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

Metacognitive sensitivity governs adaptive representational compactness in hierarchical structure learning

Rochelle Kaper¹, Megan AK Peters¹; ¹University of California, Irvine

Presenter: Rochelle Kaper

Humans form compact hierarchical representations and metacognitively evaluate their own uncertainty to cope with information-processing constraints. It is unknown how hierarchical structure, coupled with metacognitive evaluation, jointly affect predictive representations and their influence on behavior. Notably, unfamiliar stimuli require more working memory (WM) resources for encoding, but it is further unknown if such stimulus-driven WM capacity leads to more compact representations that blend, rather than distinctly represent, multiple structures. To address these gaps, we fit an analytical Successor Representation (SR) model to choice behavior from a hierarchical learning paradigm where participants learned co-occurrences embedded in a sequence across two sessions with either familiar line drawings or unfamiliar fractal stimuli. We found a tradeoff between the discount factor (γ; representing the degree of compactness) and the influence of representations on decisions (β): fractals elicited a greater γ, but line drawings drove higher likelihood of representations influencing choices, suggesting that fractal-induced reductions in available WM capacity may drive greater compactness of structure representations. Further, fractal stimuli elicited a negative relationship between metacognitive sensitivity and discount factor (γ), but a positive relationship between metacognitive sensitivity and behavioral dependence on SR (β). Adaptively, higher metacognitive sensitivity may support representational compactness and use of internal representations of hierarchical environmental structure.

Topic Area: Memory, Learning & Knowledge Structures