Action understanding as inverse planning.

By Chris L. Baker, Rebecca Saxe, Josh Tenenbaum. 2009

1,001 citations recorded by the source corpus.

Humans are adept at inferring the mental states underlying other agents' actions, such as goals, beliefs, desires, emotions and other thoughts. We propose a computational framework based on Bayesian inverse planning for modeling human action understanding. The framework represents an intuitive theory of intentional agents' behavior based on the principle of rationality: the expectation that agents will plan approximately rationally to achieve their goals, given their beliefs about the world. The mental states that caused an agent's behavior are inferred by inverting this model of rational planning using Bayesian inference, integrating the likelihood of the observed actions with the prior over mental states. This approach formalizes in precise probabilistic terms the essence of previous qualitative approaches to action understanding based on an "intentional stance" [Dennett, D. C. (1987). The intentional stance. Cambridge, MA: MIT Press] or a "teleological stance" [Gergely, G., Nádasdy, Z., Csibra, G., & Biró, S. (1995). Taking the intentional stance at 12 months of age. Cognition, 56, 165-193]. In three psychophysical experiments using animated stimuli of agents moving in simple mazes, we assess how well different inverse planning models based on different goal priors can predict human goal inferences. The results provide quantitative evidence for an approximately rational inference mechanism in human goal inference within our simplified stimulus paradigm, and for the flexible nature of goal representations that human observers can adopt. We discuss the implications of our experimental results for human action understanding in real-world contexts, and suggest how our framework might be extended to capture other kinds of mental state inferences, such as inferences about beliefs, or inferring whether an entity is an intentional agent.

Subfields: Cognitive Neuroscience, Developmental and Educational Psychology

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Cognitive Neuroscience

In Developmental and Educational Psychology

Citations in the corpus

Theory of Mind: A Neural Prediction Problem (2013, positive): “Like vision, theory of mind can be understood as an inverse problem (Baker et al., 2011; Baker et al., 2009); the challenge is to use the observable evidence (in this case human behaviors and states) to infer the invisible causal structure that gave rise to the evidence (the goals, thoughts, and…”
An Algorithmic Perspective on Imitation Learning (2018, background): “…called inverse optimal control [Kalman, 1964, Moylan and Anderson, 1973, Dvijotham and Todorov, 2010, Levine and Koltun, 2012], inverse planning [Baker et al., 2009], or structural estimation of MDPs Rust [1994] the learner tries to recover a reward function from a policy (or demonstrations of…”
Learning the Preferences of Ignorant, Inconsistent Agents (2015, methodology): “This approach is known as Inverse Reinforcement Learning (Ng and Russell 2000) in an RL setting and as Bayesian Inverse Planning (Baker, Saxe, and Tenenbaum 2009) in the setting of probabilistic generative models.”
Maximum Entropy Deep Inverse Reinforcement Learning (2015, background): “…in robotics (Argall et al., 2009) where it allows robot to learn complex behaviour from human demonstrations and also in fields of cognition (Baker et al., 2009) and preference learning (Ziebart et al., 2008) where it serves Figure 1: Fully Convolutional Neural Network for reward…”
Inferring the Why in Images (2014, background): “However, we are inspired by the observation that humans have the remarkable cognitive ability to think about other people’s thinking [2, 30].”
Reward-rational (implicit) choice: A unifying formalism for reward learning (2020, background): “Boltzmann-rational policies are widespread in psychology [13, 14, 15], economics [16, 17, 18, 17], and AI [19, 20, 21, 22, 23] as models of human choices, actions, or inferences.”
Predicting Motivations of Actions by Leveraging Text (2014, background): “However, humans do have the ability to think about other people’s thinking [3, 32].”
Rational quantitative attribution of beliefs, desires and percepts in human mentalizing (2017, methodology): “Earlier model-based accounts 14–16 used various adaptations of Markov decision processes (MDPs), special cases of POMDPs, which assume that the world state is fully observed and known to the agent.”
Probabilistic Human Intent Recognition for Shared Autonomy in Assistive Robotics (2019, methodology): “Bayesian models are shown to be effective for inference in cognitive science [4] and human-robot interaction research [36].”
Intuitive Theories (2020, background): “In subsequent experiments, Baker et al. (2009) also showed how their account can handle cases in which the agent's trajectory contradicts a simple view of rational action (e.g. when the agent heads toward B at t 2 in Figure 9c) by assuming that the agent's goals might change over time or that the…”
Prediction of Human Activity by Discovering Temporal Sequence Patterns (2014, methodology): “lated the prediction task as a decision-making process [57] and proposed to forecast human behavior by leveraging the recent advances in semantic scene labeling [35], [36] and inverse optimal control [4], [1], [56], [31].”
The Naïve Utility Calculus: Computational Principles Underlying Commonsense Psychology. (2016, background): “, [41]), we propose that these reverse inferences can be understood as a type of Bayesian inference (Box 1).”