Probing Physics Knowledge Using Tools from Developmental Psychology

By Luis S. Piloto, Ari Weinstein, TB Dhruva, Arun Ahuja, Mehdi Mirza, Greg Wayne, David Amos, Chia-Chun Hung, Matthew Botvinick. 2018

38 citations recorded by the source corpus.

In order to build agents with a rich understanding of their environment, one key objective is to endow them with a grasp of intuitive physics; an ability to reason about three-dimensional objects, their dynamic interactions, and responses to forces. While some work on this problem has taken the approach of building in components such as ready-made physics engines, other research aims to extract general physical concepts directly from sensory data. In the latter case, one challenge that arises is evaluating the learning system. Research on intuitive physics knowledge in children has long employed a violation of expectations (VOE) method to assess children's mastery of specific physical concepts. We take the novel step of applying this method to artificial learning systems. In addition to introducing the VOE technique, we describe a set of probe datasets inspired by classic test stimuli from developmental psychology. We test a baseline deep learning system on this battery, as well as on a physics learning dataset ("IntPhys") recently posed by another research group. Our results show how the VOE technique may provide a useful tool for tracking physics knowledge in future research.

Subfields: Cognitive Neuroscience, Education

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 Education

Citations in the corpus

Flexible Neural Representation for Physics Prediction (2018, background): “Finally, it will be of great interest to evaluate to what extent HRN-type models describe patterns of human intuitive physical knowledge observed by cognitive scientists [32, 35, 38].”
Machine Common Sense Concept Paper (2018, methodology): “Moreover, they demonstrated a scheme for using the same VOE method used in developmental psychology to evaluate how well the artificial models mimic child cognition [65].”
IntPhys: A Benchmark for Visual Intuitive Physics Reasoning (2019, background): “There is also recent dataset proposed by a DeepMind team [33].”
PHYRE: A New Benchmark for Physical Reasoning (2019, methodology): “Many methods are evaluated on the constrained task of predicting whether a 3D stack of blocks will topple; some recent studies instead ask models to determine whether videos of more complex scenes are physically plausible [39, 41].”
Physion: Evaluating Physical Prediction from Vision in Humans and Machines (2021, background): “Existing benchmarks that test physical prediction and reasoning often consider only a small set of principles to test – for instance, judging whether a tower of blocks will fall [29], or whether a viewed object’s trajectory violates basic physical laws [49, 46, 57].”