For this project, we explored Partially Observable Markov Decision Processes (POMDPs) and their capacity for exhibiting curiosity. Contributions included researching formulations of curiosity in the cognitive, behavioral, and biological sciences, and devising experiment environments for a virtual gridworld where curiosity-esque behavior is required to solve a puzzle. We concluded POMDPs are too limited to exhibit phenomena mirroring curiosity.
Research was performed in the Personal Robotics Lab at the University of Washington under Professor Siddhartha Srinivasa with the supervision of William Agnew and Matt Barnes.