Memory-Augmented Theory of Mind Network

TL;DR

This paper introduces ToMMY, a neural network model with memory and hierarchical attention that improves theory of mind reasoning in complex, dynamic scenarios, especially for false-belief tasks.

Contribution

It presents a novel memory-augmented neural network architecture for theory of mind reasoning, enabling better handling of complex, changing social scenarios.

Findings

Memory mechanisms enhance ToM learning efficiency.

ToMMY outperforms previous models on false-belief tasks.

Hierarchical attention improves information retrieval for reasoning.

Abstract

Social reasoning necessitates the capacity of theory of mind (ToM), the ability to contextualise and attribute mental states to others without having access to their internal cognitive structure. Recent machine learning approaches to ToM have demonstrated that we can train the observer to read the past and present behaviours of other agents and infer their beliefs (including false beliefs about things that no longer exist), goals, intentions and future actions. The challenges arise when the behavioural space is complex, demanding skilful space navigation for rapidly changing contexts for an extended period. We tackle the challenges by equipping the observer with novel neural memory mechanisms to encode, and hierarchical attention to selectively retrieve information about others. The memories allow rapid, selective querying of distal related past behaviours of others to deliberatively reason about their current mental state, beliefs and future behaviours. This results in ToMMY, a theory of mind model that learns to reason while making little assumptions about the underlying mental processes. We also construct a new suite of experiments to demonstrate that memories facilitate the learning process and achieve better theory of mind performance, especially for high-demand false-belief tasks that require inferring through multiple steps of changes.