Date of Award
2024
Document Type
Honors Thesis (Open Access)
Department
Colby College. Economics Dept.
Advisor(s)
Michael Donihue
Abstract
Two studies were conducted to evaluate the effectiveness of AI tools and keyword searchers in candidate evaluation. The first study examined the use of ChatGPT-4 and a traditional hiring committee in selecting a semifinalist pool for an Assistant Professor of Economics position. It was found that ChatGPT-4's selections were non-replicable, producing different semifinalist lists with each run, indicating significant variability and lack of reliability compared to human committees. The second study focused on predicting self-reported conscientiousness from resumes using both keyword searchers and ChatGPT-4. The results showed that neither method could accurately predict conscientiousness. Similarly, ChatGPT-4's selections were non-replicable, producing different ratings of conscientiousness with each run. These findings suggest substantial limitations in the current use of AI and keyword searchers for academic hiring and personality trait assessment.
Keywords
AI, Resume Analysis, Personality Prediction, Keyword Searcher, Hiring Decisions
Recommended Citation
Snyder, Sarah, "Comparing Generative AI, Keyword Searchers, and Hiring Committees in Selecting Semi-finalist Job Candidates in Academia" (2024). Honors Theses. Paper 1452.https://digitalcommons.colby.edu/honorstheses/1452