Artificial Intelligence and the Scientific Method: A Theoretical Examination of Discovery, Explanation and Reproducibility in AI-Driven Science
DOI:
https://doi.org/10.68854/rccqng25Keywords:
AI for science, scientific discovery, explanation, reproducibility, philosophy of science, machine learning, research integrity, automated science, epistemology of scienceAbstract
Artificial intelligence (AI) is changing the conduct of science. Models predict protein structures and propose materials, generate hypotheses, design and execute experiments, and increasingly assist in writing the papers that report the results. This paper examines the effect of these developments on the scientific method, drawing on the philosophy of science, the research-integrity and reproducibility literatures, and recent work on AI-driven discovery across the natural, engineering and social sciences. It distinguishes three functions of science that AI affects differently: discovery, the identification of what is the case; explanation, the account of why it is the case; and reproducibility, the possibility of independent confirmation. The paper argues that AI substantially accelerates discovery, threatens explanation, because predictive success without mechanistic understanding is now attainable at scale, and complicates reproducibility, since AI-driven results depend on models, data and computing conditions that are seldom reported in full or readily reproduced. It sets out a corresponding division of epistemic labour between human scientists and AI systems and a set of reporting and verification standards for AI-assisted research, and it considers the particular implications for scientific institutions in Africa, which may gain access to discovery capacity previously beyond their means while facing acute risks of irreproducibility and of dependence upon others. It argues that these three effects pull in different directions rather than reinforcing one another, so that the governance of AI-assisted research cannot treat acceleration, understanding and confirmation as a single problem. The paper sets out propositions on where predictive models may stand in for mechanistic explanation and where they may not, and proposes that African scientific institutions adopt explicit standards for the reporting and verification of AI-assisted results as a condition of building research capacity that is independent rather than dependent.
Downloads
References
Ahmed, W., Kommineni, V. K., König-Ries, B., Gaikwad, J., Gadelha, L., & Samuel, S. (2025). Evaluating the method reproducibility of deep learning models in biodiversity research. PeerJ Computer Science, 11, e2618. https://doi.org/10.7717/peerj-cs.2618
Bajorath, J. (2025). From explainable artificial intelligence to human understanding. Artificial Intelligence in the Life Sciences, 7, 100131. https://doi.org/10.1016/j.ailsci.2025.100131
Boge, F., & Mosig, A. (2025). Causality and scientific explanation of artificial intelligence systems in biomedicine. Pflügers Archiv – European Journal of Physiology, 477(4), 543–554. https://doi.org/10.1007/s00424-024-03033-9
Crockett, M. J., Bai, X., Kapoor, S., Messeri, L., & Narayanan, A. (2023). The limitations of machine learning models for predicting scientific replicability. Proceedings of the National Academy of Sciences, 120(33), e2307596120. https://doi.org/10.1073/pnas.2307596120
Crockett, M. J., & Messeri, L. (2026). AI surrogates and illusions of generalizability in cognitive science. Trends in Cognitive Sciences, 30(3), 203–215. https://doi.org/10.1016/j.tics.2025.09.012
Day, T. (2023). A preliminary investigation of fake peer-reviewed citations and references generated by ChatGPT. The Professional Geographer, 75(6), 1024–1027. https://doi.org/10.1080/00330124.2023.2190373
Del Giudice, M. (2024). The prediction-explanation fallacy: A pervasive problem in scientific applications of machine learning. Methodology, 20(1), 22–46. https://doi.org/10.5964/meth.11235
