Psychology & society

When AI-Assisted Cheating Casts Doubt on Honest Achievement

By Minsu Kim, Ph.D. · U.S.-Licensed Clinical Psychologist

What AI-assisted cheating can take away from people who followed the rules, and why that matters beyond the exam room.

Ahead of South Korea’s 2027 college entrance exam, the Ministry of Education announced stronger measures to prevent cheating. The possibility of unauthorized AI use brings an old question back into view. What exactly are we protecting when we protect the integrity of an exam?

People were finding ways to cheat long before generative AI arrived. Hidden notes and electronic devices offered different means to the same end: getting outside help on an exam meant to show what a person could do independently. AI can go further by interpreting a question and generating an answer rather than simply retrieving or relaying one. But the concern that stays with me is what happens afterward. When people no longer trust how a qualification was earned, even those who earned it honestly may find themselves having to prove it again.


Why the reaction goes beyond the rule itself

Cheating on a classroom exercise and cheating on an exam that determines university admission or eligibility for a professional license both violate rules. The consequences are different. A place at a university may go to one applicant instead of another. In professional licensing, there is a further question: can the public rely on the qualification when choosing whom to trust with their health, finances, or legal affairs?

It is easy to understand why people who spent years preparing would feel angry about an unfair advantage. But you do not have to be a candidate to have a stake in the outcome. Most of us cannot independently evaluate every professional we turn to for help. We rely, in part, on a process that has checked qualifications on our behalf. A breach can leave us wondering how much confidence to place in that process, even if the exam itself has nothing to do with our own careers.


What my psychology license represents

When I was working toward my first psychology license in Maryland, the licensing exam was one step in a much longer process. I completed graduate training, supervised clinical work, and an internship accredited by the American Psychological Association. While I was in my postdoctoral fellowship, I took the Examination for Professional Practice in Psychology (EPPP) as part of the Maryland licensing process, which also included a state jurisprudence exam. Each step gave someone a different opportunity to evaluate what I knew and how I worked.

The breadth of the EPPP content is worth noting. Someone might reasonably ask why a clinical psychologist needs to be tested on development, cognition, or research methods. In practice, two people who say “I can’t concentrate” may need very different things. One may have had attention difficulties since childhood; another may have started struggling after a period of poor sleep or sustained stress. Knowledge from several areas of psychology helps me decide what to ask next and what evidence would actually change my judgment. Supervision adds something a written exam cannot: someone can ask how I arrived at a conclusion, notice what I missed, and help me recognize when I need to consult or refer.

A license cannot promise that I will always get those decisions right. It tells the public that I met a set of requirements for entering the profession and can be held accountable to professional standards afterward. Its value comes from what those requirements allow others to verify, not simply from how difficult they were to complete.


What a compromised score can no longer tell us

Suppose someone passes a professional exam honestly but later struggles to apply that knowledge in practice. That raises a question about the assessment itself: did it adequately test abilities the job requires? Now suppose someone passes the same exam using unauthorized AI assistance. That raises a different question: whose performance does the score reflect? The candidate may already know the material. The score, however, can no longer establish what that person knew or could do independently under the required conditions. Improving the exam’s content addresses the first problem; safeguarding the conditions addresses the second.

This distinction makes it easier to see who else bears the cost. Imagine that you earned the same credential honestly and then heard someone ask, “How do we know you didn’t cheat, too?” You have not lost any of the knowledge or judgment you worked to develop. What you may have lost is the ability to point to the credential as evidence that others readily accept. Once that trust erodes, the work of proving what you earned can fall back on you, even though you did nothing wrong.


When justified anger turns into blanket suspicion

Public reactions can have more than one source. A person may sincerely object to an unfair advantage and also feel some satisfaction when a high-status group draws criticism. Psychologists use the term schadenfreude for pleasure at another person’s misfortune. Research on social comparison suggests that envy can contribute to this response in some circumstances. That research does not tell us why a particular person criticized an exam, and anger over unfairness needs no explanation in terms of envy.

The shift I pay attention to is the one from “Someone may have cheated on this exam” to “That qualification doesn’t mean much.” The first statement calls for evidence and a fair investigation. The second can quietly turn suspicion of a particular act into a judgment about people who were never implicated. It is possible to take misconduct seriously without treating everybody who holds the credential as suspect.


What prevention should protect

Prevention is hard to defend in advance. Its costs are visible: staff time, revised procedures, and sometimes inconvenience for candidates. When a measure works, what we observe may be a day on which nothing happened, and we cannot easily tell whether the measure made the difference. After an incident, the costs of doing nothing are easier to see. That may help explain why institutions sometimes act more decisively after a problem becomes public, even when the risk was foreseeable.

A practical place to start is to ask what candidates must be able to do without assistance, then design the assessment around that answer. If independent judgment matters, an exam should give candidates a way to demonstrate it under conditions the institution can reasonably verify. Where responsible use of tools is part of the work, that can be assessed separately, with the same tools and clear rules for everyone. Institutions can also test safeguards before rolling them out, tell candidates what to expect, and give them a fair way to challenge a mistaken finding. More surveillance, by itself, answers none of the questions about what an exam actually measures.

I think of the people who spent years learning their profession and took the exam expecting their work to be judged by the same standards as everyone else’s. They deserve a process whose results they can stand behind without having to defend themselves against someone else’s misconduct. That is one of the most important reasons to protect the integrity of an exam.