Decision Theory with a Human Face

The aim of this research is the production of a book on decision making that applies to situations in which we don't know what all the alternative are and have not made up our mind about those that we are aware of.

The first part of the book will present, explain and then reformulate the core of Bayesian decision theory, currently the standard theory of decision making. The focus will be on foundational questions and in particular on two problems: one normative, the other epistemological. The first is what might be termed the problem of justification, namely that of providing grounds for believing that people should follow its prescriptions. The second is the problem of interpretation, which in its starkest form is the question of how we can have knowledge of non-observable mental states such as beliefs and preferences. A solution to this problem is crucial to the applicability of decision theory, for unless we can determine what someone believes or desires, we cannot use the theory to say what he should do or explain what he was observed to do.

The second part will address the problem of decision making in the kinds of situations described above. The main thought here is that since people cannot plan for all contingencies, a theory which gives advice on the presumption that agents can make their minds up about everything in advance is going to be of limited help. Consequently, this part of the project will attempt to address three main research questions. Firstly, how people should change their mind about the possibilities they face as they become aware of previously unforeseen features of them? Secondly, how should they make up their mind about what probability and desirability to attach to prospects in the first place? And thirdly, how should they make decisions when they have not made their mind up about all relevant features of the decision problem, including how likely are the various contingencies upon which the success of the their decisions depend and how desirable are the possible consequences of the choices they make.

In addition to its benefits to the academic community, the proposed research has the potential to benefit decision making in many areas of social life and especially those in which decisions typically have to be made under the kinds of non-ideal conditions that will be investigated. For example, much public policy decision making is made with partial information about the possible consequences of adopting one or another policy and with incompletely formed public preferences over them, but using tools appropriate to conditions of complete information. The emphasis in the proposed research project on ways in which both the factual and normative inputs to decisions can be formed and revised has the potential to encourage policy decision makers to think about the robustness of policy in the face of such changes.

The proposal to publish a book on decision making under non-ideal conditions is itself motivated by the desire to increase the impact of decision theoretic work on decision makers in both public and private institutions. Much decision theoretic research is inaccessible to this audience in virtue of its highly technical nature. In contrast the book will be written in a relatively non-technical way, making it possible for a much wider group of people to draw lessons from it.