Kieran Gibson does some nifty experiments, to check if the subjects (MTurk users) made coherent decisions. They do not. Here is the key chart:
He also rules out ceiling effects and unexpected utilities by using probabilistic rewards. All the VNM criteria really.
My explanation is that this result it due to bounded rationality. Most people can’t hold the joint reward function in their mind and will choose or alternate between the individual functions instead—which leads to averaging instead of adding.
[Linkpost] Competing Motivations: When More Incentives Lead To Less Effort
Link post
Kieran Gibson does some nifty experiments, to check if the subjects (MTurk users) made coherent decisions. They do not. Here is the key chart:
He also rules out ceiling effects and unexpected utilities by using probabilistic rewards. All the VNM criteria really.
My explanation is that this result it due to bounded rationality. Most people can’t hold the joint reward function in their mind and will choose or alternate between the individual functions instead—which leads to averaging instead of adding.
Link to PDF Download
See also Coherent decisions imply consistent utilities