 | , - |
 | Dauphine A409 |
Événement en anglais.
Assessment Aggregation under Conflict of Interest (with Yoko Kawada, Yuta Nakamura and Noriaki Okamoto)
Collective evaluation problems under conflict of interest (e.g., the peer-grading system) are attracting increasing attention across many academic fields, such as economics, computer science, sociology, and education. The problem of assessment aggregation under conflict of interest is not limited to peer evaluation situations. In this paper, we allow arbitrary interest relationships between the evaluators and evaluatees. Then, the prevention of self-interested behavior among recipients of the evaluation outcome and of biased inputs to the evaluation process should be separately formulated.
Strategic-Networking-Proofness prevents outcome rank improvement of an evaluatee when a new interest relationship is formed between the evaluatee and some evaluator, and Interest-Exclusion prevents evaluators' potentially biased inputs from being counted in the evaluation process. We found that these two requirements are incompatible with Unanimity. By constructing concrete rules, we identify two thresholds of weaker conditions such that a desirable aggregation rule exists if and only if those are imposed. One is the winning-rate rule, which extends the Borda count, and the other is an extended plurality rule. Our main impossibility theorem applies to natural extensions of the aggregation rules used in the peer-evaluation literature. By Strategic-Networking-Proofness, we prevent favoritism via networking, connecting network theory and social choice theory.
Inscription possible prochainement.
