Ph.D. Dissertation Defense_Merve Özer

Date: July 20, 2026, Monday

Time: 09:30 am

Place: FMAN 1127 Meeting Room

Presenter: Merve Özer Uysal

Dissertation Advisor: Prof. Nihat Kasap

Title: ESSAYS ON HUMAN RESOURCES DECISION-SUPPORT WITH EXPLAINABLE ANALYTICS

Abstract: This thesis develops an integrated human resources analytics framework for a leading
call center operating in Türkiye. The analysis draws on a structured dataset
comprising 11,159 employees observed over the 2022-2024 period, including narrowbandwidth
personality traits, job outcomes, and demographic measures, complemented
by job interview videos. The thesis examines how personality traits drive
employee lifetime value, shape turnover risk, and can be extracted directly from
candidates’ voices. The study first translates predicted Employee Lifetime Value
(ELTV) into a constrained selection model that maximizes each candidate’s expected
net contribution, showing that this approach yields higher operational contribution
than random and performance-only selection benchmarks. Turnover risk is
then monitored within a defined intervention window using an explainable, dynamic
early-warning model, validated against a more complex deep-learning benchmark,
then expected tenure, estimated from characteristics measured at the point of hire, is
modeled separately through survival analysis. Finally, the study examines whether
the personality traits underlying these decisions can be predicted from the acoustic
parameters of interview speech (such as emotional intensity, pitch, and speaking
rate), offering a low-cost screening layer. The findings show that personality traits
can provide a consistent, explainable, and actionable decision-support infrastructure
addressing who should be hired, who is likely to remain, and what information these
decisions should be grounded in.