How trial correlations and feedback shape sequential decision-making

To make the best decisions, organisms must flexibly accumulate information, accounting for what is relevant and ignoring what is not. Many decision-making studies focus on sequences of independent trials in which the evidence gathered to make a choice, as well as the resulting actions and feedback, are irrelevant to future decisions. Two-alternative forced choice tasks (2AFC) are often used to characterize strategies subjects use to make decisions. Normative theories, which model ideal observers, have been developed for such tasks when rewards provide the sole evidence (e.g., two-armed bandit tasks). Less is known about how observers should integrate probabilistic rewards interspersed with noisy evidence to inform their decisions in future correlated trials. To understand decision-making under more natural conditions, we extend drift-diffusion models to obtain the normative form of evidence accumulation in a series of 2AFC trials with the correct choice evolving as a two-state Markov process. We analyze 3 different feedback cases: withholding trial-to-trial feedback, giving probabilistic trial-to-trial signal, and giving probabilistic trial-to-trial reward. Ideal observers integrate noisy evidence within a trial until reaching a decision threshold and bias their initial belief depending on the evidence accumulated and feedback received on previous trials. Optimal observers accumulate more evidence on early trials and make faster decisions on later trials. Gains in performance are primarily due to biases in initial beliefs that lead to faster decisions even when feedback is lacking. Feedback shapes trial-to-trial decision strategies determining whether decisions are immediate, or a result of past and present evidence, depending on whether the feedback is strong enough to overcome the volatility of changes between trials. Our findings are also consistent with experimentally observed response trends, showing decreased reaction times when correct choices are repeated and in response to prior trial rewards.

Name thesis:degree_name

Doctor of Philosophy

Level thesis:degree_level

Doctoral

Discipline thesis:degree_discipline

Mathematics

Grantor

University of Houston

Year dc:date.issued

2020

Author dc:creator

• Nguyen, Khanh Phuong

Advisor dc:contributor.advisor

• Josić, Krešimir

Committee members dc:contributor.committeemember

• Kilpatrick, Zachary P.
• Ott, William
• Perepelitsa, Mikhail

• decision-making
• drift-diffusion model
• correlations
• feedback

Statement dc:rights

• The author of this work is the copyright owner. UH Libraries and the Texas Digital Library have their permission to store and provide access to this work. UH Libraries has secured permission to reproduce any and all previously published materials contained in the work. Further transmission, reproduction, or presentation of this work is prohibited except with permission of the author(s).

Language dc:language.iso

eng

Handle dc:identifier.uri

https://hdl.handle.net/10657/10253

OAI identifier oai:identifier

oai:uh-ir.tdl.org:10657/10253