Citation: Han, H. (2026). Persuasion in AI-driven science: A case of drug design. Studies in History and Philosophy of Science, 119. https://doi.org/10.1016/j.shpsa.2026.102219.

Abstract: In AI-driven science, there is an empirical asymmetry between AI researchers and experimenters, stemming from their different capabilities to control material resources and generate experimental data. This paper develops an “activity-centric approach to persuasion” by focusing on the case of AI-driven drug design to address the asymmetrical relationship between AI researchers and experimenters. First, AI researchers can emphasise several values to persuade experimenters to conduct epistemic activities for pursuing specific AI models. Second, “annotating” the outputs of AI models effectively facilitates the activities of experimenters. Finally, persuasion in AI-driven science proceeds through a reciprocal iteration of the activities between AI researchers and experimenters, which requires an analysis of the pursuit of an AI model at the organisational level. This activity-centric approach to persuasion would provide a useful framework for analysing interdisciplinary research projects, especially those involving asymmetrical relationships.

Citation: Emanuele R., & L. Zuchowski. (2026). What can Artificial Intelligence learn from medicine? Generative analogies and reliable machine learning systems. Studies in History and Philosophy of Science, 119. https://doi.org/10.1016/j.shpsa.2026.102202.

Abstract: In the past few years, machine learning (ML) has been widely (and to an extent, successfully) implemented in medicine. However, uncertainties surrounding ML have made it difficult to establish the bases of its epistemic and methodological warrants. In the literature, a parallel has been drawn between medicine and ML, suggesting that we should model epistemic and methodological standards for ML on the standards of clinical translation. By developing tools from Hesse’s work, we characterize the nature of this parallel as a generative analogy between the process of clinical translation and the process of building ML systems. We identify more precisely the epistemic and methodological warrants of clinical translation that are typically only mentioned when appealing to the analogy, and we show in which sense such warrants apply analogically to the context of ML. In particular, we interpret warrants of clinical translation in reliabilist terms, and we show how this can inform a new form of ML reliabilism, which is distinct from (though compatible with) existing reliabilist accounts in philosophy of AI.

Citation: Cañas, D. (2026). Three epistemological desiderata for HPS practitioners. Studies in History and Philosophy of Science, 119. https://doi.org/10.1016/j.shpsa.2026.102201.

Abstract: I outline an epistemology of evidential reasoning in the history and philosophy of science (HPS). Drawing upon some concrete works in HPS scholarship, I formulate three novel epistemological desiderata for using historical case studies as evidence for philosophical claims about science, to wit: independent historical evidence, metahistorical criticism, and disciplinary alignment. These desiderata bring out epistemic qualities of evidential reasoning that contribute to achieving its primary goal of conferring justification upon philosophical conclusions on the basis of historical evidence. In this way, my proposed epistemology tackles the methodological problem of vicious circularity and the metaphysical problem of disciplinary unsuitability that allegedly jeopardise evidential reasoning, thereby vindicating its positive epistemic status.

Citation: Thorstad, D. (2025). The complexity–coherence trade-off in cognition. Mind, 134(534), 422–457. https://doi.org/10.1093/mind/fzaf015.

Abstract: I present evidence for a systematic complexity–coherence trade-off in cognition. I show how feasible strategies for increasing cognitive complexity along three dimensions come at the expense of a heightened vulnerability to incoherence. I discuss two normative implications of the complexity–coherence trade-off: a novel challenge to coherence-based theories of bounded rationality and a new strategy for vindicating the rationality of seemingly irrational cognitions. I also discuss how the complexity–coherence trade-off sharpens recent descriptive challenges to dual process theories of cognition.

Citation: Sung, D. (2026). On epistemic risks in research method validation. Studies in History and Philosophy of Science, 119. https://doi.org/10.1016/j.shpsa.2026.102206.

Abstract: Justin Biddle and Quill Kukla’s taxonomy of epistemic risk maps the theoretical terrain in ways that can assist researchers balance the cost of various types of errors. Based on a case study of method validation in toxicology, I argue that research method validation involves what I call “method mischaracterization risk.” I show that this risk fits into Biddle and Kukla’s taxonomy but is distinct from any previously identified risk (e.g., inductive risk). I also draw from the case study to illustrate that research method validation is vulnerable to a particular social-epistemic dilemma known as the Ibsen predicament. By combining insights drawn from pragmatist accounts of scientific methodology and Daniel Steel’s notion of epistemic constraint, I argue that method characterization risk can be adequately managed when research methods are construed as self-correcting systems of epistemic practice.

