Decisionmaking under uncertainty

Decision theory purports to study human decisions descriptively and provide a normative framework for rational decision-making. The elements of decision theory are quite simple: a choice between different courses of action; some knowledge about different outcomes or consequences of these options; and, finally, an evaluation of each outcome, that is a value attached to every consequence based on preferences. Generally four types of practical decision problems can be distinguished: a decision under certainty; a decision under risk; a decision under uncertainty; and a decision under ignorance. In the case of certainty we know the outcomes of different choices and the only challenge is to be clear about one’s preferences. In the case of risk we know the outcomes (benefits and adverse effects) and the probability of various outcomes. In the case of uncertainty we know the possible outcomes but have no objective ground to estimate their probability. In the case of ignorance we do not even know what adverse effects to anticipate or we don't know their magnitude or relevance and have no clue of their probability.

When both the utility and the probability of the various outcomes of a decision are known, maximizing expected utility is generally advocated as a rational decision rule. However, this is not the case with the Precautionary Principle (PP), which applies to decisions under uncertainty.

Risk management based on quantitative risk assessment and the setting of quantitative norms and standards for acceptable risk for different activities has become the dominant paradigm in the risk policies of many nation States. This approach is often regarded as scientific, because it draws on empirical evidence. It is, however, not a purely objective endeavor because it employs normative assumptions about the types of harms that should be addressed; the level of risk that is acceptable; the choice of a limited set of risk dimensions that are considered in the judgment of acceptability; the implicit choice to consider the unquantifiable as well as the distribution of benefits and harms to be irrelevant.
The usefulness of this quantitative approach is further limited by lack of agreement about the utilities or indicators to be used in the risk assessment to compare outcomes for different decision options (for example dollars lost/saved, lives lost/saved, species lost/saved, years of life lost/gained, etc.) and how to weigh them if different indicators are used simultaneously. Finally, scientific uncertainties and knowledge gaps that hamper the ability to reliably assign probabilities to the various outcomes.

Different rational decision strategies have been developed for decisions where the probability of outcomes is unknown. What approach is the best depends, however, on one’s attitude towards risk, that is whether one is for instance risk-averse, risk-tolerant, or risk-seeking.
For instance, maximin is the strategy that chooses the option that has the best (that is: the least severe) worst-case scenario. It makes sense if we have little to win and a great deal to lose, but it tends to prevent us from taking advantage of opportunities. Such a strategy seems the only rational course when we are gambling with outcomes that affect not only us, but also others. It would be unjust to let others suffer unnecessarily from my unlucky choices. One may note that the maximin strategy already contains the seeds of precaution. Closely related to maximin is the difference principle: one society is better off than another if the worst-off members of the former do better than the worst-off of the latter. Maximin allows the most disadvantaged members of society to be harmed if the overall society benefits; the difference principle would forego an overall benefit to the society if it harmed the most disadvantaged members.

Arguing from an ethical point of view, one may say that in certain types of situations the use of decision theory prescribes the course of action that is both rational and ethical. One could even say that decision theory not only may, but also should be used in ethics. People who have moral goals, should seek to realize them rationally. If the goals should be achieved, then rationality should control the relationship between means and ends. There is, however, an important proviso to this claim: some important types of situation demand close attention to morally relevant aspects and facts that are not routinely captured in decision theory.

The PP has arisen from unresolved problems of the existing decision support approaches outlined above. When the bounds of the possible outcomes are not known and no credible ground exists for the quantification of probabilities, and ethical dimensions of inter- and intra-generational equity are at stake, the other decision principles fail to satisfactorily address these problem characteristics. For exactly these cases, the PP offers a rational alternative. Because the PP applies to those cases where serious adverse effects and surprises can occur with an unknown probability, it is rational to follow a 'better safe than sorry' strategy. Failing to take precautionary measures in a timely manner could result in devastating and irreversible consequences. Such consequences might have been avoided by proactive and anticipatory interventions whose costs are justifiable in comparison to the damages and losses that could occur.

References
UNESCO COMEST (2005) The Precautionary Principle, UNESCO, Paris.