Results 1 -
3 of
3
Learning the Preferences of Ignorant, Inconsistent Agents
"... An important use of machine learning is to learn what people value. What posts or photos should a user be shown? Which jobs or activities would a person find rewarding? In each case, observations of people’s past choices can inform our inferences about their likes and preferences. If we assume that ..."
Abstract
-
Cited by 1 (1 self)
- Add to MetaCart
An important use of machine learning is to learn what people value. What posts or photos should a user be shown? Which jobs or activities would a person find rewarding? In each case, observations of people’s past choices can inform our inferences about their likes and preferences. If we assume that choices are approxi-mately optimal according to some utility function, we can treat preference inference as Bayesian inverse plan-ning. That is, given a prior on utility functions and some observed choices, we invert an optimal decision-making process to infer a posterior distribution on utility func-tions. However, people often deviate from approximate optimality. They have false beliefs, their planning is sub-optimal, and their choices may be temporally in-consistent due to hyperbolic discounting and other bi-ases. We demonstrate how to incorporate these devia-tions into algorithms for preference inference by con-structing generative models of planning for agents who are subject to false beliefs and time inconsistency. We explore the inferences these models make about pref-erences, beliefs, and biases. We present a behavioral experiment in which human subjects perform prefer-ence inference given the same observations of choices as our model. Results show that human subjects (like our model) explain choices in terms of systematic de-viations from optimal behavior and suggest that they take such deviations into account when inferring pref-erences.
Learning the Preferences of Bounded Agents
"... A range of work in applied machine learning, psychology, and social science involves inferring a person’s preferences and beliefs from their choices or decisions. This includes work in economics on Structural Estimation, which has been used to infer beliefs about the rewards of education from observ ..."
Abstract
- Add to MetaCart
(Show Context)
A range of work in applied machine learning, psychology, and social science involves inferring a person’s preferences and beliefs from their choices or decisions. This includes work in economics on Structural Estimation, which has been used to infer beliefs about the rewards of education from observed work and education choices [1] and preferences for health outcomes from smoking behav-
Ten Challenges in Highly-Interactive Dialog Systems
"... Systems capable of highly-interactive dialog have re-cently been developed in several domains. This paper considers how to build on these successes to make sys-tems more robust, easier to develop, more adaptable, and more scientifically significant. ..."
Abstract
- Add to MetaCart
(Show Context)
Systems capable of highly-interactive dialog have re-cently been developed in several domains. This paper considers how to build on these successes to make sys-tems more robust, easier to develop, more adaptable, and more scientifically significant.