Decision Making and Belief Change in the face of Severe Uncertainty: A Confidence-Based Approach – DUSUCA

The project will involve a combination of theoretical methods developed in economic decision theory, conceptual analysis from epistemology, philosophy of decision, moral philosophy and philosophy of science, and experimental techniques from behavioural economics and beyond. It will confront with current practice in relevant parts of fields such as engineering, climate science, risk analysis and statistics.

How should you decide in the face of severe uncertainty? This project develops a reply that recognises the role of confidence in belief, and guides decision according to the following maxim: the more important the decision, the more confidence is required in a belief for it to be mobilised in the decision. Moreover, evaluating theories on a range of normative criteria, the approach developed here turns out to provide the best existing answer to this question.

When should one defer a decision? We provide the first account that identifies and characterises a role for confidence in deferral, which that is faithful to the simple intuition that lack of confidence can be a reason to defer. It can explain cases of deferral that remain mysterious under standard accounts.

How can and should the conclusions of the Intergovernmental Panel on Climate Change (IPCC) inform environmental policy making? This question is non-trivial – and indeed, has to do not been answered – largely because the IPCC uses a specifically-developed uncertainty language which does not provide the precise probability values required by the standard normative model of decision making (Bayesianism). We show that the answer seems to be: according to the confidence-based approach developed in this project. Indeed, this model turns out to be the only existing decision procedure that can fully use the resources provided by the IPCC without requiring richer data that cannot be expressed in their language. Moreover, it provides concrete recommendations for future uncertainty reporting.

In situations of severe uncertainty, beliefs may be imprecise but they may also vary over actors, as may their values. How should such beliefs and values be aggregated to inform a decision on behalf of society? In the face of well-known impossibility results, we show that such reasonable aggregation is possible, proposing some new aggregation methods for imprecise conflicting beliefs and conflicting values.

There is a plethora of proposed representations of beliefs and theories of decision making that claim to go beyond the Bayesian benchmark in their treatment of severe uncertainty. But which, if any, should a rational decision maker use? The first explicit attempt at identifying normative criteria for non-Bayesian models of belief and decision, and at evaluating the main accounts in the philosophical and economic literature according to them, yields an answer: the confidence approach developed in this project. It turns out to have the strongest normative credentials of all proposals in this literature.

This approach requires judgements not only about the beliefs a decision maker (be it an individual or a society) holds, but of his confidence in his beliefs. Can such judgements be provided for the toughest decisions we face, such as those concerning environmental policy, for instance? The answer is somewhat surprising: not only they can be provided, but they already are! The present state of scientific knowledge about climate change and its impacts is periodically summarised by the Intergovernmental Panel on Climate Change (IPCC), but their reports are written in a specifically-developed uncertainty language. Not only does this language (and hence their reports) provide precisely the confidence information required by the confidence-based approach developed in this project, but this is the only existing decision procedure that can fully use the resources provided by the IPCC without requiring richer data that cannot be expressed in their language. This not only suggests that climate policy is an ideal area where the proposed approach can be applied, but, in the light of the previous finding, it provides a perhaps unexpected vindication of IPCC practices.

Further work in the project will explore this interface, as well as progressing on the other central question, belief change.

1. R. Bradley, C. Helgeson, B. Hill, Climate Change Assessments: Confidence, Probability and Decision, Philosophy of Science, forthcoming.

2. E. Danan, T. Gajdos, B. Hill, J-M. Tallon, Robust Social Decisions, American Economic Review 106(9): 2407-25, 2016.

3. B. Hill, Incomplete Preferences and Confidence, Journal of Mathematical Economics, 65: 83-103, 2016.

4. C. Helgeson, E Thompson, R Frigg, Expert Judgement for Climate Change Adaptation, Philosophy of Science, 83 : 1110–1121, 2016.

How should governments decide in the face of radical uncertainties such as those involved in climate change, energy policy, genetically modified organisms or nanotechnologies? How should scientists incorporate new, yet inconclusive evidence into their current state of knowledge and uncertainty? The standard normative theory of decision making and belief change in philosophy and economics, Bayesianism, seems incapable of giving satisfactory replies to these questions. The central aim of this project is to deal with the fundamental issue of which notion of rationality is appropriate for the new risks and severe uncertainties that we as a society are faced with today. A normative theory of decision making and belief change based on the idea that one's confidence in one's beliefs has a role to play in decision and learning will be developed. Consequences will be drawn for public decision making, and the ground shall be laid for applications to concrete decisions. Awareness of these fundamental questions and possible replies shall be promoted among actors in the decision making process.