Bob and Alice Go to a Bar: Reasoning About Future With Probabilistic Programs

08/09/2021

∙

by   David Tolpin, et al.

∙

13

∙

share

Agent preferences should be specified stochastically rather than deterministically. Planning as inference with stochastic preferences naturally describes agent behaviors, does not require introducing rewards and exponential weighing of behaviors, and allows to reason about agents using the solid foundation of Bayesian statistics. Stochastic conditioning is the formalism behind agents with stochastic preferences.