Mapping the evolution of algorithmic HRM (AHRM): a multidisciplinary synthesis

The Fourth Industrial Revolution's digital transformation is profoundly impacting HRM, leading to an increased use of algorithms in HR strategy and practice. This "Era of Algorithmic HRM (AHRM)" offers increased efficiency and productivity but necessitates changes in organizational systems and upskilling strategies. This study addresses two objectives: first, to clarify the concept of AHRM; and second, to map its evolution from 2000 to 2022. The study builds upon existing literature highlighting the link between digital transformation, strategic HRM (SHRM), and the resource-based view (RBV), emphasizing the shift towards more sophisticated AHRM technologies. Previous research often conflates terms like digital HRM, e-HRM, and HR analytics, lacking construct clarity. This study uses a multidisciplinary synthesis to define AHRM and then employs a science mapping review to chart its evolution over time, addressing the research question: What are the dominant, emerging, and state-of-the-art themes in published research from 2000 to 2022 that examine organizational tools for algorithmic HRM (AHRM)?

The paper reviews ten previous studies analyzing algorithmic applications in HRM, highlighting their diverse methodologies, focuses (AI, machine learning, blockchain, big data), and time ranges. These reviews demonstrate the growing interest in AI and related technologies within HRM. Four articles broadly explore AI in HRM, covering various HR functions and efficiency improvements. Two focus on blockchain in HR, examining its implications for employee-system interactions and fraud detection. Five others concentrate on specific HR functions (recruitment, performance evaluation, remuneration, training & development, smart technologies). The reviewed studies collectively show increasing researcher interest in AI and its applications to HRM systems, functions, and strategies.

The study uses a theory-led research method informed by a General Framework for Algorithmic Decision-Making, synthesizing Meijerink et al.'s (2021) AHRM definition with Herrmann's (2022) AI taxonomy. This framework distinguishes between automated (prescriptive analytics, robotic process automation, metaheuristics) and augmented (descriptive, diagnostic, predictive analytics using machine learning, deep learning, and big data) decision-making. A scientometric science mapping approach (using SciMAT Version 1.1.05) analyzes bibliometric data from the Scopus database. The Scopus search, detailed in the paper, targeted peer-reviewed journal articles, books, chapters, conference papers, and proceedings published between 2000 and 2022, focusing on English-language publications related to AHRM. The search yielded 1672 publications after removing duplicates and false positives. The analysis included co-occurrence of keywords to map the evolution of AHRM themes across four periods.

The science mapping reveals four key findings. First, considerable confusion persists regarding terminology related to AI and its applications in HRM. Second, the majority of relevant publications appear in computer science and engineering journals, indicating limited engagement from HRM-focused researchers. Third, most research focuses on non-strategic HR functions (scheduling, recruitment, learning & development), with less attention to strategically integrating AHRM into broader human capital strategies. Fourth, recruitment is the most automated AHRM application, while others (workforce planning, allocation & scheduling, learning & development) predominantly involve augmented decision-making. These findings are summarized in Table 2 and Fig. 9, which presents a General Framework for Algorithmic HRM Tools, integrating data-driven and theory-driven perspectives. The study also reveals an increase in AHRM publications since 2006, with a significant surge in 2020, driven by applications in HR scheduling, allocation, and recruitment. The literature shows most applications of algorithmic tools occur within a micro perspective, focusing on individual HRM functions. However, some literature indicates a movement towards multi-objective HRM applications. Fig. 5 illustrates an evolutionary map of AHRM themes across different periods, showing shifts in keyword co-occurrences and the emergence of new themes.

The findings highlight the need for a more integrated and multidisciplinary approach to AHRM research and practice. The lack of HRM-focused research suggests a need for greater awareness and engagement from HRM scholars. The focus on non-strategic functions underscores the need to explore AHRM's potential for maximizing human capital strategies. The distinction between automated and augmented decision-making in various HRM functions provides valuable insights for practitioners. The study's frameworks can help to clarify terminology and guide the development of more effective and ethically sound AHRM applications. The study suggests that increased collaboration between HRM experts and technology specialists is essential for leveraging the full potential of AHRM. This includes addressing ethical considerations related to fairness, bias, privacy, and transparency.

This study provides a comprehensive overview of AHRM, offering a refined definition and a detailed evolutionary map. It highlights the multidisciplinary nature of the field, the current focus on specialized applications, and the need for greater emphasis on strategic HRM integration. Future research should focus on bridging the gap between technical and HRM perspectives, exploring more integrative AHRM applications, and addressing ethical considerations. The emergence of generative AI necessitates further exploration of its potential benefits and risks within the context of AHRM.

The study's reliance on the Scopus database, excluding Google Scholar, may limit the comprehensiveness of the literature review. The use of a review methodology means that no primary data was collected. While the study covers the period from 2000 to 2022, the rapid pace of innovation in AHRM, particularly the recent emergence of generative AI, warrants further investigation into the implications of these developments for the field.