AI didn't break integrity, it exposed credentialism

Last year a browser extension made college students a simple promise: it would read the quiz in front of them in their course and answer it with one click. It arrived in the middle of the alarm over chatbots. Commentators declared the essay dead, faculty wondered whether any unsupervised assessment could still be trusted, and academic integrity became the main way higher education talked about AI.

Phil Hill, founder of Phil Hill and Associates and publisher of the On EdTech newsletter, thinks that lens points at the wrong problem. He agrees some of the cheating is real, and he also sees a moral panic. In his reading, AI has not created a crisis of integrity. It has exposed a system that built everything around earning the credential and left learning optional. If he is right, better detection will not fix it. Institutions will have to decide what they are actually assessing.

Why shortcuts were always the rational move

Hill's argument starts with incentives. If the credential is the goal, and success means moving students toward it, then any shortcut to the credential looks attractive. "We shouldn't be surprised that that's happening," he said. Cheating is the most visible shortcut. The deeper issue is that many courses were never built to show whether a student had learned anything, only whether they had finished.

He does not treat credentialism as new, or as a problem that belongs only to higher education. At the Blackboard users conference in July, Stanford's David Labaree spoke about where credentialism came from and why it is a core problem. Labaree did not connect it to AI, but Hill said hearing him sharpened his own thinking. Hill traces the pattern to an American habit: name one ideal path, then treat everything else as the alternative.

His example comes from his years as an Air Force electrical engineer at the end of the Cold War. He wanted to keep doing hands-on engineering and research. The career structure pushed officers toward acquisition, which meant managing large contractors and moving through a fixed sequence: Wright-Patterson Air Force Base, then Washington. By his account, the Soviet military had career tracks for people who wanted to stay operational.

He sees the same binary in American education. It runs from Bill Bennett's early-1980s work defining an ideal K-12 path into college, through what Hill describes as the Obama administration's push for a college degree for all. Even then, he noted, some people were asking how to support everyone headed straight into the workforce.

States are already redefining what counts

Darren, host of the Hard Truths in Higher Ed series on Signals in Higher Ed, pointed to a correction that has drawn little attention. Almost every state has set a goal for the share of adults holding a postsecondary credential, and many of those goals now count far more than degrees. In his home state of Michigan, the 60 by 30 goal counts a skill certificate the same as a college degree.

Hill sees this as one of the system's real strengths. States act as laboratories, and he pointed to the Texas coordinating board's work on alternative credentials, pathways and ways to measure and support them. The weak spot is what happens next. Good ideas tend to stay inside the state that produced them, and Hill does not see comparable effort at the federal level or in work that brings states together.

That matters for the AI debate. A broader definition of credentials does not, on its own, cure credentialism. A certificate can be chased as a box to check just as easily as a degree. Wider recognition only helps if what gets recognized reflects demonstrated learning, and that puts the question right back on assessment.

Removing the measure is not the same as fixing it

The University of Michigan offered a test case. Darren noted the announcement that incoming freshmen would get pass or fail marks instead of letter grades in the fall semester, a policy framed around mental health. He saw a possible correction, one that could turn students toward belonging and learning, and something worth studying if more schools tried it in the first year.

Hill was more skeptical. "I think they're trying to do a shortcut to deal with something as opposed to a deep rethinking," he said, while adding that he truly hopes he is wrong. His concern is a familiar pattern: a measure with well-known flaws gets removed instead of improved, and the gap creates bigger problems.

He pointed to the move away from standardized testing across much of higher education over the past four or five years. In California, he said, faculty are overwhelmed and say they are being asked to teach middle school to incoming students. For Michigan's change to be useful, in his view, it has to be replaced with something that actually encourages learning.

This is where Hill parts ways with a simple anti-grading stance. He is not defending grades as the thing that counts. He is saying institutions have leaned on weak signals of learning, and pulling the signal without building a better one does not bring learning back into focus.

Three models that put learning back at the center

Asked what shows promise, Hill named three approaches. None is new, and he called none of them a panacea. Together they shift the emphasis from completion to demonstration.

1. Skills and competency-based education. Hill said it has not reshaped higher education as much as he had hoped, but where it exists, at Western Governors University and elsewhere, it has shown real benefits. The key, he said, is finding a way to show that a student has the skill "we think you have." The open problem is embedding it in mainstream education.
2. Authentic assessment. Even in courses organized the traditional way, faculty can evaluate what students know with better methods than heavy reliance on multiple choice and other easy, flawed formats. Hill was clear that multiple choice still has a place.
3. Institution-wide AI frameworks, with the University of Sydney as his lead example. Its white paper sets out principles for transforming education in the age of AI that, in his words, go "well beyond the trivial human in the loop type of what I think are banalities."

Hill said he cannot yet point to results from Sydney, but he called its framework one of the best he has seen. It accepts that students need preparation for a world full of AI, and it deals head-on with the downsides instead of treating AI as purely good or purely a threat.

As Hill describes it, Sydney's assessment framework runs two streams. Open assessments assume students can or will use AI, and are designed so the work still produces learning. A separate stream excludes AI, which in practice means proctored, in-person testing and similar formats.

The two-stream design is the most practical answer to the problem Hill laid out at the start. It does not try to keep AI out of every assessment, which the quiz-answering extension suggests is a losing fight. It also does not give up on verifying what a student can do alone. Each assessment has to say up front which kind it is and what it is meant to show.

The risk of swapping one narrow metric for another

Darren raised a trend he sees at many institutions: more work-based learning, closer employer partnerships, and stronger pressure to show that a degree leads to a job. Hill welcomed parts of it. Paying attention to what graduates do after college is good, and employer input reinforces the move toward skills and competencies over the letter A or B on a transcript.

He added two caveats. First, the field should not swing from treating everything as a traditional credential to treating everything as work-based learning. Not every program should be judged by the job a student lands right after finishing. Second, learning has value of its own: learning with others, forming a worldview, becoming an educated citizen.

Darren agreed, calling for a recommitment to the liberal arts. The link back to AI is direct. Trading grade-chasing for placement-chasing would keep the credential problem alive in a different outfit.

The bottleneck is scale, not ideas

Hill returned to Texas as the closest thing to a statewide reckoning. He said it treats the problem as multidimensional and uses multiple metrics, recognizing that some students need immediate jobs and others go on to graduate school, with particular focus on community colleges and regional universities. He sees plenty of good work elsewhere too, but says it sits too often at the margins.

His diagnosis: higher education is not short of ideas. Competency-based programs, authentic assessment and frameworks like Sydney's already exist. What is missing, he said, is the willingness to apply lessons already learned to the mainstream fast enough to meet what he called an existential challenge.

Darren pushed further. He tells both academics and edtech companies that pilots are "the death of your business." Half measures signal that the change does not really matter, and they rarely get enough resources to produce results. In a risk-averse culture, he said, leaders sometimes have to take big swings and even put their position at risk.

For provosts, assessment leads and faculty developers, the takeaway from Hill's argument is not about catching AI misuse. It is about going back to ground zero: what each course is trying to evaluate, and whether its assessments actually show it. Hill admits higher education is good at fooling itself and may find ways to keep doing so. For now, he credits AI with forcing a conversation he says "we should have been having ten years ago."