Given once, 2020

Fair Division is Hard even for Amicable Agents

Even when agents have different tastes, fair division with minimal sharing remains hard

Fair division · Computational social choice

An advanced research talk, for graduate students and up.

Abstract

Draft, not yet checked by me

Indivisible goods cannot always be split fairly, but letting agents share a small number of goods can bring back both fairness and efficiency. How little sharing actually does the job is the question. Sandomirskiy and Segal-Halevi showed that envy-free, fractionally Pareto-optimal allocations with minimal sharing can be found efficiently for any fixed number of agents whose valuations are far from identical (the "amicable" case). In joint work with Aditi Sethia, I show that this does not scale: with many agents, the problem is NP-complete even at degeneracy one and with no sharing at all. I also give another way to enumerate consumption graphs with few sharings.

The slides

The slides for this one are not online.

Where I have given it

All appearances