Del Pup, F., & Atzori, M. (2024). Toward improving reproducibility in neuroimaging deep learning studies. Frontiers in Neuroscience, 18, 1509358. https://doi.org/10.3389/fnins.2024.1509358
Duede, E. (2023). Deep learning opacity in scientific discovery. Philosophy of Science, 90(5), 1089–1099. https://doi.org/10.1017/psa.2023.8
Freiesleben, T., & Zezulka, S. (2026). The benchmarking epistemology: Validity theory for evaluating machine learning models. Philosophy of Science. Advance online publication. https://doi.org/10.1017/psa.2026.10280
Ikeuchi, R., & Imai, T. (2026). An empirical assessment of data sharing and computational reproducibility in sociology. Royal Society Open Science, 13(9), 260949. https://doi.org/10.1098/rsos.260949 Jedlička, P. O. (2026). Quo vadis? Scientific discovery in the age of artificial intelligence. Teorie vědy / Theory of Science, 48(1), 9–44. https://doi.org/10.46938/tv.2026.728
Kapoor, S., & Narayanan, A. (2023). Leakage and the reproducibility crisis in machine-learning-based science. Patterns, 4(9), 100804. https://doi.org/10.1016/j.patter.2023.100804
Kondo, T. S., & Diwani, S. A. (2023). Artificial intelligence in Africa: A bibliometric analysis from 2013 to 2022. Discover Artificial Intelligence, 3(1), 34. https://doi.org/10.1007/s44163-023-00084-2
Li, K., Rubungo, A. N., Lei, X., Persaud, D., Choudhary, K., DeCost, B., Dieng, A. B., & Hattrick-Simpers, J. (2025). Probing out-of-distribution generalization in machine learning for materials. Communications Materials, 6(1), 9. https://doi.org/10.1038/s43246-024-00731-w
Mehta, N., Felder, R. M., Kodish, E., Imrey, P. B., & Frankel, R. M. (2026). A call for clarity: A unified checklist for reporting use of large language models in writing scientific manuscripts. Research Integrity and Peer Review, 11(1), 24. https://doi.org/10.1186/s41073-026-00212-3
Messeri, L., & Crockett, M. J. (2024). Artificial intelligence and illusions of understanding in scientific research. Nature, 627(8002), 49–58. https://doi.org/10.1038/s41586-024-07146-0
Muhammad, Y. U., Abubakar, I., & Jiya, E. A. (2026). Data sovereignty and the leaky pipe: Charting a course for data governance and information security in Nigeria and Africa. Journal of Basics and Applied Sciences Research, 4(2), 224–232. https://doi.org/10.4314/jobasr.v4i2.23
Picazo-Sanchez, P., & Ortiz-Martin, L. (2026). Evaluating the integrity of LLM-generated citations: Prevalence and risks of fabricated references in scientific literature. Data, 11(5), 122. https://doi.org/10.3390/data11050122
Schoch, D., Chan, C., Wagner, C., & Bleier, A. (2024). Computational reproducibility in computational social science. EPJ Data Science, 13(1), 75. https://doi.org/10.1140/epjds/s13688-024-00514-w
Selan, J. (2026). (De)generative AI and research integrity. Humanities and Social Sciences Communications. Advance online publication. https://doi.org/10.1057/s41599-026-08248-y Søgaard, A. (2023). On the opacity of deep neural networks. Canadian Journal of Philosophy, 53(3), 224–239. https://doi.org/10.1017/can.2024.1
Tobias, A. V., & Wahab, A. (2025). Autonomous 'self-driving' laboratories: A review of technology and policy implications. Royal Society Open Science, 12(7), 250646. https://doi.org/10.1098/rsos.250646
Tong, S., Mao, K., Huang, Z., Zhao, Y., & Peng, K. (2024). Automating psychological hypothesis generation with AI: When large language models meet causal graph. Humanities and Social Sciences Communications, 11(1), 896. https://doi.org/10.1057/s41599-024-03407-5
Xu, P., Ji, X., Li, M., & Lu, W. (2023). Small data machine learning in materials science. npj Computational Materials, 9(1), 42. https://doi.org/10.1038/s41524-023-01000-z
Yu, B. (2024). After computational reproducibility: Scientific reproducibility and trustworthy AI. Harvard Data Science Review, 6(1). https://doi.org/10.1162/99608f92.ea5e6f9a
Downloads
Published
Issue
Section
License
Copyright (c) 2026 IIAR Journal of Artificial Intelligence and Emerging Technologies
This work is licensed under a Creative Commons Attribution 4.0 International License.
