Dynamic decision-making is a complex process that relies on our ability to generate, evaluate and implement a variety of strategies. Understanding how people navigate this process is a difficult problem that requires a wide range of methodologies. This study details a combination of behavioral experiments, computational modeling, and neuroimaging that complement each other in describing how people engage in dynamic decision-making under uncertainty. To create opportunities to observe such decision-making participants were taught two categories by sorting sine-wave gratings selected by an adaptive, real-time computational model, PINNACLE 2.0. During the protocol, participants were presented with a challenging, dynamic decision-making task that experimentally prolongs strategy exploration while fMRI data were collected. During this task, participants displayed a broad range of behavior indicative of a variety of explicit strategy use and evaluation. Accounting for the results of this experiment proved challenging for existing computational models of category learning. In developing better accounts of the data, and how people navigated this task, we rule out several previously successful models. The models considered include exemplar, and rule-based models that include parallel and sequential strategy representations, incremental rule modification and rule replacement mechanisms, and single-step and hierarchical rule structures. Through competitive model fitting, we further the development of the PINNACLE architecture culminating in version 2.1a. This version represents the best account of participant behavior to-date. Finally, by contrasting successful and unsuccessful learning on this task, we describe preliminary evidence of the neural correlates of decision-making under uncertainty. How well people do on the task is a function of their performance expectation. Those with high expectations engage in more strategy generation, evaluation, and replacement, and tend to succeed by finding better rules whereas those with lower expectations tend to settle for less successful ones.

Creator

• Reuveni, Ben

DOI

• https://doi.org/10.21985/n2-2fjw-vf41

Subject

• Cognitive psychology
• Psychology
• Neurosciences

Language

• en

Alternate Identifier

• http://dissertations.umi.com/northwestern:15045
• etdadmin_upload_732002

Keyword

• Memory Systems
• Decision Making
• Computational Neuroscience
• Learning
• Cognitive Modeling
• Category Learning

Date created

• 2020-01-01

Resource type

• Dissertation

Rights statement

• In Copyright