Learning Probabilistic Responsibility Allocations for Multi-Agent Interactions

2026-04-13

AI Summary

This work addresses the challenge of modeling human responsibility allocation in multi-agent interactions to enhance the social compliance and trustworthiness of autonomous systems. It proposes the first framework that formulates responsibility assignment as a multimodal probability distribution, integrating conditional variational autoencoders with multi-agent trajectory prediction to learn context-aware responsibility distributions conditioned on both scene and agent-level contexts. By incorporating a differentiable optimization layer, the model enables end-to-end training without requiring ground-truth responsibility labels. Experiments on the INTERACTION driving dataset demonstrate that the approach achieves strong predictive performance while offering interpretable insights into agent interactions through the lens of responsibility attribution.

Technology Category

• Multiagent Systems: Adversarial Agents
• Humans and AI: Human-Aware Planning and Behavior Prediction
• Machine Learning: Multimodal Learning

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization

Abstract

Human behavior in interactive settings is shaped not only by individual objectives but also by shared constraints with others, such as safety. Understanding how people allocate responsibility, i.e., how much one deviates from their desired policy to accommodate others, can inform the design of socially compliant and trustworthy autonomous systems. In this work, we introduce a method for learning a probabilistic responsibility allocation model that captures the multimodal uncertainty inherent in multi-agent interactions. Specifically, our approach leverages the latent space of a conditional variational autoencoder, combined with techniques from multi-agent trajectory forecasting, to learn a distribution over responsibility allocations conditioned on scene and agent context. Although ground-truth responsibility labels are unavailable, the model remains tractable by incorporating a differentiable optimization layer that maps responsibility allocations to induced controls, which are available. We evaluate our method on the INTERACTION driving dataset and demonstrate that it not only achieves strong predictive performance but also provides interpretable insights, through the lens of responsibility, into patterns of multi-agent interaction.

Problem

• responsibility allocation
• multi-agent interactions
• probabilistic modeling
• human behavior
• social compliance

Innovation

• probabilistic responsibility allocation
• conditional variational autoencoder
• multi-agent trajectory forecasting
• differentiable optimization layer
• interpretable interaction modeling