11 - Imprecise Bayesianism

Summary

INTRODUCTION

The second and third parts of the book developed a model of rational decision making based on a number of idealisations. Firstly, that agents’ attitudes – their beliefs, desires and preferences – are consistent both in themselves and with respect to one another. Secondly, that agents are logically omniscient, in that they believe all logical truths and endorse all the logical consequences of their attitudes. Thirdly, that they are fully aware of all relevant prospects, including what options they have, what the possible states of the world are and what the potential consequences of exercising their options are. And, lastly, that agents are maximally opinionated, in that they have determinate belief, desire and preference attitudes to all prospects under consideration.

These assumptions had a purpose: they simplified the discussion enough to allow the presentation of a comprehensive theory of rationality. There is no doubt, however, that they are unrealistic. All of us violate some of them most of the time and most of them some of the time. Anyone interested in a descriptively accurate account of human reasoning and decision making will therefore want to dispense with them. And indeed there has been a great deal of work in this area over the last few decades, broadly under the banners of bounded rationality and behavioural economics, which has emphasised the extent to which our thinking and choosing are affected by factors such as context, framing, emotions and biases.

Normative theories of rationality also need to dispense with these assumptions; not in order to become more descriptively accurate (that is not their goal) but in order to provide prescriptions relevant to real agents. The aim is not to abandon full rationality for a bounded variant that more accurately corresponds to the way human decision makers are, but to characterise (full) rationality for agents who are bounded in the sense that they are not maximally opinionated, logically omniscient, fully aware or even completely consistent. So each of these assumptions will need to be relaxed to some degree, both in order to examine what modifications are required to the theory of rationality developed so far and in order to say something about the rational way to behave when you know yourself to be bounded in these ways.