Learning from Experts: A Survey

TL;DR

This survey reviews methods of information transfer from experts to non-experts, focusing on game-theoretic and Bayesian decision-theoretic approaches, highlighting their differences and implications.

Contribution

It provides a comprehensive comparison of two main theoretical frameworks for expert-non-expert communication, clarifying their assumptions and relationships.

Findings

Different premises lead to distinct models of expert communication

Limited connection between game-theoretic and Bayesian approaches

Highlights need for integrated frameworks in expert information transfer

Abstract

The survey is concerned with the issue of information transmission from experts to non-experts. Two main approaches to the use of experts can be traced. According to the game-theoretic approach expertise is a case of asymmetric information between the expert, who is the better informed agent, and the non-expert, who is either a decision-maker or an evaluator of the expert's performance. According to the Bayesian decision-theoretic approach the expert is the agent who announces his probabilistic opinion, and the non-expert has to incorporate that opinion into his beliefs in a consistent way, despite his poor understanding of the expert's substantive knowledge. The two approaches ground the relationships between experts and non-experts on so different premises that their results are very poorly connected.