## Abstract

It is suggested to look on probabilistic risk quantities and concepts through the prism of accepting one of the views: whether a true value of risk exists or not. It will be argued that discussions until now have been primarily focused on closely related topics that are different from the suggested one.

In general, the values of risks are not known precisely and the analyst has the option to consider that convergence to a precise value of risk is possible in the limit. That is, the true value exists but due to limited time, resources or other limitations in assessing probabilities it is not known at the time being. Following this prospective, a single probability distribution over a set of possible outcomes can be chosen that is tacitly regarded as a ‘true’ or ‘ideal’ model of uncertainty. After that, computing other probabilistic risk measures of interest becomes a rather easy mathematical exercise. In fact, accepting the true-value view does not make the adherent use only a single probability distribution as a model of uncertainty. One can introduce a class of distributions in which a particular member is considered a plausible candidate to be an ideal distribution. Then, a risk quantity of interest can be computed for each distribution-candidate and after that, lower and upper bounds can be constructed as assessments of risk. This is a robust way of compensating for ignorance that studies the sensitivity of derived numerical values to variations in probability distributions.

In general, the values of risks are not known precisely and the analyst has the option to consider that convergence to a precise value of risk is possible in the limit. That is, the true value exists but due to limited time, resources or other limitations in assessing probabilities it is not known at the time being. Following this prospective, a single probability distribution over a set of possible outcomes can be chosen that is tacitly regarded as a ‘true’ or ‘ideal’ model of uncertainty. After that, computing other probabilistic risk measures of interest becomes a rather easy mathematical exercise. In fact, accepting the true-value view does not make the adherent use only a single probability distribution as a model of uncertainty. One can introduce a class of distributions in which a particular member is considered a plausible candidate to be an ideal distribution. Then, a risk quantity of interest can be computed for each distribution-candidate and after that, lower and upper bounds can be constructed as assessments of risk. This is a robust way of compensating for ignorance that studies the sensitivity of derived numerical values to variations in probability distributions.

Original language | English |
---|---|

Publication date | 2016 |

Publication status | Published - 2016 |

Event | SRA Europe 2nd Nordic Chapter Meeting - Gothenburg, Sweden Duration: 14 Nov 2016 → 15 Nov 2016 |

### Conference

Conference | SRA Europe 2nd Nordic Chapter Meeting |
---|---|

Country | Sweden |

City | Gothenburg |

Period | 14/11/2016 → 15/11/2016 |