Cao Bo, Greiner Russell, Greenshaw Andrew, Sui Jie
Department of Psychiatry, University of Alberta, 4-142A Katz Group Centre for Research, 11315 - 87 Ave NW, Edmonton, AB, T6G 2B7, Canada, 1 7804929576.
Department of Computing Science, Faculty of Science, University of Alberta, Edmonton, AB, Canada.
J Med Internet Res. 2025 Aug 18;27:e66100. doi: 10.2196/66100.
Recent applications of artificial intelligence (AI) and machine learning in medicine, psychology, and social sciences have led to common terminological confusions. In this paper, we review emerging evidence from systematic reviews documenting widespread misuse of key terms, particularly "prediction" being applied to studies merely demonstrating association or retrospective analysis. We clarify when "prediction" should be used and recommend using "prospective prediction" for future prediction; explain validation procedures essential for model generalizability; discuss overfitting and generalization in machine learning and traditional regression methods; clarify relationships between features, independent variables, predictors, risk factors, and causal factors; and clarify the hierarchical relationship between AI, machine learning, deep learning, large language models, and generative AI. We provide evidence-based recommendations for terminology use that can facilitate clearer communication among researchers from different disciplines and between the research community and the public, ultimately advancing the rigorous application of AI in medicine, psychology, and social sciences.
人工智能(AI)和机器学习在医学、心理学及社会科学领域的最新应用引发了常见的术语混淆。在本文中,我们回顾了系统评价中的新证据,这些证据记录了关键术语的广泛误用,尤其是“预测”被用于仅证明关联或回顾性分析的研究。我们阐明了何时应使用“预测”,并建议对未来预测使用“前瞻性预测”;解释了模型可推广性所必需的验证程序;讨论了机器学习和传统回归方法中的过拟合和泛化;阐明了特征、自变量、预测因子、风险因素和因果因素之间的关系;并阐明了AI、机器学习、深度学习、大语言模型和生成式AI之间的层次关系。我们为术语使用提供基于证据的建议,以促进不同学科的研究人员之间以及研究界与公众之间更清晰的交流,最终推动AI在医学、心理学和社会科学中的严谨应用。