Human active learning in dynamic physical microworlds
Many aspects of our physical environment are hidden. For example, we can't confidently determine the masses of objects using our eyes alone. In this work, we examine how people actively “experiment” within the physical world to discover the latent physical properties of objects. We presented participants with moving objects in “microworlds” that operate according to continuous spatiotemporal dynamics similar to everyday physics (i.e., forces of gravity, friction, etc...). Participants were asked to interact with objects in the microworlds to identify the masses or forces of attraction/repulsion that governed the objects’ movement. A quantitative analysis of the information generated by the learner's actions shows that people selectively emphasise evidence which reveals the physical property consistent with their inquiry goal. In addition, inference about the objects’ latent properties was more accurate among “active” learners who freely interacted with the physical system over passive movie watchers including a yoked set who watched replays of an active learner's interactions. Active participants performed a variety of human-codeable “natural experiments” that revealed the objects’ properties with varying success. I will end my talk by sketching ongoing work exploring how these experimental strategies may be discovered through play, and mental simulation.
Add to your calendar
18 August 2017 - Neil Bramley: Seminar
Informatics Forum 4.31/4.33