Modeling Intention Inference as a Basis of Imitation Learning: Analysis of Pendulum Swing-Up Task

Abstract

Learning an action from others require to infer their underlying intentions. Psychological studies have reported behavioral evidences that young children do infer others’ underlying intentions by observing their actions. The objective of the present study is to propose a mechanistic account for how intention inference is possible by observing others’ actions. For this purpose, we performed a series of simulations in which two agents control pendulums for different tasks and goals, and analyzed which types of features is informative to infer their latent intentions. Our analysis showed that a type of fractal dimension of the pendulum movements is sufficiently informative to classify the types of agents. With respect to its invariant nature, our results suggest that the fine-grained movement patterns such as the fractal dimension reflect the structure of the underlying intentions.