Choice with errors
- Start date: 01 September 2012
- End date: 31 August 2015
Substantial experimental and empirical evidence has been gathered in recent years showing the inadequacy of thestandard rational agent paradigm to capture observed choice behaviour.
The core idea of this project is that though the agent's behaviour is partly governed by "true" preferences (expressing the agent's welfare); the agent may make errors. Errors can be conceptualised as the failure on the part of the agent to apply preference maximisation (ie to be rational) over the entire set of physically available alternatives (and hence to act in a globally optimal way). In other words, the agent may fail to consider some courses of action physically available to him: for example, they may not be salient, or be excluded from assessment on psychological, ideological, moral or social grounds (before their "utility" is even considered).
The research will show that this framework subsumes a large number of existing models of bounded rationality, and proposes to develop a formal probabilistic general model that captures this idea. Further, a number of more specific and policy relevant models will be analysed, in particular to answer questions regarding the so called "dependency culture" and the design and welfare evaluation of schemes inspired by the nudge approach.