IRGA 2021 IAB@R

The topic of recruitment aided by Artificial Intelligence (AI, hereinafter), also known as “predictive recruitment,” falls within the broader field of analytical tools used in human resources management (HR analytics), whose advantages and limitations are increasingly being discussed (Angrave et al., 2016). This is not yet a well-established field of research, and there is a significant gap between the role these tools play in corporate practices and the research published in academic journals (Marler & Boudreau, 2017; Woods et al., 2019). In practice, every stage of HR processes is affected by the introduction of AI-based algorithms, and solutions are growing exponentially. According to data collected in France, 50% of HR professionals were using at least one tool incorporating AI algorithms in 2019 (e.g., automated applicant tracking software). The predictive recruitment market is growing rapidly, driven by specialized startups (Esayrecrue, Kudoz, Goshaba, etc.). The main arguments put forward by proponents of predictive recruitment tools are that they enable a faster, more efficient, and more inclusive selection process (free of discriminatory biases), due to their “objectivity” and their ability to eliminate interpersonal human judgment biases linked, for example, to the activation of stereotypes. However, it must be acknowledged that studies on the validity and impacts of AI tools applied to recruitment remain very rare. This is despite the fact that the stakes are high and particularly sensitive, especially in the context of promoting diversity and combating discrimination. This fight is not only an ethical imperative but also a legal obligation, the parameters of which are specified in the Labor Code. The law specifies that recruiters are required to use objective and relevant selection methods and must be able to justify their choices if necessary.

The IAB@R project aims to partially bridge this gap between the promises made and the lack of information regarding the effectiveness, relevance, and objectivity of predictive recruitment solutions. This objective is part of a broader examination of the impact of algorithmic recommendations on decision-making. In the field of management science, predictive recruitment algorithms are a prime example of a “management tool” (Vaujany, 2006), whose deployment and adoption influence the attitudes and behaviors of organizational actors, almost regardless of their effectiveness. In this context, we aim to examine both the psychological aspect (users’ perceived effectiveness) and the instrumental aspect (the tool’s ability to make selections free of discriminatory bias). Specifically, the project aims to address two major questions.

1. How do users react when they encounter predictive recruitment solutions?
2. To what extent are predictive recruitment solutions technically capable of reducing discriminatory biases?

The IAB@R project is a cross-disciplinary initiative: it draws on research questions from the fields of management science (recruitment management and tools), organizational psychology (user responses to algorithmic recommendations), and statistics (the effectiveness of predictive models in personnel selection). An initial exploratory study conducted in 2019 by the project leader used a cross-disciplinary literature review to provide an overview of biases associated with algorithmic recruitment (Lacroux & Martin-Lacroux, forthcoming). This study made it possible to identify, among the many research questions, those whose economic and social implications appear to be the most significant, namely the problem of “black boxes,” that is, issues related to the explainability of algorithmic decisions (Villani et al., 2018) and the question of the adoption of algorithmic decision-support solutions in the field of human resources management.