Theoretical Philosophy

My research in theoretical philosophy centres on six connected themes:

1. The origins of our core cognitive concepts—truth, knowledge, and understanding—and what they do for us.
2. The relation between reasons and causes, both in epistemology (justification vs. causal explanation) and in the philosophy of action (reconciling voluntary agency and the causal determination of the will).
3. The rationality of conceptual change and conceptual engineering and their theoretical underpinning by different accounts of concepts and meaning.
4. The systematicity of thought and the limits of theoretical virtues such as precision, depth, consistency, and coherence.
5. The epistemology of artificial cognition, especially how best to conceptualize the emergent capacities of large language models in light of 1–4.
6. Philosophy’s relation to the human sciences and its nature as a humanistic discipline.

My first monograph, The Practical Origins of Ideas: Genealogy as Conceptual Reverse-Engineering (OUP 2021), develops a method of “conceptual reverse-engineering” to uncover the practical challenges at the root of our most abstract and elusive concepts. Through lenses from Nietzsche and Bernard Williams, the book explores the genealogy of the notion of truth and how norms of accuracy and sincerity evolved with our capacity for belief and assertion to become defining virtues of thought and communication. The book also reconstructs, through lenses from E.J. Craig and Miranda Fricker, the genealogy of knowledge, testimony, and epistemic injustice, and reconciles it with knowledge-first epistemology (see also Genealogy and Knowledge-First Epistemology: A Mismatch?). A series of more recent papers on understanding (Why We Care About Understanding: Competence through Predictive Compression) complements this earlier work on the genealogy of the core mindware structuring thought and talk.

This interest in genealogical explanation grew out of my work on Wittgenstein, who sharply distinguished reasons from causes, and insisted that causes merely explain while only reasons justify. This divide not only creates a problem over how to reconcile the determination of our will by reasons with the determination of our bodies by causes, but also bars genealogies—as causal explanations—from having any effect in the space of reasons. I engaged with this divide in a string of papers (Two Orders of Things; Wittgenstein on the Chain of Reasons; Davidsonian Causalism and Wittgensteinian Anti-Causalism) before finding a way to accommodate the normative significance of causal explanation (How Genealogies Can Affect the Space of Reasons; From Paradigm-Based Explanation to Pragmatic Genealogy). This cleared a path from the explanation to the evaluation and engineering of concepts.

During my three years at Oxford, I then wrote my Habilitation, Reasons for Reasons: A Theory of Reasons for Concept Use, which later appeared under the title: The Ethics of Conceptualization: Tailoring Thought and Language to Need (OUP 2025). Building on an earlier article (Function-Based Conceptual Engineering and the Authority Problem), the book intervenes in debates on conceptual engineering by establishing a framework for the rationality of conceptual change. It develops an epistemology of second-order reasons for concept use, which provide reasons to frame first-order reasons for belief and action in certain terms rather than others. The book reexamines the view that concepts must always aspire to theoretical virtues like precision and consistency, arguing that in certain contexts, apparent vices such as vagueness or superficiality actually serve us better. A central case study is the concept of voluntariness, which reconciles reasons and causes precisely by remaining superficial.

I then led a four-year SNSF Ambizione project exploring the rationales for cognitive systematization and reappraising the systematicity debate between classical and connectionist theories of mind (Explainability through Systematicity). I showed that transformer-based connectionist models now rise to Fodor’s “systematicity challenge,” but still struggle with more demanding forms of systematicity and with the asystematicity of the truth in certain domains (Can AI Rely on the Systematicity of Truth?).

A unifying thread running through my work is the relevance of the human sciences to philosophy. My first monograph theorized the necessity of this engagement, and I have since put this into practice by drawing on anthropology and ethnography in my second book and by co-editing a volume on the relationship between philosophy and history, entitled: Bernard Williams on Philosophy and History (OUP 2025).

With the five-year SNSF Starting Grant I was recently awarded, I am reconnecting with my academic background in cognitive and computational linguistics through a project on artificial cognition. By working through three foundational case studies—concept formation, understanding, and reasoning—the project aims to decouple the indispensable functional roles of these concepts from their anthropocentric assumptions. This requires bridging the gap between philosophy and the emerging field of mechanistic interpretability. For an example of this, see Mechanistic Indicators of Understanding in Large Language Models.