Too Many Cooks: Bayesian Inference for Coordinating Multi‐Agent Collaboration

Abstract

Collaboration requires agents to coordinate their behavior on the fly, sometimes cooperating to solve a single task together and other times dividing it up into sub-tasks to work on in parallel. Underlying the human ability to collaborate is theory-of-mind (ToM), the ability to infer the hidden mental states that drive others to act. Here, we develop Bayesian Delegation, a decentralized multiagent learning mechanism with these abilities. Bayesian Delegation enables agents to rapidly infer the hidden intentions of others by inverse planning. We test Bayesian Delegation in a suite of multiagent Markov decision processes inspired by cooking problems. On these tasks, agents with Bayesian Delegation coordinate both their high-level plans (e.g., what sub-task they should work on) and their low-level actions (e.g., avoiding getting in each other's way). When matched with partners that act using the same algorithm, Bayesian Delegation outperforms alternatives. Bayesian Delegation is also a capable ad hoc collaborator and successfully coordinates with other agent types even in the absence of prior experience. Finally, in a behavioral experiment, we show that Bayesian Delegation makes inferences similar to human observers about the intent of others. Together, these results argue for the centrality of ToM for successful decentralized multi-agent collaboration.

Resonating Minds—Emergent Collaboration Through Hierarchical Active Inference

“…The Overcooked domain has been previously tackled using Deep Reinforcement Learning to train a collaborative policy from minimal domain knowledge [11,39]. A BToM approach has also been used in a simplified version of the task [40]. In the latter approach, Wu et al propose Bayesian Delegation, a Bayesian model that enables (joint) task inference using knowledge about sub-tasks with pre-and post-conditions to determine partial orderings of different recipes.…”

“…In the latter approach, Wu et al propose Bayesian Delegation, a Bayesian model that enables (joint) task inference using knowledge about sub-tasks with pre-and post-conditions to determine partial orderings of different recipes. Each "agent simulates a fictitious centralized planner that controls the actions of all agents working together on the same sub-task" ([ [40], p. 422]) in order to determine the best action to take for the agent. That is, the approach is to search for the most suitable collaborative action in order to complete the task in as few steps as possible.…”

“…In the present work, we ask how successful collaboration can emerge based on cognitively simpler means, without explicitly representing and solving an inherently complex multi-agent decision or planning problem. We use the Overcooked domain in the setting presented in [11] and additionally evaluate our model in the version discussed in [40] (Results). Our approach is related to the one presented in [40] as we also employ a probabilistic BToM model to identify and adapt to another agent's underlying goals on-the-fly.…”

“…We use the Overcooked domain in the setting presented in [11] and additionally evaluate our model in the version discussed in [40] (Results). Our approach is related to the one presented in [40] as we also employ a probabilistic BToM model to identify and adapt to another agent's underlying goals on-the-fly. We also provide domain knowledge in the form of possible (action) intentions, such as fetching a certain item or preparing an ingredient, and their likelihoods given an observation of the environment.…”

Balancing Performance and Human Autonomy With Implicit Guidance Agent

“…One type of intuitive collaborative agent is the supportive agent , which helps a human by predicting the human’s objective and planning an action that would best help achieve it. In recent years, there have been agents that can plan effectively by inferring human subgoals for a partitioned problem based on the Bayesian Theory of Mind ( Wu et al, 2021 ). Other agents perform biased behavior for generic cooperation, such as communicating or hiding their intentions ( Strouse et al, 2018 ), maximizing the human’s controllability ( Du et al, 2020 ), and so on.…”