# Conclusions of the Workshop

The workshop closed with a group work session in which five groups were asked the same question: *Is there, based on the presentations in the workshop, a common ground for risk-informed decision-making for structures, and how can that be described?* Each group listed (a) the main principles that should be followed when assessing risk acceptance and (b) what should be avoided and where the main sources of inconsistency lie. The flipcharts of the groups are documented in the [group work slides](https://folk.ntnu.no/jochenk/JCSS/Files/BinderGW.pdf); the following is a synthesis.

## Common ground

The groups arrived, with different emphasis, at a rather consistent set of principles:

1. **Risk is the expected value of consequences, given the available knowledge.** The risk estimate is knowledge-based and therefore not objective; the state of information and the epistemic uncertainties have to be made explicit, and the analysis should be presented as knowledge-based rather than as "subjective" in a dismissive sense.
2. **Risk acceptance is a decision, and we always accept risk, implicitly or explicitly.** Prescriptive safety provisions contain implicit acceptance decisions. The relevant question is not only whether an activity is acceptable, but which mitigation measures, including the 0-alternative, should be chosen; acceptance is the attribute that determines when to stop mitigating.
3. **Relative comparison rather than absolute limits.** Within the ALARP region, acceptance follows from a comparison of alternatives. For life safety the marginal life-saving cost principle (SWTP) is the appropriate tool; it should be validated with data. Upper bounds for individual or societal risk may be needed as a reference or backstop, i.e. as a reality check, but not as the primary criterion.
4. **A holistic and societal perspective.** All negative outcomes, including indirect consequences, and all relevant mitigation actions should be captured; attributes beyond fatalities (cost, environment, reputation) matter. Life-safety risk and financial risk should be treated differently: for life safety the focus is on mitigation actions and their effectiveness, for financial risk a target safety level can be used within a decision problem.
5. **Optimisation as the basis.** Risk-informed decisions should identify the alternative with the maximum expected utility, with economic and sustainability principles accounted for. Acceptance criteria are minimum requirements; more may be needed on economic grounds.
6. **Proportionality, equal treatment and pragmatism.** Similar cases should be treated similarly, measures should be proportionate, and the system boundaries of the analysis should be stated. Risk evaluation is a process with different stakeholders involved.
7. **Risk aversion exists, but F–N criterion lines are not an appropriate way to represent it.** Large accidents may carry disproportionately larger societal costs; this is better accounted for in the consequence model than through an aversion factor. (One group would exclude risk aversion from the criterion altogether.)
8. **Transparency and standardisation.** Assumptions and their effects should be documented and clearly communicated. Risk acceptance should be standardised so that the responsibility is not borne by individual engineers; at the same time, if the code writer is the decision maker, some optimality is lost.

## What should be avoided

* F–N criteria and their misuse; oversimplification such as risk matrices; "doing the usual".
* "Check-box" analyses, and risk analyses carried out only to verify a criterion whose outcome is already known.
* Using a risk tool to justify predetermined preferences; biased or unidirectional assessments; unclear communication of assumptions.
* Neglecting uncertainties; underestimating risk through experience-based or mainstream solutions, and overestimating risk at the expense of the environment and the safety budget.
* Replacing probability by "possibility", a zero-risk attitude, and the inclusion of non-accountable (e.g. political) aspects in the estimate.
* Externalising disadvantages to third parties without compensation or further risk reduction.
* Focusing on what is best known (e.g. detailed FE modelling) rather than on what is important; lack of resolution and of resources for quality assurance; underestimating the time needed.

## Open questions

* Can risk acceptance criteria be used to release resources where the risk is already below the limit?
* Where should upper limits for individual and societal risk enter: as a backstop only, or as a formal requirement?
* How can the principles above be transferred into codes and standards without losing the decision-theoretic basis?

## Follow-up

The discussion is continued in the JCSS workshop *Societal Preferences on Risk Acceptance*, University of Waterloo, 21–22 October 2026, which revisits the foundations of risk acceptance criteria (Life Quality Index, societal willingness to pay, target reliabilities) and their regulation in different world regions.