AI perception in personnel selection: Stakeholder perspective
How to cite
Kandul, Serhiy; Leicht-Deobald, Ulrich. 2023. "AI perception in personnel selection: Stakeholder perspective." AEA RCT Registry. AEARCTR-0011208. https://doi.org/10.1257/rct.11208-1.1
Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.
* Primary investigator
Abstract
AI-based personell selection tools are efficient. However, they can generate disparities in selection rates across groups of applicants and therefore violate fairness requirements. In this project we investigate how people perceive this efficiency-fairnes tradeoffs. After having observed people's general fairness evaluation and pferences over efficient and fair algorithms, in this study we are interested in how the stakeholder perspective affects the AI perceptions in HR context. For this purpose, we compare the tradeoff perceptions by the UK managers and students in the role of job applicants. Based on the prominent fairness theory in organizational justice, we expect that managers prefer more efficient algorithms, while students favour les efficient algorithms with lower disparity rates.
Intervention
We manipulate stakeholder perspective and see how this affects people's preferences over efficient and fair selection algorithms
Experimental design
We present participants with a series of selection algorithms. The algorithms are defined by two major features: efficiency, measured as the quality of the selection, and faireness, measured by the degree of disparity in selection rates across gender. There is always efficiency-fairnes tradeoffs. We manipulate the stakeholder perspective (between-subjects): the managers take the role of HR managers in the company, while students play as job applicants and the degree of the efficiency-fairness tradeoffs (within-subject). We then compare the fairness perceptions and choices over algorithms of managers and students.
Design details
Cited by
No citation data available for this registration yet.