How do people learn to solve a hard problem? Gautam Agarwal, PhD-@Seay 2.108

How do people learn to solve a hard problem?

While neural networks approach human levels of performance in many complex tasks, they require much more training than humans. This may be because only humans can infer and apply generalizable principles from prior experience. However, the statistics that underlie human skill learning are poorly understood and hard to investigate in the large state spaces found in most complex tasks. We thus designed a cognitive task that is constrained enough to densely sample human strategies, but complex enough to compel intelligent search. We launched the game as a smartphone-based app to collect data from ~10k human participants. We find humans (1) search a highly restricted subset of the policy space; (2) attempt even poor solutions many times before discarding them; (3) arrive at the optimal policy suddenly and unpredictably with a “leap of insight”. Our data suggest a “top-down” learning process by which humans propose explanatory solutions which they replace only upon collecting sufficient evidence to the contrary.

Attendees of this workshop are invited to play the app beforehand (available at hexxed.io) and complete at least the first level. After I motivate our approach and describe our findings, we will have an open discussion of issues like: What are the strengths and weaknesses of games as a tool to study skill learning? What does a satisfying account of skill learning look like?

Bio:
Starting as an experimentalist studying olfactory neuropil of flies, Gautam turned to theory believing that it offered a unique language to express the staggering complexity of brains. Dr. Agarwal studied neural oscillations as a postdoc with Dr. Friedrich Sommer at the Redwood Center at UC Berkeley (2011-2014). Gautam then lived in Lisbon, where he worked with Dr. Zach Mainen to develop a behavioral task for deconstructing human intelligence. Gautam is an Assistant Professor of Neuroscience at the Claremont Colleges, where he is thinking about ways of communicating theoretical neuroscience in the context of a liberal arts education.