Date: Novermber 14, 2025 (Friday)
Time: 15:30-17:00
Venue: MB142, 1/F, Main Building, HKU
Speaker:
Professor James Joyce , C. H. Langford Collegiate Professor of Philosophy, University of Michigan
Abstract:
After introducing the notion of credal accuracy, and explaining why formal epistemologists use strictly proper scoring rules to measure it, I will introduce some well-known accuracy scores and discuss some of their strengths and weaknesses. A key question will be this: should an acceptable measure of credal accuracy make it the case that having veridical learning experiences always improves credal accuracy, or can learning truths make one’s credences less accurate. As we will see, some scores have this property and others do not. Recently, Peter Lewis and Dan Fallis have argued that all acceptable rules should have this property. I shall argue that they are wrong. I will close by bringing up a paradoxical case that I don’t know how to handle.
