Name: Vikram Sahgal
Masters Thesis Defense Meeting
Date: Thursday, April 17th, 2025
Time: 12:00 pm
Location: JS Coon 148 or Virtual

Zoom Link: https://gatech.zoom.us/j/96073960929?pwd=vuAaJaRX1LrbaTVwfNbopleJX4BBpy.1

Thesis Chair/Advisor:
Rick Thomas, Ph.D. (Georgia Tech)

Thesis Committee Members:
Christopher Stanzione, Ph.D. (Georgia Tech)
Title: COST SENSITIVITY IN EXTERNAL INFORMATION SEARCH AND HYPOTHESIS TESTING
Abstract: This thesis investigates how individuals integrate cost considerations into hypothesis-guided information search during diagnostic decision-making. Building on cognitive models such as HyGene's hypothesis-guided search and theories of information foraging, this research examines how participants adaptively trade informational utility with access costs. Using the Medical Diagnosis Game (MDG), participants engaged in simulated clinical decision tasks requiring them to generate hypotheses, select diagnostic medical tests, and make final diagnoses under varying test costs. Participants completed a cost-free training phase in a two-phase experiment to learn cue-diagnosis-test associations. In the test phase, costs were introduced for each diagnostic test, requiring participants to strategically evaluate and tradeoff expected informational utility with information access cost and potential in-game monetary penalty for incorrect diagnoses. Results showed that participants adjusted test selection behavior in response to beliefs in hypotheses based on information they had sampled and the cost structures. Participants preferred to select higher-cost tests when their expected information utility justified the expense. They also demonstrated increased search termination when initial test results yielded max posterior belief gains. Moreover, participants who were more consistent in their early-stage decision consistency, as captured by entropy-based measures, exhibited better overall diagnostic accuracy. These findings support the roles of hypothesis generation, belief updating, and cost sensitivity in hypothesis-guided search. Thus, participants adapted their search strategies in response to the expected informativeness of the tests and cost constraints. The results contribute to a growing body of literature linking decision variability, information foraging, and search termination behavior to learning and cognitive limitations in applied contexts such as medical diagnosis.