• Corpus ID: 3454285

Machine Theory of Mind

  title={Machine Theory of Mind},
  author={Neil C. Rabinowitz and Frank Perbet and H. Francis Song and Chiyuan Zhang and S. M. Ali Eslami and Matthew M. Botvinick},
  booktitle={International Conference on Machine Learning},
Theory of mind (ToM; Premack & Woodruff, 1978) broadly refers to humans' ability to represent the mental states of others, including their desires, beliefs, and intentions. We propose to train a machine to build such models too. We design a Theory of Mind neural network -- a ToMnet -- which uses meta-learning to build models of the agents it encounters, from observations of their behaviour alone. Through this process, it acquires a strong prior model for agents' behaviour, as well as the… 

Theory of Mind From Observation in Cognitive Models and Humans

A cognitive ToM framework that uses a well-known theory of decisions from experience to construct a computational representation of ToM is proposed and the potential of the IBL observer model to improve human-machine interactions is discussed.

Cognitive Machine Theory of Mind

A theoreticallygrounded, pre-existent cognitive model is used to demonstrate the development of ToM from observation of other agents’ behavior and the IBL observer is able to infer the agent’s false belief and pass a classic ToM test commonly used in humans.


  • Computer Science
  • 2021
It is shown that multi- agent deep reinforcement learning models that model the mental states of other agents achieve significant performance improvements over agents with no such ToM model, and that the modeling of theory of mind in multi-agent scenarios is very much an open challenge.

Memory-Augmented Theory of Mind Network

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.

Learning Theory of Mind via Dynamic Traits Attribution

A new neural ToM architecture is proposed that learns to generate a latent trait vector of an actor from the past trajectories and thenmultiplicatively modulates the prediction mechanism via a ‘fast weights’ scheme in the prediction neural network, which reads the current context and predicts the behaviour.

Robot Learning Theory of Mind through Self-Observation: Exploiting the Intentions-Beliefs Synergy

—In complex environments, where the human sen- sory system reaches its limits, our behaviour is strongly driven by our beliefs about the state of the world around us. Accessing others’ beliefs,


Under the PR2 framework, decentralized-training-decentralized-execution algorithms are developed that are proved to converge in the self-play scenario when there is one Nash equilibrium and experiments show that it is critical to reason about how the opponents believe about what the agent believes.

Solving the Baby Intuitions Benchmark with a Hierarchically Bayesian Theory of Mind

This approach achieves near-perfect accuracy on most benchmark tasks, outperforming deep learning and imitation learning baselines while producing interpretable human-like inferences, demonstrating the advantages of structured Bayesian models of human social cognition.

A Brain-Inspired Model of Theory of Mind

A Brain-inspired Model of Theory of Mind (Brain-ToM model) is proposed, and the model is applied to a humanoid robot to challenge the false belief tasks, two classical tasks designed to understand the mechanisms of ToM from Cognitive Psychology.

Theory of Mind Modeling in Search and Rescue Teams

A modular ToM model which observes team behaviors and infers their mental states in a urban search and rescue (US&R) task and proved superior to the average judgments of human observers on all four tests of inference and better than 90th percentile observers on three of the four.



Rational quantitative attribution of beliefs, desires and percepts in human mentalizing

Social cognition depends on our capacity for ‘mentalizing’, or explaining an agent’s behaviour in terms of their mental states. The development and neural substrates of mentalizing are well-studied,

Bayesian Theory of Mind: Modeling Joint Belief-Desire Attribution

This work presents a computational framework for understanding The- ory of Mind (ToM): the human capacity for reasoning about agents’ mental states such as beliefs and desires, and expresses the predictive model of belief- and desire-dependent action at the heart of ToM as a partially observable Markov decision process (POMDP), and reconstructs an agent’s joint belief state and reward state using Bayesian inference.

Does the chimpanzee have a theory of mind?

Abstract An individual has a theory of mind if he imputes mental states to himself and others. A system of inferences of this kind is properly viewed as a theory because such states are not directly

Mirror neurons and the simulation theory of mind-reading

Psychological Reasoning in Infancy.

This evidence indicates that when infants observe an agent act in a simple scene, they infer the agent's mental states and then use these mental states, together with a principle of rationality (and its corollaries of efficiency and consistency), to predict and interpret theAgent's subsequent actions and to guide their own actions toward the agent.

Modeling Human Understanding of Complex Intentional Action with a Bayesian Nonparametric Subgoal Model

This work model how humans infer subgoals from observations of complex action sequences using a nonparametric Bayesian model, which assumes that observed actions are generated by approxi- mately rational planning over unknown subgoal sequences.

Building machines that learn and think like people

It is argued that truly human-like learning and thinking machines should build causal models of the world that support explanation and understanding, rather than merely solving pattern recognition problems, and harness compositionality and learning-to-learn to rapidly acquire and generalize knowledge to new tasks and situations.

Game Theory of Mind

It is shown it is possible to deduce whether players make inferences about each other and quantify their sophistication on the basis of choices in sequential games, and exactly the same sophisticated behaviour can be achieved by optimising the utility function itself (through prosocial utility), producing unsophisticated but apparently altruistic agents.

The Naïve Utility Calculus: Computational Principles Underlying Commonsense Psychology

Help or Hinder: Bayesian Models of Social Goal Inference

A model for how people can infer social goals from actions, based on inverse planning in multiagent Markov decision problems (MDPs), is proposed and behavioral evidence is presented in support of this model over a simpler, perceptual cue-based alternative.