Abstract strategy learning underlies flexible transfer in physical problem solving
- Kelsey Allen, Department of Brain and Cognitive Sciences, Massachusetts Institute of Technology, Cambridge, Massachusetts, United States
- Kevin Smith, Department of Brain and Cognitive Sciences, Massachusetts Institute of Technology, Cambridge, Massachusetts, United States
- Ulyana Piterbarg, Department of Brain and Cognitive Sciences, MIT, Cambridge, Massachusetts, United States
- Robert Chen, Department of Brain and Cognitive Sciences, MIT, Cambridge, Massachusetts, United States
- Josh Tenenbaum, Brain and Cognitive Sciences, MIT, Cambridge, Massachusetts, United States
AbstractWhat do people learn when they repeatedly try to solve a set of related problems? In a set of three different physical problem solving experiments, participants consistently learn strategies rather than generically better world models. Participants selectively transferred these strategies when the crucial context and preconditions of the strategy were met, such as needing to ``catapult'', ``support'', ``launch'' or ``destabilize'' an object in the scene to accomplish their goals. We show that these strategies are parameterized: people can adjust their strategies to account for new object weights despite no direct interaction experience with these objects. Taken together, these results suggest that people can make use of limited experience to learn abstract strategies that go beyond simple model-free policies and are instead object-oriented, adaptable, and can be parameterized by model-based variables such as weight.