Valuating the Impacts

The goal of this phase is to map indicator impacts that are measured by some natural scales (e.g. tonnes of CO2, costs in euros, biodiversity index value, expert evaluation scale, etc.) onto a unified value scale, with the aim to make the impacts comparable with each other. In the value scale, the worst possible impact typically gets a value of 0 and the best impact a value of 1 (or 100) and the other impacts values between these, indicating their relative impact compared to the end points.

Value functions can be used to indicate how the measurement values are mapped into the value scale. Then, it is very important to consider whether the impacts can be mapped linearly to the value scale or should some nonlinear value function be used (e.g. in cases of diminishing returns or thresholds in the impact scale).

The endpoints of the impact scale should reflect the worst and best relevant outcomes within the decision context. These can be defined:

Locally, when the worst impact in the set of alternatives gets a value of 0 and the best one a value of 1 (or 100); or
Globally, when the endpoints of the impact scale are the least and most desirable but still plausible impacts. They can be, for example, a clearly unacceptable and technically maximum feasible performance level, respectively, or the endpoints of some natural scale (e.g. grading from A to F, or percentage scale).

The local value functions are suitable in situations with a clear predefined set of alternatives, and global value functions in situations in which there are clear natural endpoints of the scales, but the set of alternatives may not be defined yet.

The valuation of impacts is linked to the weighting of the criteria so that the value scales within criteria determine how strongly differences in performance matter within a single criterion, but the criteria weights determine how much each criterion matters relative to other criteria.

Please note that in some decision analytical methods (e.g. outranking methods) the impacts are not explicitly valued, but their values are derived implicitly through the process.

NetworkNature Task Force 1 AI usage guidance

Outcome of the phase: A score for each impact on a unified scale (e.g. 0–1).

Follow up Questions for the AI

Ask the following questions to get more information from the AI:

Can you describe in detail the role of the impact values in decision analysis?
Can you elaborate the linkages between the impact value scales and the weights of the criteria?
Can you elaborate in which kinds of cases assumption of linearity is valid?
Can you give examples of defining endpoints of the impact scales for the value functions?

Benefits and Risks of Using AI

Benefits (+)

- Support for finding example cases of the valuation of the impacts in similar cases.

- Support for estimating value functions from the given data.

- Support for carrying out the valuation in a case of large datasets.

Benefit Scoring: + Low

Risks (!!)

- AI giving generic impact estimates without being able to take the current decision context into account.

- Loss of transparency due to estimating values from historical data.

- Inability to take the personal stakeholder preferences into account.

- False feeling of objectivity when using AI to assess the impacts – the values of the impacts (especially regarding the threshold or nonlinearity) depend on both the context and stakeholders’ preferences.

Risk Scoring: !! Medium

Questions to Better Understand the Benefits and Risks of AI

Questions to make use of the benefits

I have this criterion/indicator: [name of the criterion/indicator]. Can you suggest different ways in which the valuation of impacts can be implemented in relation of this criterion?

The value functions themselves are typically context dependent, but AI can give suggestions on how to assess them.
I have this problem: [problem description including criteria and alternatives]. Can you give me examples of how earlier studies have valued the impacts within a similar context?

After getting the response, the decision analyst should carefully analyse whether the conditions of the earlier study are sufficiently similar to allow the results here.
I have these criteria in my problem: [list of criteria]. What kinds of differences there are typically between the stakeholders when evaluating these impacts?
AI can be applied to roughly estimate how different stakeholders value different impacts

Questions to avoid the risks

AI giving generic impact estimates without being able to take the current decision context into account.
Loss of transparency due to estimating values from historical data.
Inability to take the personal stakeholder preferences into account.
False feeling of objectivity when using AI to assess the impacts – the values of the impacts (especially regarding the threshold or nonlinearity) depend on both the context and stakeholders’ preferences.

Questions to ask from the AI

(After getting suggestions of earlier studies about the impact valuation) What issues should be considered when adapting the results of these studies to this case?

It is essential to understand the prerequisites of using the results of other cases of impact valuation this case

Questions NOT to ask

I have a problem with these impacts of alternative to criteria: [Impact matrix]? Can you make the mapping of the impact values into the value scale?

The values of the impacts are typically very context dependent and depend on the stakeholder preferences. In addition, the relation between the impacts and their values can be nonlinear and there could be thresholds, and the response from AI can be very biased if not being able to take these issues into account.

decision analysis home

applying a decision analysis model

Valuating the Impacts

Follow up Questions for the AI

Ask BIONEXT

Benefits and Risks of Using AI

Benefits (+)

Risks (!!)

Questions to Better Understand the Benefits and Risks of AI

Previous Stage

Next Stage