Keynotes | K&Ts | GACs | Talks | Posters | Search

Poster C in Poster Session C: Wednesday, August 5, 9:30 – 11:15 am, Kimmel Center, Shorin & Rosenthal Rooms

Beyond search: reframing problem solving through cached solutions

Shuze Liu¹, Samuel J. Gershman¹; ¹Harvard University

Presenter: Shuze Liu

A classical perspective in cognitive science characterizes problem solving as search over a well-defined representation. When representations are not given, they are assumed to be chosen through similar incremental search in the space of possible representations, guided by plan utility or collisions in simulation. Yet, for many naturalistic problems, the unboundedness of problem features often makes such search computationally intractable. Drawing on insights from naturalistic decision making and design cognition, we propose that humans instead rely on solution-driven strategies: candidate solutions are recalled and reused, thus determining which problem features must be represented. Across two experiments in different domains, increasing representational demands led participants to rely more on cached solutions, representing enough information to evaluate them while failing to identify the global optimum---even when such optimal solutions enabled representing less information. These results suggest that solution-driven reasoning plays a central role in problem solving under large-world complexity.

Topic Area: Decision-Making, Cognitive Control & Event Cognition