Explaining Human Oversampling Biases on Full Information Optimal Stopping Problems: a Behavioural, Computational and Neuroimaging Investigation. / van de Wouw, Sahira.

2022.

Research output: ThesisDoctoral Thesis

Unpublished

Documents

Abstract

An optimal stopping problem can be defined as a situation in which a decision-maker has to choose a time to take a given action. Within this thesis I look at a specific type of optimal stopping problem called the full information problem on which contrasting human behaviour has been reported. In full information problems, the decision-maker first learns the probability distribution that will generate the decision options, after which option values from this generating distribution are presented in sequence, and the decision-maker has to decide when to stop sampling and choose an option, under the condition that rejected options cannot be returned to later. The decision-makers' sampling rate is then compared to that of an optimal model to determine any sampling biases (undersampling or oversampling). My novel contribution to the literature is to show that human oversampling biases on these kinds of full information problems extend from the mate choice domain to other decision-making domains including image-based domains such as trustworthiness, foods and holiday destinations, as well as number-based domains such as smartphone prices. Furthermore, I describe how the moments of the generating distribution influence both the decision-makers' and the optimal model's sampling rate, and show that a correct specification of the generating distribution is crucial for correctly identifying sampling biases. Finally, I present neuroimaging evidence indicating that similar areas in the so-called decision network are activated when a decision-maker samples too few or too many options on a full information problem.
Original languageEnglish
QualificationPh.D.
Awarding Institution
Supervisors/Advisors
Award date1 Mar 2022
Publication statusUnpublished - 2022
This open access research output is licenced under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported License.

ID: 44383585