Citation: Ladyman, J., & Nefdt, R.M. (2026). Are AI language models scientific models of language?. European Journal for Philosophy of Science, 16. https://doi.org/10.1007/s13194-026-00770-9.

Abstract: There has been a recent surge of philosophical interest in AI large language models (LLMs) and their relevance for the understanding of human language. In this article, we take the perspective of the philosophy of science on modelling in general to consider whether LLMs are (good) models of language. We establish that LLMs can be considered scientific models, and that as such they are good scientific models of certain aspects of human language, in so far as they can capture the modal structure of natural language. We argue that they can do this even if they do not reveal (all) its causal structure, nomological structure or mechanisms.

Citation: Figdor, C. (Forthcoming). The 21st Century Mind-Body Problem. Philosophy of Science, 1–17. https://doi.org/10.1017/psa.2026.10274.

Abstract: This paper articulates how recent advances in comparative behavioral and biological research are changing the nature of the mind/body problem from a human mind/human body problem to a problem of conceptualizing and organizing diverse types of minds in a hierarchical relational structure that is non-contingently related to phylogeny. It also discusses the impact of this refinement on our assessments of modal claims about possible artificial (AI) consciousness.

Citation: Holman, B. An ethical obligation to ignore the unreliable. Synthese, 198 (suppl 23), 5825–5848. https://doi.org/10.1007/s11229-019-02483-4.

Abstract: Stephen John has recently suggested that the ethics of communication yields important insights as to how values should be incorporated into science. In particular, he examines cases of “wishful speaking” in which a scientific actor (e.g. a tobacco company) endorses unreliable conclusions in order to obtain the consequences of the listener treating the results as credible. He concludes that what is wrong in these cases is that the speaker surreptitiously relies on values not accepted by the hearer, violating what he terms “the value-apt ideal”. I expand on this view by integrating into it Miranda Fricker’s account of testimonial injustice. I find testimonial injustice can arise in a manner unanticipated by Fricker, specifically that a credibility excess given to a speaker typically reduces the ability of others in the epistemic community to transmit knowledge, a phenomenon I term “collateral epistemic injustice”. I argue that this possibility entails that receivers have an ethical obligation to assign credibility judiciously. A further consequence of this view is that the value-apt ideal is insufficient. Because audience members have an obligation to assign credibility judiciously, a speaker cannot merely rely on shared values, they must also be open about the extent to which their conclusions depend on those values. Thus, both wishful speaking and obscuring the value-dependent nature of a conclusion makes one an unreliable source of information. Accordingly, other community members have an ethical obligation to ignore a speaker that frequently engages in either.

Citation: Wu, S. Can arguments save armchair philosophy?. Synthese, 208. https://doi.org/10.1007/s11229-026-05766-9.

Abstract: Experimental philosophers have argued that the diversity of intuitions elicited by thought experiments poses a significant challenge to armchair philosophy, given its traditional reliance on such intuitions. In response, some defenders of traditional philosophical methodology have proposed an “argument defense,” according to which philosophers rely primarily on arguments rather than intuitions when evaluating thought experiments. This paper critically examines this strategy. Drawing on existing experimental work—both direct studies examining the influence of arguments and indirect evidence from cross-cultural research on reasoning and argument evaluation—it argues that responses to arguments may also exhibit diversity, leaving open the possibility that such diversity is no less significant than that found in intuitions. Although the current evidence remains limited and inconclusive, it raises doubts about the assumption that arguments can straightforwardly vindicate armchair methodology. It also suggests the need to expand experimental philosophy beyond intuitions to include the empirical study of arguments.

Citation: Duede, E. Forthcoming. AI and the decentering of disciplinary creativity. Philosophy of Science, 1—12. https://doi.org/10.1017/psa.2026.10262.

Abstract: This paper examines the role of artificial intelligence in scientific problem-solving, with a focus on its implications for disciplinary creativity. Drawing on recent work in the philosophy of creativity, I distinguish between creative approaches and creative products, and introduce the concept of disciplinary creativity—the creative application of discipline-specific expertise to a valued problem within that field. Through two cases in mathematics, I show that while computation can extend disciplinary creativity, certain approaches involving AI can serve to displace it. This displacement has the potential to alter (and, perhaps, diminish) the value of scientific pursuit.