Some variations of the secretary problem

Researchers Sarthak Agrawal and Sanjeev Saxena have published a new paper titled 'Some variations of the secretary problem' that introduces two practical twists to the classic optimal stopping problem. The first variation models real-world scenarios where candidates might reappear—each interviewee has a fixed probability p of appearing a second time. The researchers characterized the optimal threshold rule for this scenario, showing how the additional information from repeated appearances can be leveraged to improve selection performance beyond the classical 37% success rate.

The second, potentially more impactful variation redefines success from selecting the single best candidate to selecting any one of the top three candidates. This model acknowledges that in many real hiring or selection processes, several top candidates may be nearly equally qualified. The analysis reveals that this relaxed criterion significantly increases the probability of a successful hire and, interestingly, shifts the optimal stopping threshold earlier in the interview sequence. This provides a mathematical foundation for more pragmatic AI-driven decision systems where perfection is less critical than securing a high-quality outcome.

These algorithmic advancements move the secretary problem from a theoretical puzzle to a tool with direct applications. The models offer a formal framework for building smarter AI agents in recruitment platforms, investment systems, and real-time recommendation engines where sequential, irreversible decisions must be made under uncertainty. The work bridges discrete mathematics, game theory, and practical algorithm design.