Great paper by Douven on IBE (vs Bayesianism). An ergodicity argument against Bayes à la @ole_b_peters and @nntaleb—Bayes may be "optimal", but only because it does well on exponentially unlikely sequences. https://www.researchgate.net/profile/Igor_Douven/publication/282945311_Inference_to_the_Best_Explanation_What_Is_It_And_Why_Should_We_Care/links/5623a07508aea35f2686845c/Inference-to-the-Best-Explanation-What-Is-It-And-Why-Should-We-Care.pdf

Douven's paper summarizes lovely recent work he's done with Schupbach on the *psychology* of explanation-making, but pulls no punches in arguing that non-Bayesian updating may be normatively (i.e., philosophically) better, and not just a heuristic or shortcut.

Would be curious to hear @nescio13's impressions (and digressions) on Douven's argument here, which is claims to rescue the old-school Inference to Best Explanation literature of van Frassen, Harman, et al—even Popper!

Or @davidbkinney—I see you liking this thread, David. Another great feature of Douven's paper is that he doesn't just punt IBE to the prior, but asks if we have good evidence (or normative justification) for having it affect how we update in the presence of information.