University Recruitment: When AI makes articles easy to produce, do publications still have to be counted?

Two researchers model through game theory the dilemma of young researchers, between understanding in depth or inflating their CV with AI, and that of recruiters who must choose what to evaluate.

For a long time, a well-written paper containing a solid theoretical result proved that its authors mastered their field. With AIs capable of producing this type of text, this signal is much cheaper to produce and no longer allows candidates to be sorted according to their actual understanding, especially in mathematics and theoretical sciences. Saptarshi Pal and Peter S. Park draw the consequences for the academic job market.

A dilemma on both sides

The young researcher must choose between investing in a deep understanding of his or her field or relying on AI to align publications. The recruiter, on the other hand, may favor candidates who are less prolific but who pass an interview designed to probe their understanding, or demand both volume and success in the interview. The authors formalize these choices in a game theory model, identify all possible balances and simulate the behaviors that populations of candidates and recruiters end up adopting.

The work does not decide for a single policy, but provides a framework for evaluating recruitment practices that can continue to reward actual understanding rather than the amount of work posted.