Artificial intelligence (AI) and machine learning (ML) are already changing how organizations make employee selection decisions (EEOC, 2016).
AI/ML-based approaches can help organizational decision-making by leveraging computational algorithms to efficiently collect and synthesize large volumes of applicant data (e.g., Das et al., 2018).
However, despite the growing interest in semi- or fully automating hiring processes with AI/ML there is a degree of apprehension toward AI/ML-based decision-making approaches with concerns about fairness, transparency, privacy, and insufficient human interaction (Gonzalez et al., 2019).
As such we have an ongoing research program to understand how candidates react to AI in the hiring process. Our research has been focused around two strategies to improve reactions toward AI/ML-driven hiring processes. First, the concern about the lack of human interaction might be improved if AI/ML algorithms augmented human decision-making, rather than being the sole decision-making source. Second, by grounding AI/ML algorithms in well-established psychological theories, we hypothesized that predictions could be understood in terms of relations between psychological constructs and make them easier for organizations to understand, explain, and (importantly) defend. If these two strategies could be adopted in the development of AI/ML-based talent decision-making processes, would people react more favorably toward such tools?
To date, our research program has consisted of three studies, all of which have either been published in scientific research outlets or accepted for presentation at research conferences. In all three studies, we manipulated different variables in written hiring scenarios and measured candidate reactions such as perceived fairness and transparency, company culture, and how respondents would behave in the scenario.
Here is a brief summary of the three studies:
- Study 1 - Participants imagined applying for a job where the hiring decision would be made by either a human or an AI. We also manipulated whether participants imagined being hired or turned down for the job by that decision-maker.
- Study 2 – Participants compared a strictly human-based hiring process, where a recruiter or a hiring manager would be involved in the entire process, to four other types of AI-based hiring processes. In the AI-based processes we crossed (a) whether the AI was developed using psychological theory (i.e., theory-driven) versus a strictly data-driven approach like data mining, and (b) whether the process was managed by just an AI or by a human plus an AI, which we call an augmented process. Participants read all five scenarios and reported their reactions to each hiring process.
- Study 3 - Participants imagined going through a hiring process where the process would be overseen either solely by a hiring manager, solely by an AI, or by both. We also manipulated whether participants imagined themselves in the screening stage or the final stage of the selection process.
The major highlights across these three studies are below:
- People prefer people. Participants favored a solely human-based decision-making approach, relative to AI-based approaches.
- Candidate experience affects the bottom line. Not only did negative reactions to AI affect intentions to accept job offers, our research also suggests these reactions may spill over beyond the hiring process. Compared to people who were hired by a human decision-maker, people who were hired with an AI-based approach reporting greater distrust of the company and less willingness to promote the company and its products.
- AI needs to be explainable. Data-driven AI/ML were viewed as less transparent than human decision-makers, whereas theory-driven AI/ML and human decision-making approaches were perceived as having similar transparency.
- AI should augment human decision making, not replace it. People reported greater interactional and procedural justice, perceived the process to have more of a personal touch, expressed greater trust in the types of data used for decision-making, and felt more comfortable when AI/ML-based approaches supplemented human decision-making, compared to when AI/ML was used alone.
- Familiarity begets acceptance. Candidates who are more familiar with AI react more favorably to it during the hiring process.
The research showed that while applicants may prefer human decision-makers over AI/ML-based approaches, this could be mitigated by (a) augmenting human decision-making with AI/ML, rather than replacing it entirely, and (b) by relying upon theory when designing AI/ML-based solutions, rather than opting for primarily data-driven approaches.
Going forward, we need to be seeing AI as an ally to our talent decision making processes.
Going forward, we need to view rigorously-developed AI as an ally to our talent decision making processes. Perhaps future research may explore strategies for building trust in AI/ML decision-making, maintaining data transparency and understanding how individual differences (e.g., gender) affect reactions to AI/ML.
Gonzalez, M. F., Capman, J. F., Oswald, F. L., Theys, E. R., Tomczak, D. L. (2019). “Where’s the I-O?” Artificial Intelligence and Machine Learning in Talent Management Systems. Personnel Assessment and Decisions, 5(3), 33-44.
About the AuthorFollow on Linkedin Visit Website More Content by Aon's Assessment Solutions