Can Legal and Professional Personnel Selection Principles be Met With Machine Learning (Artificial Intelligence)?
基于人力资源管理文献，评估机器学习能否满足人员选拔的科学、专业和法律原则，发现其能基本满足且不增加法律风险，但大型语言模型可能带来作弊和文本伪造问题。
ABSTRACT The purpose of this article is primarily to evaluate whether machine learning (a form of artificial intelligence) can meet scientific, professional, and legal principles of personnel selection based on the rapidly accumulating research literature in Human Resource Management (HRM). It does so by addressing a series of questions in terms of what is known in the current research literature on these principles and related topics and by proposing a research agenda for what else needs to be studied. The review shows that there is enough current scientific evidence on the value of ML to support its use. ML tools are able to meet the basic principles of personnel selection and better meet many other very important principles. In addition, ML for personnel selection decisions is not a “black box” and can be understood and explained. It does not increase the legal risks from hiring, although it may require some additional steps in a few jurisdictions. There are uses of ML in selection that most organizations should be considering. Customized procedures will require good data, but generic vendor products may also serve some needs. Additional expertise may be required of HRM professionals, but not necessarily at a high level that would require new staffing or consultant expense. The availability of large language models (LLMs) may render unproctored remote assessments vulnerable to cheating, and all narrative candidate information provided in the future susceptible to AI‐generated text, resulting in more questions than research answers at this stage. The article ends with practical and theoretical implications for the use of ML in selection.