Use this URL to cite or link to this record in EThOS:
Title: Adaptive sampling of information in perceptual decision problems
Author: Cassey, Thomas C.
Awarding Body: University of Bristol
Current Institution: University of Bristol
Date of Award: 2013
Availability of Full Text:
Access from EThOS:
When humans and other primates make decisions, cognitive, sensory, and physical restrictions often prevent the simultaneous extraction of information from all available sources. In such scenarios, the decision maker must decide how to divide their attention to optimise performance. This thesis studies a class of decision problem in which a decision maker must solve a comparative two alternative decision problem, where multiple sources of information must be sampled sequentially, over the course of a fixed period of time, before a decision is made. Critically, the thesis considers the situation in which the different sources of information have different levels of noise. To better understand such problems, optimal observer analysis is used to determine how an ideal observer should behave in the task to optimise their accuracy. Analysis of this model shows that, for maximised accuracy during decision making, the optimal observer should allocate sampling time to different sources in proportion to their noise levels. Predictions from the ideal observer model are compared to behavioural data from human observers who performed a corresponding perceptual decision-making task, revealing several differences between ideal observer predictions and human behaviour. These differences suggest that in addition to the quality of information sources, multiple other factors also influence the sampling strategy. To explain these differences, assumptions made during the derivation of the ideal observer model are modified and new predictions are compared to the experimental data. A simplified model for solving the decision problem is derived and compared to the ideal observer model as well as existing models of decision making. This analysis shows that, under certain circumstances, the simplified decision model approximates the ideal observer. In decisions in which both alternatives are equally uncertain, however, the model is equivalent to a previously developed mechanistic model of decision making.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available