Calculating the Results

In this phase, the decision analysis model is used to calculate the results based on the information gathered and modelled in the previous phases.

Calculation is typically based on this input:

Impacts of alternatives
Value functions and valuation of impacts
Stakeholder preferences and weights
Possible uncertainty information

The purpose of this phase is to combine these elements according to the selected decision analysis approach and generate the outputs required for comparing the alternatives.

Possible results of the calculations:

Overall values or utility scores for the alternatives
Rankings of alternatives
Dominance or outranking relationships
Classification of alternatives into categories
Measures of uncertainty or confidence

The outcome depends on the selected decision analysis method

The results may be presented separately as a stakeholder-specific results or a group-level results embedding the results of all the stakeholders.

The calculations should be carried out in a transparent and reproducible manner.

Issues to be ensured by the decision analyst:

The model has been correctly implemented, and all input data have been correctly transferred into the model
Assumptions are documented
Calculation procedures are traceable and the results are mathematically consistent

Although the model generates the results, these should not yet be interpreted as recommendations or decisions. The purpose of this phase is simply to calculate the consequences implied by the assumptions, impact assessments, and preferences incorporated into the model.

Outcome of the phase: Results calculated with the decision analysis model, such as scores or values of alternatives, rankings or classifications, stakeholder-specific and/or group results, comparison matrices and uncertainty measures. These outputs provide the basis for the subsequent comparison and interpretation of alternatives.

Follow up Questions for the AI

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

How are results calculated in different decision analysis methods?
What are the key assumptions behind aggregation of criteria and how the aggregation procedure works?
What calculation errors are common in decision analysis models?
How can the correctness of calculations be verified?
How should stakeholder-specific and group-level results be generated?
How can uncertainty information be incorporated into the calculations?

Benefits and Risks of Using AI

Benefits (++)

- Assisting in implementation of calculation procedures.

- Helping identify inconsistencies in model inputs.

- Explaining mathematical procedures used in the model.

- Supporting generation of result tables and visualisations.

- Automating repetitive computational tasks

Benefit Scoring: ++ Medium

Risks (!!!)

- Mathematical or computational errors in the calculations or use of incorrect aggregation procedures.

- Misinterpretation of the selected decision analysis method and of the assumptions embedded in AI-generated calculations leading to erroneous results.

- Lack of transparency regarding how results were generated.

- False confidence in apparently precise numerical results.

- Use of methods that are inconsistent with the selected decision analysis approach

Risk Scoring: !!! High

Questions to Better Understand the Benefits and Risks of AI

Questions to understand the benefits

Prerequisite for each of these questions is that there are data available from previous phases to calculate the results.

I have these impacts, value functions, and stakeholder weights. How should the results be calculated using this decision analysis method?
Can you verify whether there are inconsistencies in the input data?
What assumptions are embedded in this calculation procedure?
Can you identify potential calculation errors in this model specification?
Can you help visualize the calculated results?

Questions to avoid the risks

Mathematical or computational errors in the calculations or use of incorrect aggregation procedures.
Misinterpretation of the selected decision analysis method and of the assumptions
embedded in AI-generated calculations leading to erroneous results.
Lack of transparency regarding how results were generated.
False confidence in apparently precise numerical results.
Use of methods that are inconsistent with the selected decision analysis approach.

Questions NOT to ask

Calculate the results using the calculation approach that gives the best results.

The calculation procedure should always be explicitly defined by the decision analyst and be consistent with the selected decision analysis method. AI can support implementation, checking, and explanation of calculations, but it should not independently choose modelling assumptions or aggregation procedures. All calculations should be transparent, reproducible, and independently verifiable.
Can we assume these calculated results are correct?

AI verifying its own results in not a good idea.

Questions to ask the AI

Can you explain every calculation step used to generate the results?

Unlike some earlier phases, calculation errors can directly affect all subsequent results and conclusions. Even small mistakes in formulae, aggregation methods, or input handling may substantially alter rankings or recommendations.
How can the calculations be independently verified and which parts of the calculation should be checked manually?
What are the limitations of using AI for these calculations?

Decision analysis home

applying a decision analysis model