Why the greatest risk of AI in higher education is the erosion of learning.

Source: Fast Company

Public debate focuses on cheating. The authors argue the deeper risk is structural: as AI performs more knowledge work, universities risk hollowing out the very ecosystem that produces expertise.

The surface narrative is about student misconduct.

The real shift is institutional automation.

AI is now embedded in:

• Admissions screening

• Academic advising

• Scheduling and risk scoring

• Tutoring and feedback

• Research synthesis

• Experimental design

So the real concern is no longer whether students use AI. It’s whether universities quietly redesign themselves around AI efficiency.

Because the real shift here is about cognitive offloading at scale.

Universities face pressure from all sides:

• Budget constraints

• Competition for enrollment

• Demands for measurable output

• Political scrutiny

• Administrative cost creep

AI promises efficiency, scalability, and productivity.

If the metric is output (degrees, publications, throughput), automation looks rational.

If the metric is formation (judgment, struggle, mentorship), automation becomes destabilizing.

The incentive tilts toward efficiency. Formation becomes invisible.

This isn’t really about cheating. It’s about friction.

Learning is slow.

Struggle is inefficient.

Mentorship is labor-intensive.

AI removes friction.

But friction is where competence forms. When:

• Drafting is automated

• Feedback is machine-generated

• Research cycles are partially agentic

• Tutoring is synthetic

The university risks producing credentials without deeply formed expertise.

The erosion isn’t sudden. It’s gradual.

The pipeline thins quietly.

Universities now face a fork. They either adopt:

Model 1

University as an output machine.

Optimize for productivity.

Adopt AI wherever it increases measurable performance.

or:

Model 2

University as an ecosystem.

Preserve mentorship.

Protect productive struggle.

Design AI use around formation, not replacement.

Because we’re long past the “Can AI write essays?” question.

The real issue we’re facing now is:

“What disappears when humans no longer practice the work that forms them?”

Because automation in and of itself isn’t the risk. It’s outsourcing the process that builds judgment.

And judgment doesn’t scale as easily as software.