Navigating Automated Hiring: Perceptions, Strategy Use, and Outcomes Among Young Job Seekers

TL;DR

The paper examines how young computer science job seekers perceive automated employment decision tools (AEDTs) and how these perceptions relate to their strategies and early job outcomes. Using a two-part survey of 448 CS students across three universities, the authors quantify procedural fairness and willingness to be evaluated by varying levels of automation and evaluation types, and link these to awareness, strategies, and outcomes. They find that AI-based hiring is perceived as less fair, with greater reductions in fairness for less technical tasks, though a human-in-the-loop mitigates these concerns; referrals and higher family income emerge as the strongest predictors of job offers, underscoring persistent socioeconomic inequities. The findings have implications for auditing, policy, and worker-centered design of AEDTs, suggesting reforms that ensure transparency, accountability, and strategies that do not rely on privilege to achieve hiring success.

Abstract

As the use of automated employment decision tools (AEDTs) has rapidly increased in hiring contexts, especially for computing jobs, there is still limited work on applicants' perceptions of these emerging tools and their experiences navigating them. To investigate, we conducted a survey with 448 computer science students (young, current technology job-seekers) about perceptions of the procedural fairness of AEDTs, their willingness to be evaluated by different AEDTs, the strategies they use relating to automation in the hiring process, and their job seeking success. We find that young job seekers' procedural fairness perceptions of and willingness to be evaluated by AEDTs varied with the level of automation involved in the AEDT, the technical nature of the task being evaluated, and their own use of strategies, such as job referrals. Examining the relationship of their strategies with job outcomes, notably, we find that referrals and family household income have significant and positive impacts on hiring success, while more egalitarian strategies (using free online coding assessment practice or adding keywords to resumes) did not. Overall, our work speaks to young job seekers' distrust of automation in hiring contexts, as well as the continued role of social and socioeconomic privilege in job seeking, despite the use of AEDTs that promise to make hiring "unbiased."

Figures (5)

• Figure 1: Participant self-reported household income from $0 to $1,500,000, with a median of $100,000.
• Figure 2: Participants' procedural fairness and willingness perceptions for different evaluation types and levels of automation were rated on a Likert scale (1: process seems very unfair / unwilling to be evaluated this way --- 5: seems very fair / willing to be evaluated this way). Notably, both measures performed similarly with participants. Responses reflected that some human involvement was always preferred, and AI-only review was least-preferred across evaluation types. However, there was a greater tolerance for AI involvement in more technical evaluations (i.e., in coding assessments compared to interviews). Bars represent 95% confidence intervals.
• Figure 3: Participants reported applying to 0 to 600 jobs, with 50% applying to more than 20.
• Figure 4: Participants used referrals in anywhere from 0% to 100% of applications, with over half of participants having at least one referral.
• Figure 5: Participants reporting higher annual household incomes were also more likely to receive one or more job offers.