2026-05-22T15:42:43Zhttps://keep.lib.asu.edu/oai/request

oai:keep.lib.asu.edu:node-2023452025-08-18T22:22:09Zoai_pmh:alloai_pmh:repo_items

202345 https://hdl.handle.net/2286/R.2.N.202345 http://rightsstatements.org/vocab/InC/1.0/ All Rights Reserved 2025 102 pages Masters Thesis Academic theses en Gajavalli, Sri Harsha Hasan, Rakibul Bansal, Srividya Choi, YooJung Arizona State University Partial requirement for: M.S., Arizona State University, 2025 Field of study: Software Engineering Artificial Intelligence (AI) systems increasingly rely on personal data to make high-stakes decisions in contexts such as college admissions and hiring, raising concerns about fairness and potential social harms. This thesis investigates how individuals perceive the fairness of Artificial Intelligence systems based on the types of personal data used in these decisions. A large-scale survey was conducted with 400 United States (U.S.) college students, who evaluated 14 potential harms across six common data attributes (e.g., demographics, personality traits, emotional state) in both educational and employment contexts. Findings indicate that attributes such as emotional state, personality traits, and disability status were perceived as especially sensitive. Harms such as bias, stereotyping, and manipulation were rated as the most severe. Perceptions of fairness also varied significantly by context and participant demographics, including gender, race, and academic discipline. Based on these results, the thesis proposes a practical framework for identifying perceived data-use harms and offers design recommendations—such as avoiding sensitive features, increasing transparency, and enhancing user agency—to help align Artificial Intelligence systems with public expectations of fairness and ethical data practices. Computer Science AI in Education AI in Employment Algorithmic fairness Human-Centered AI Perceived Fairness Risks Privacy Harms Perceptions of Fairness and Privacy in AI-Driven Recruitment and Education - Examining the Use of Candidate Attributes in Algorithmic Decision Making