@StatMLPapers I like @cerullig recasting multi-armed bandits as an ML with observed data problem. Kind of surprised this hasn't been done before in this setting, but the dependence on risk preference is also well-explored in microeconomics, and I'd think someone did it in that context.