Conference on Cognitive Computational Neuroscience

What goes on in your head when you are solving a problem? Advances towards this question will allow us to better understand and predict complex human behaviour as well as obtain estimates for latent abilities such as how deep a person plans. We choose to contribute to this question by studying problem solving in puzzles: well-controlled and tractable environments, in particular the game of Rush Hour. We present three distinct process models for solving puzzles, fit these on subjects' data and discuss the strengths and shortcomings of each. We find that models which assume a constant depth of planning fail to capture some important trends in summary statistics. We hypothesize that people dynamically change their depth of planning and propose models to incorporate this. This work introduces psychologically-plausible process models of human behaviour to the field of human problem solving using rigorously studied puzzles.