Behavioural experiments by Ollie Hulme’s group in Copenhagen show that people change their implied risk preferences when choosing between repeated multiplicative versus additive gambles. The shifts are consistent with a model assuming that the subjects intend to maximise the time-averaged growth rate of wealth; however, such a model is not cognitively realistic.

In this talk, Colm Connaughton (London Mathematical Laboratory) presents Lay Len Ching’s recent work exploring the extent to which a class of heuristic decision models known as fast-and-frugal trees can be trained to play the Copenhagen experiment. The findings suggest that simple heuristics may provide a cognitively plausible model for how people adapt to the changing risk dynamics in the experiment. This talk can be thought of as a follow-on from Özgür Şimşek’s seminar on the design of fast-and-frugal trees.

The seminar was hosted by James King (Science Practice) and Emilie Rosenlund Soysal (London Mathematical Laboratory).