesign. Keep what is already working, enhance what needs modest improvement, and focus deeper redesign where the educational value is highest.
6. Implement and reassess. Assessment redesign should become an improvement cycle rather than a one time response to AI.

The institutional question has changed

Generative AI has accelerated a conversation higher education needed to have anyway. Institutions can respond by trying to determine whether a particular tool contributed to a particular assignment, or they can ask a more durable question about the quality of the evidence on which academic decisions depend.

For academic leaders, faculty, assessment professionals, and teaching and learning teams, the goal is not to make assessment technology proof. It is to make learning visible enough that faculty can still make credible judgments about what students know, can do, and are becoming capable of doing.