When should the lawful evidence presented to a judge become proof? This paper argues that Bayesian decision theory can answer that question—not by reducing legal judgment to arithmetic, but by making explicit what is at stake when courts risk punishing the innocent or letting the guilty go free. Its central idea is this: judges should minimize expected justice costs. From that single thought, the paper derives a probability threshold for conviction, a principled prior probabi

My reply with Holger Andreas to the commentaries on our lead article "Factual Difference-Making" is now out. The reply refines our analysis of causation such that it can account for all the counterexamples raised in the commentaries. I would like to thank Cei Maslen, as well as Justin Sytsma and Ed Mares, for outstanding editorial support. I am also grateful for the insightful commentaries by Ned Hall, Christopher Hitchcock, Brad Weslake, Rose McDonald, Tom Wysocki, Vera Hoff

I had the pleasure to moderate the session on Protecting Vulnerable and Underserved AI Users at the 2026 conference of the International Association for Safe and Ethical AI. You can watch the whole session here: https://virtual.oxfordabstracts.com/event/public/75968/session/196861 .