The Reasons that Agents Act: Intention and Instrumental Goals
Francis Rhys Ward, Matt MacDermott, Francesco Belardinelli, Francesca Toni, Tom Everitt
TL;DR
The paper tackles how to rigorously define AI intention by grounding it in structural causal influence models and linking subjective beliefs and utilities to intentional action. It shows that an operational notion of intention can be captured behaviorally and equivalently under robust optimality, bridging to notions like actual causality and instrumental goals. The work establishes a formal framework for inferring intentions in RL agents and language models, with connections to safety concepts such as instrumental control incentives. The approach provides practical criteria for evaluating intent in real-world ML systems and aligns philosophical insights with computational models to support accountability and safety.
Abstract
Intention is an important and challenging concept in AI. It is important because it underlies many other concepts we care about, such as agency, manipulation, legal responsibility, and blame. However, ascribing intent to AI systems is contentious, and there is no universally accepted theory of intention applicable to AI agents. We operationalise the intention with which an agent acts, relating to the reasons it chooses its decision. We introduce a formal definition of intention in structural causal influence models, grounded in the philosophy literature on intent and applicable to real-world machine learning systems. Through a number of examples and results, we show that our definition captures the intuitive notion of intent and satisfies desiderata set-out by past work. In addition, we show how our definition relates to past concepts, including actual causality, and the notion of instrumental goals, which is a core idea in the literature on safe AI agents. Finally, we demonstrate how our definition can be used to infer the intentions of reinforcement learning agents and language models from their behaviour.