Abstract Details

Title Predicting Open-Ended Decisions: Understanding the Differential Roles of Semantic Memory and Preference

Topic Aging, Dementia, and Behavioral Neurology

Presentation(s) Behavioral and Cognitive Neurology Posters (7:00 AM-5:00 PM)

Poster/Presentation
Number 032

Objective Decisions made in a controlled testing environment may not correlate with decision-making in the real world. For decisions such as how to spend one's evening, the space of potential options and goals is less well-defined and may require generation by the decision-maker. Here we take a step toward predicting open-ended choices by expanding current models of decision-making to incorporate processes enabling option generation.

Background Real-world decisions are often open-ended, with goals and/or choice options conceived by decision-makers themselves. These requirements, however, are largely absent from classical decision-making models. Here we address this gap by (1) developing a neurally-inspired model of decisions in which choice options are self-generated, and (2) confirming its neural substrates via functional MRI (fMRI).

Design/Methods Using a between-subjects design in 2,811 participants, we compared choices from an “external menu condition” (EMC), where an external menu is present, with those in an “internal menu condition” (IMC), where it is not, for 6 categories of real-world goods. An additional 256 subjects completed a semantic fluency task over these categories. We then developed a retrieval-constrained valuation (RCV) model in which valuation, captured by a multinomial logit function, operates on an “internal menu” generated probabilistically via censored random walk on an associative network. We then conducted fMRI scans of 32 subjects while they performed the EMC, IMC, and fluency tasks.

Results Compared to models that excluded semantic memory or preference, the RCV model showed dramatic improvement in out-of-sample prediction, accurately identifying the entire response profile across categories (mean R² of 0.94, range 0.87 to 0.97). On fMRI, consistent with the model, IMC demonstrated greater engagement of valuation systems than fluency (p < 0.05, corrected), and greater engagement of semantic retrieval systems than EMC (p < 0.05, corrected).

Conclusions Together these results take a step toward unraveling cognitive mechanisms underlying adaptive decision-making in the real-world.

Authors/Disclosures
Andrew Kayser, MD, PhD, F�鶹��ýӳ�� (UCSF) PRESENTER	Dr. Kayser has received personal compensation in the range of $500-$4,999 for serving as a Consultant for Boehringer Ingelheim. The institution of Dr. Kayser has received research support from NIH. The institution of Dr. Kayser has received research support from VA RR&D.
	No disclosure on file
	No disclosure on file

�鶹��ýӳ��

�鶹��ýӳ��