Research

My research broadly aims to understand—and sometimes improve—how social scientists and policymakers reason from evidence. I am particularly interested in the role of similarity and analogy in inference and decision-making. 

Peer-Reviewed Publications

(2026). Inferring Unobservable Relations by Analogy. European Journal for Philosophy of Science.

• I show that there is a particular kind of analogical inference that has not been appreciated by philosophers and cognitive scientists. I argue that this kind of inference plays an important role in scientific — especially economic — practice.

(2024). Formal Models and Justifications of Democracy. Synthese.

• I show that Hong and Page’s [2004] computational results are not robust under both minor and more substantial parameter changes. I then examine the consequences this has for the use of the models in political philosophy, especially with respect to Landemore’s [2013] defense of democratic theory.   

(2020). How to Think About Analogical Inference: A Reply to Norton . Studies in History and Philosophy of Science Part A.

• I note some issues with John Norton’s “material theory” of analogy and propose some ways forward.

Book Reviews

(2024). Book Review: Economizing Our Lives. Philosophy of the Social Sciences. 

• Book review of Erik Angner’s How Economics Can Save the Worldthat cautions against metaphorically framing all aspects of our lives in terms of economics.

Under Review (please email me for drafts)

No Reproducibility for AI-Assisted Systematic Reviews? No Problem [under R&R]

• I argue that a strict kind of reproducibility is not necessary for trustworthy systematic reviews.

  Selected as Best Paper at LMU’s Epistemology of Medicine Conference.

  Edited version accepted at AI4Metascience Workshop at NeurIPS 2026.

Easy Similarity, Hard Representation

• I distinguish between two kinds of similarity judgments made when reasoning analogically from models: an easy and a hard kind. I then show how some of the most forceful recent objections against the similarity view of model representation are substantially mitigated when this distinction is taken into account.