PhD student Vael Gates and Professors Anca Dragan and Tom Griffiths published “How to Be Helpful to Multiple People at Once” in the journal Cognitive Science.

The authors consider the problem of deciding how to assist multiple recipients with very different preferences. They develop a quantitative model of what people consider desirable behavior, characterizing participants’ preferences by inferring which combination of “metrics” (maximax, maxsum, maximin, or inequality aversion) best explain participants’ decisions in a drink‐choosing task. The paper helps constrain the space of desirable behavior in both leaders and assistive artificial intelligence systems.