Racial discrimination in the online hiring task

In this study, we asked the returned participants (N=751) to read about an organization called IMB Consulting that had just completed its hiring process. The participants were shown 30 AI-generated photos and hypothetical resumes of applicants who varied across four dimensions: race, gender, GPA, and hometown. Participants viewed each candidate’s image and resume and were informed whether the candidate was hired or not. Each returned participant was randomly assigned to one of three treatments: 1) discrimination against Blacks, 2) discrimination against Whites, and 3) discrimination against Asians. In the third treatment, AI-generated photos of East, Southeast, and South Asians were randomized across participants. The treatments differ only in terms of the correlation between race and the likelihood of being hired. In all three treatments, we design the GPA (+.78) and being male (+0.2) to be positively correlated with the likelihood of being hired. In treatment 1, being Black has a negative correlation (-0.2) with the likelihood of being hired; in treatment 2, being White has a negative correlation (-0.2) with the likelihood of being hired, and the same is true for being Asian in treatment 3.

After completing the task, participants were asked to “please note anything that stood out to [them] about the hiring process.” The study coded whether participants naturally mentioned any inequality in the hiring process. We account for any potential incorrect mentions by participants of bias against the racial group that was favored in the treatment. Participants could receive a score of -1, 0, or +1. A score of -1 represents an incorrect mention of bias against the racial group that was, in fact, favored in their treatment. A 0 represents the participant not mentioning any bias, and +1 represents a correct mention of the bias against the racial group disfavored in their treatment. This score is named “relative naturalistic notice bias.”

We also elicited participants’ attention to inequality by directly asking participants to rate the level of racial and gender discrimination in the IMB Consulting’s hiring procedure, which we called “absolute bias judgment.” We further asked them five questions about their willingness to invite a third-party agency to investigate the company’s hiring process. We named these variables “the desire to investigate” (DTI)