Experimental Design
This study uses a between-subject experimental design. Participants acting as employers will be randomly assigned to one of two applicant environments: an AI environment, in which all applicant profiles are drawn from job seekers who completed the relevant tasks with access to generative AI, or a no-AI environment, in which all applicant profiles are drawn from job seekers who completed the tasks without access to generative AI.
Each participant will make 10 independent and incentivized hiring decisions. In each round, the participant will observe a pair of applicant profiles drawn from a previously conducted job-seeker experiment. The profiles contain information on applicants’ gender, age, high-school academic track, intermediate-task performance, and AI-use condition. Participants will predict the advanced-task performance of both applicants and select one applicant to hire.
To construct employers’ beliefs and hiring decisions, applicant profiles may differ along one or more observable characteristics. The order of the 10 hiring rounds will be randomized, and applicant pairs will not be repeated for the same participant.
Hiring decisions are incentivized. One of the 10 rounds will be randomly selected for payment, and the participant’s bonus will depend on the advanced-task performance of the applicant selected in that round. After completing the hiring decisions, participants will answer a post-experiment questionnaire measuring their beliefs about the performance of male and female applicants under AI and no-AI conditions, as well as their knowledge of and attitudes toward generative AI.