Abstract

     Artificial intelligence (AI) has the potential to change the future of human resource management (HRM). Scholars from different disciplines have contributed to the field of AI in HRM but with rather insufficient cross-fertilization, thus leading to a fragmented body of knowledge. In response, we conducted a systematic, interdisciplinary review of 184 articles to provide a comprehensive overview. We grouped prior research into four categories based on discipline: management and economics, computer science, engineering and operations, and others. The findings reveal that studies in different disciplines had different research foci and utilized different methods. While studies in the technical disciplines tended to focus on the development of AI for specific HRM functions, studies from the other disciplines tended to focus on the consequences of AI on HRM, jobs, and labor markets. Most studies in all categories were relatively weak in theoretical development. We therefore offer recommendations for interdisciplinary collaborations, propose a unified definition of AI, and provide implications for research and practice.

Introduction

     Artificial intelligence (AI) is among the most influential technologies changing the labor market (e.g., Huang & Rust, 2018). On the one hand, AI can have negative consequences, such as eliminating over 45% of all jobs (Berg, Buffie, & Zanna, 2018) and increasing social inequality (e.g., Levy, 2018). On the other hand, it may also provide benefits, such as upgrading or augmenting jobs instead of replacing them (e.g., Autor, 2015). Taken together, it is fair to say that AI will have a significant impact on the future of human resource management (HRM), and the application of AI in HRM has great potential (Malik, Budhwar, Patel, & Srikanth, 2020; Malik, De Silva, Budhwar, & Srikanth, 2021).

     AI-HRM is a topic beyond the field of HRM because of its interdisciplinary nature, i.e., the development of AI-based HR tools depends on progress in technical fields, while implementations of such AI tools and consequences of AI implementations rely on knowledge from social science. Scholars from various disciplines have contributed to AI–HRM knowledge. For example, computer science (CS) scholars developed AI algorithms to solve HRM problems (e.g., Anandarajan, 2002). Economists discussed AI’s impacts on labor markets (e.g., Berg et al., 2018). Psychologists found that AI usage did not demotivate job candidates during recruitment (Van Esch, Black, & Ferolie, 2019) but might induce higher employee turnover (e.g., Brougham & Haar, 2020). Medical scholars revealed that medical employees were not ready for AI usage (e.g., Abdullah & Fakieh, 2020). Although substantial research exists on AI-HRM topics in various disciplines, each discipline approached the topic from a different perspective, paying little attention to synthesizing interdisciplinary knowledge. This is unfortunate because interdisciplinary knowledge and collaboration are particularly important for successful AI implementation (Fountaine, McCarthy, & Saleh, 2019) and talent development in the AI era (Pejic-Bach, Bertoncel, Meško, & Krstić, ۲۰۲۰). In response to this gap, a comprehensive interdisciplinary review can help synthesize the rather scattered knowledge and encourage disciplinary cross-fertilization by reducing misunderstanding and avoiding “reinventing the wheel.”

Discussion

     The review suggests that the field of AI-HRM is still in its infancy, despite its rapid growth in recent years. The field is rather fragmented, with studies from different disciplines covering a wide variety of topics. While CS and EO papers focused more on developing AI tools to facilitate HRM, ME and OT papers were more interested in general issues related to AI usage, particularly in topics related to AI-induced job loss or job changes. Our critical evaluation reveals that scholars need to pay attention to theoretical and methodological rigor. In terms of theories, all disciplines were rather weak in theoretical developments, indicating that the current field is perspective- and practice-oriented. The evaluation of methods identified data validity was the most notable methodological pitfall. We also call for a more rigorous and standardized approach for new AI methods. Based on the literature review, we discuss key issues, propose specific recommendations for future research, and elaborate on theoretical and managerial implications in the following.