Machine Learning Reconfigures Science From Inference to Prediction

Machine learning has fundamentally reoriented scientific practice from inference toward prediction, according to a new analysis of 4.9 million publications and 255 ML techniques by Malena Méndez Isla et al. The study identifies a core-periphery structure with physical sciences at the methodological core and health sciences as the largest adopters, and documents two waves of displacement: 2015-2021 driven by deep learning and post-2022 driven by reliance on external commercial models.

For AI and data-science practitioners, the paper highlights shifting trade-offs between predictive performance, interpretability, and reproducibility that affect model selection, evaluation, and research design. According to the arXiv preprint by Malena Méndez Isla et al., the authors analyze 4.9 million publications and 255 ML techniques to map how machine learning reshaped scientific practice across disciplines. The paper reports a core-periphery semantic structure with the physical sciences at the methodological core and health sciences as the largest adopters. It finds predictive techniques concentrated in computer science while inferential approaches remain distributed across applied fields, and documents two waves of displacement: 2015-2021 driven by deep learning and post-2022 driven by reliance on external commercial models, per the arXiv preprint.