Bayesian Inference of Self-intention Attributed by Observer

Abstract

Most of agents that learn policy for tasks with reinforcement learning (RL) lack the ability to communicate with people, which makes human-agent collaboration challenging. We believe that, in order for RL agents to comprehend utterances from human colleagues, RL agents must infer the mental states that people attribute to them because people sometimes infer an interlocutor's mental states and communicate on the basis of this mental inference. This paper proposes PublicSelf model, which is a model of a person who infers how the person's own behavior appears to their colleagues. We implemented the PublicSelf model for an RL agent in a simulated environment and examined the inference of the model by comparing it with people's judgment. The results showed that the agent's intention that people attributed to the agent's movement was correctly inferred by the model in scenes where people could find certain intentionality from the agent's behavior. CCS CONCEPTS• Computing methodologies → Theory of mind; KEYWORDS Reinforcement learning, Bayesian inference, Public self-awareness, Theory of mind, PublicSelf model, Human-agent interaction ACM Reference Format:

Conveying Intention by Motions With Awareness of Information Asymmetry

“…Although the simulation study over 2,500 conditions is complementary to the user study with a small number of conditions, the data-driven artificial observers have an inference structure different from people, which can yield different results. For example, in our previous study ( Fukuchi et al, 2018 ), people occasionally doubted the assumption that the actor intends to get either an apple or a pear and lost confidence on their answers, which never happens in simulation. Moreover, people can be affected by former trials while the artificial observers does not change after training phase.…”

“…where a is an action performed by the actor, and belief b t is a probability distribution representing the probability that the environmental state is s t The PublicSelf model extended BToM to an actor's secondorder belief inference (Fukuchi et al, 2018). We can consider PublicSelf as a computational model of an actor's public self-awareness.…”

“…To formalize this claim, we developed a novel method to generate legible motion with awareness of information asymmetry. This method is based on our previously proposed PublicSelf model, which is a computational model of public self-awareness that infers the mental states attributed to the agent by an observer ( Fukuchi et al, 2018 ). By explicitly distinguishing the observations and beliefs of an acting agent and its observer, PublicSelf can accurately predict a human’s inference of an agent’s mind in information-asymmetric situations with partial observability, but it was not applied to generating an agent’s behavior.…”

“…The PublicSelf model extended BToM to an actor’s second-order belief inference ( Fukuchi et al, 2018 ). We can consider PublicSelf as a computational model of an actor’s public self-awareness.…”

Explaining Intelligent Agent’s Future Motion on Basis of Vocabulary Learning With Human Goal Inference

“…In particular, the behavior of a person who gives instructions provides plenty of information about her/his goal. An IA motion can also affect the g human because a person sometimes tries to infer an artificial agent's mental states on the basis of mere observations of IA behavior [37], [38]. When a person tries to infer a g agent to provide instructions, an IA may need to infer how its own motion is considered by people.…”