The market for lemons in higher education: What fake bibliographies reveal about AI and credibility

 

By Richard Sebaggala (PhD)

 

In economics, one of the most enduring insights is that markets collapse when information asymmetries exist. George Akerlof’s “Market for Lemons” has shown how buyers who cannot distinguish between good and bad used cars distrust the market as a whole. The credibility of the seller becomes crucial. Once trust has been eroded, assurances are no longer enough to restore trust. The seller must show with words and deeds that they know more than the buyer and use this knowledge responsibly. Education is also a market, even if it is not always seen in this light. Professors sell specialised knowledge, and students are the consumers. The same problem of information asymmetry now arises with the use of artificial intelligence in teaching.

A recent case at the National University of Singapore illustrates this problem. A professor assigned his literature students a reading list of six scholarly works. The students tried to locate the references but realised that none of them existed. The list had been compiled from a machine-generated bibliography. When the professor was confronted with this, he admitted that he “suspected” that some of the material came from an AI tool. At first glance, the incident seemed insignificant, as no grades were affected and the exercise was labelled as “optional” However, from a business perspective, the consequences were serious. The relationship of trust between professor and student was weakened. Students realised that even those who set the rules for the use of AI did not always know how to use the technology responsibly.

The irony is clear. Professors often warn students against outsourcing their learning to AI, citing the danger of hallucinations, fake citations or shallow thinking. But the professor who published a reading list of non-existent works made the same mistake. When the gatekeeper is unable to distinguish fact from fiction in his own assignments, students rightly question his authority to penalise them for similar transgressions. The situation is similar to that of a car dealer who asks buyers to trust his inspections but fails to recognise defective vehicles. In the long term, such failures undermine the credibility of the entire market, in this case higher education itself.

Economists also speak of signalling. People and institutions send out signals to create credibility. A degree signals competence; a guarantee signals trust in a product. Professors signal expertise through carefully designed syllabi, carefully constructed reading lists, and rigorous assessments. When students discover that a reading list is nothing more than an unchecked AI output, the signal is reversed. What should have conveyed care and competence instead conveys carelessness and over-reliance on poorly understood tools. The signal spreads: When a professor makes such a mistake, students will wonder how many others also rely on AI without educating themselves about it. If the experts appear confused, why should the rules they set be legitimate?

The economics of education depends on credibility. Students cannot directly test the quality of teaching the way they can test the durability of a chair or the performance of a phone. They have to trust their teachers. The value of their tuition, time and intellectual effort rests on the assumption that professors know what they are doing. This assumption is a fragile contract. When AI is abused, the contract comes under pressure. The information asymmetry is no longer just between professors and students, but also between the people and the technology that both groups are trying to control. If professors are unable to demonstrate their expertise, their advantage dwindles. The mentor runs the risk of becoming a middleman who could be displaced by the tools he or she does not know how to use.

This is why the debate about AI at universities cannot be reduced to prohibiting or controlling its use by students. The future will require AI skills and universities should recognise this. Professors have a responsibility not only to set rules, but also to model responsible use. This requires checking sources, cross-checking results, disclosing the use of AI and explaining its limitations and strengths. Just as central banks maintain market confidence by consistently demonstrating expertise, professors support the learning market by showing that they can use these tools with care and transparency.

The episode at NUS is more than just a minor embarrassment. It shows that the teaching profession risks losing credibility when those who are supposed to guide students appear unsure, careless or inconsistent in their use of technology. Students notice the double standard. They see that their own use of AI is strictly regulated while professors’ experiment without consequence. They hear over and over that critical thinking is important but are given assignments based on untested material. They are told that integrity is essential, yet they see the lines blurring. Economics tells us what happens as a result: Trust declines and the value of exchanges between teachers and learners diminishes.

To avoid this outcome, universities need to advocate for AI literacy rather than bans. Professors should lead by example and signal through their practise that they can guide students responsibly. This is not just a technical issue, but one of institutional credibility. Without it, the education market risks a similar loss of trust as Akerlof’s used car market. Students may begin to question why they should trust their teachers at all when the signals are inconsistent and the asymmetry so obvious. When that happens, the value of higher education itself is diminished in a way that is far more damaging than a single incorrect reading list.

To think like an economist, one must shed illusions about authority and examine the incentives and signals at work. Professors cannot warn their students about AI while abusing it themselves. They need to understand that credibility is a currency in the marketplace of learning. Once squandered, it is very difficult to regain.