Various decision analysis methods are available to support decision-making (e.g. Multi- Attribute Value/Utility Theory, Analytic Hierarchy Process, Outranking methods, etc). They differ from each other, for example, in terms of how preference information is elicited, how preference relations between alternatives are represented, and how results are produced.
The choice of the method should depend on the needs of the case, as no approach is universally suitable for every problem.
Often, one has to make compromises between these issues when selecting the approach.
In group decision-making situations, there are also various ways to deal with the stakeholder preferences, which affect the choice of the decision analysis approach:
Outcome of the phase: A decision on which approaches will be applied in the process
Ask the following questions to gather more information from the AI:
- Suggestions for methods to be used based on the requirements of the case.
- Analysis of the differences between the approaches.
- Summarising the advantages and weaknesses of different methods.
Benefit Scoring: ++ Medium
- AI is not able to adapt the suggestions of the approach according to the characteristics of the case.
- AI is unable to compare the approaches in a structured way.
- In the literature, it is often explained why a certain approach is suitable for the problem, but it is not described why other approaches are less suitable. In addition, in the literature, suitability is often compared to another single approach, when only the difference from that approach is explained.
- Lack of analytical rigour in the comparison of the methods. For example, when asked to list weaknesses, AI may mention a weakness for only one approach even if it applies to several approaches.
- Universal interpretation of weaknesses and advantages. For example, AI can consider “subjectivity of preferences” as a weakness, which can be true in some cases. However, in many cases the whole point of decision analysis is to take the different views of the stakeholders (which are subjective preferences by nature) into account. In such cases, “subjectivity of preferences” is actually an advantage
Risk Scoring: !! Medium
AI can give reasonable answers, but if the decision context is too loosely defined, the answers can be too general. These kinds of questions should always be followed by “Why” questions and questions about the disadvantages of using the method.
