Are You Doing What I Think You Are Doing? Criticising Uncertain Agent Models

TL;DR

This paper introduces a novel frequentist hypothesis testing algorithm for multiagent systems, enabling agents to evaluate the correctness of hypothesized behaviors with high accuracy and scalability, learning distributions dynamically during interactions.

Contribution

The work presents a new algorithm for testing the correctness of agent behavior hypotheses, incorporating multiple metrics and adaptive distribution learning with asymptotic guarantees.

Findings

High accuracy in hypothesis correctness evaluation

Scalable performance with low computational costs

Effective learning of distribution during interactions

Abstract

The key for effective interaction in many multiagent applications is to reason explicitly about the behaviour of other agents, in the form of a hypothesised behaviour. While there exist several methods for the construction of a behavioural hypothesis, there is currently no universal theory which would allow an agent to contemplate the correctness of a hypothesis. In this work, we present a novel algorithm which decides this question in the form of a frequentist hypothesis test. The algorithm allows for multiple metrics in the construction of the test statistic and learns its distribution during the interaction process, with asymptotic correctness guarantees. We present results from a comprehensive set of experiments, demonstrating that the algorithm achieves high accuracy and scalability at low computational costs.