Sven Ove Hansson

Although it has often been claimed that all the information contained in second‐order probabilities can be contained in first‐order probabilities, no practical recipe for the elimination of second‐order probabilities without loss of information seems to have been presented. Here, such an elimination method is introduced for repeatable events. However, its application comes at the price of losses in cognitive realism. In spite of their technical eliminability, second‐order probabilities are useful because they can provide models of important features of the world that are cognitively more plausible than those that can be obtained with single‐level probabilities.

Moral theory has mostly focused on idealized situations in which the morally relevant properties of human actions can be known beforehand. Here, a framework is proposed that is intended to sharpen moral intuitions and improve moral argumentation in problems involving risk and uncertainty. Guidelines are proposed for a systematic search of suitable future viewpoints for hypothetical retrospection. In hypothetical retrospection, a decision is evaluated under the assumption that one of the branches of possible future developments has materialized. This evaluation is based on the deliberator’s present values, and each decision is judged in relation to the information available when it was taken. The basic decision rule is to choose an alternative that comes out as morally acceptable (permissible) from all hypothetical retrospections.

Clear-cut cases of decision-making under risk (known probabilities) are unusual in real life. The gambler’s decisions at the roulette table are as close as we can get to this type of decision-making. In contrast, decision-making under uncertainty (unknown probabilities) can be exemplified by a decision whether to enter a jungle that may contain unknown dangers. Life is usually more like an expedition into an unknown jungle than a visit to the casino. Nevertheless, it is common in decision-supporting disciplines to proceed as if reasonably reliable probability estimates were available for all possible outcomes, i.e. as if the prevailing epistemic conditions were analogous to those of gambling at the roulette table. This mistake can be called the tuxedo fallacy . It is argued that traditional engineering practices such as safety factors and multiple safety barriers avoid this fallacy and that they therefore manage uncertainty better than probabilistic risk analysis (PRA). PRA is a useful tool, but it must be supplemented with other methods in order not to limit the analysis to dangers that can be assigned meaningful probability estimates.

Mainstream risk analysis deviates in at least two important respects from the rationality ideal of mainstream economics. First, expected utility maximization is not applied in a consistent way. It is applied to endodoxastic uncertainty, i.e. the uncertainty (or risk) expressed in a risk assessment, but in many cases not to metadoxastic uncertainty, i.e. uncertainty about which of several competing assessments is correct. Instead, a common approach to metadoxastic uncertainty is to only take the most plausible assessment into account. This will typically lead to risk-prone deviations from risk-neutrality. Secondly, risks and benefits for different persons are added to form a total value of risk. Such calculations are used to support the view that one should accept being exposed to a risk if it brings greater benefits for others. This is in stark contrast to modern Paretian welfare economics, that refrains from interindividual comparisons and does not require people to accept a disadvantage because it brings a larger advantage for others. (Published Online July 11 2006).

A value statement such as “she is a good teacher” is categoryspecified, i.e., the criteria of evaluation are specified as those that are applicable to a given category, in this case the category of teachers. In this study of categoryspecified value statements, certain categories are identified that cannot be used to specify value aspects. Special attention is paid to categories that are constituted by functional characteristics. The logical properties of value statements that refer to such categories are shown to differ significantly from the corresponding properties in social choice theory.

An experiment, in the standard scientific sense of the term, is a procedure in which some object of study is subjected to interventions that aim at obtaining a predictable outcome or at least predictable aspects of the outcome. The distinction between an experiment and a non-experimental observation is important since they are tailored to different epistemic needs. Experimentation has its origin in pre-scientific technological experiments that were undertaken in order to find the best technological means to achieve chosen ends. Important parts of the methodological arsenal of modern experimental science can be traced back to this pre-scientific, technological tradition. It is claimed that experimentation involves a unique combination of acting and observing, a combination whose unique epistemological properties have not yet been fully clarified.

Mainstream moral theories deal with situations in which the outcome of each possible action is well-determined and knowable. In order to make ethics relevant for problems of risk and uncertainty, moral theories have to be extended so that they cover actions whose outcomes are not determinable beforehand. One approach to this extension problem is to develop methods for appraising probabilistic combinations of outcomes. This approach is investigated and shown not to solve the problem. An alternative approach is then developed. Its starting-point is that everyone has a prima facie moral right not to be exposed to risk. However, this right can be overridden if the risk-exposure is part of an equitable system for risk-taking that works to the advantage of the individual risk-exposed person.

Although it has often been claimed that all the information contained in second-order probabilities can be contained in first-order probabilities, no practical recipe for the elimination of second-order probabilities without loss of information seems to have been presented. Here, such an elimination method is introduced for repeatable events. However, its application comes at the price of losses in cognitive realism. In spite of their technical eliminability, second-order probabilities are useful because they can provide models of important features of the world that are cognitively more plausible than those that can be obtained with single-level probabilities.

Available methods such as technology assessment and risk analysis have failed to predict the effects of technological choices. We need to give up the futile predictive ambitions of previous approaches and instead base decisions on systematic studies of alternative future developments. It will then be necessary to cope with mere possibility arguments, i.e., arguments in which a conclusion is drawn from a mere possibility that a course of action may have certain consequences. A five-step procedure is proposed for the assessment of such arguments. It includes a search for mere possibility arguments pointing in different directions, a scientific evaluation that may lead to the specification or refutation of some of these arguments, two symmetry tests, evaluation of the seriousness of the arguments in terms of novelty, spatio-temporal unlimitedness and interference with complex systems, and hypothetical retrospection that aims at finding a course of action that will be defensible in retrospect.