Analysing uncertainties: Towards comparing Bayesian and interval probabilities

Two assumptions, commonly made in risk and reliability studies, have a long history.
The first is that uncertainty is either aleatoric or epistemic. The second is that standard
probability theory is sufficient to express uncertainty. The purposes of this paper are to
provide a conceptual analysis of uncertainty and to compare Bayesian approaches with
interval approaches with an example relevant to research on climate change. The
analysis reveals that the categorisation of uncertainty as either aleatoric or epistemic is
unsatisfactory for practical decision making. It is argued that uncertainty emerges from
three conceptually distinctive and orthogonal attributes FIR i.e., fuzziness, incompleteness (epistemic) and randomness (aleatory). Characterisations of uncertainty, such as
ambiguity, dubiety and conflict, are complex mixes of interactions in an FIR space. To
manage future risks in complex systems it will be important to recognise the extent to
which we ‘don’t know’ about possible unintended and unwanted consequences or
unknown–unknowns. In this way we may be more alert to unexpected hazards. The
Bayesian approach is compared with an interval probability approach to show one way
in which conflict due to incomplete information can be managed.