Slightly behind the pack, it seems, I’ve suddenly started hearing about “ergodicity economics”, presented as an alternative to expected utility (EU) theory. Commenter James asked me about it here, and I also received from a colleague a copy of a paper in Nature, by Ole Peters, who appears to be the main developer of this idea. The essential idea of ergodicity is that the long-run distribution of outcomes for a dynamic process should match the uncertainty of the process at any point in time. You can get something more precise in Wikipedia. Expected utility starts with preferences over uncertainty at a point in time. Peters argues that things are better understood in terms of evolution over time. I haven’t followed all of the details of this argument as yet.
What piqued my interest is that the discussion involves a lot of discussion of probability weighting and particularly the idea that extreme low-probability outcomes may be overweighted. The most famous expression of this idea is the cumulative prospect theory put forward by Kahneman and Tversky in 1992. Their original prospect theory applied the same weighting function to all events, which raises a number of difficulties. These were resolved using the idea of rank-dependent probability weighting which I proposed in a paper in 1982 (under the name ‘anticipated utility’ and now usually called rank-dependent utility or RDU) .
The underlying reasoning is that, in a dynamic process repeated over time, taking low-probability extreme risks will (very probably) catch up with you. I’m pretty sure I made an argument of this kind in support of RDU back in the 1980s, but I haven’t been able to locate it for now.
This is one of many independent rediscoveries of the rank-dependent approach, with a variety of motivations. I think this reflects the fact that the RDU is, in some sense, the natural generalization of EU.