Is Agency Identifiable? Apr 19, 2023 10 min read AI Safety Identifiability in IRL One of my favorite papers is this one, titled "Occam's razor is insufficient to infer the preferences of irrational agents". It relates to an area of