In this phase, sources of uncertainty relating to the decision problem, input data, model assumptions, and outcomes are systematically recognised and described.
Image:
Objectives of this phase:
- Identify different types of uncertainties affecting the decision problem.
- Understand how uncertainty may influence the assessment and comparison of alternatives.
- Distinguish between reducible (epistemic) and irreducible (aleatory) uncertainties.
- Prepare for subsequent analysis (e.g. sensitivity analysis and risk assessment).
Uncertainties may arise from:
- Limited or incomplete data.
- Measurement errors or variability.
- Model structure and methodological choices.
- Future developments and external conditions.
- Stakeholder judgments and preferences.
Depending on the characteristics of the case, these uncertainties may be modelled explicitly by using dedicated methods for modelling uncertainties or implicitly in the sensitivity analysis. Explicit methods for modelling uncertainties include those relying on probabilities and expected utilities (e.g. multi-attribute utility theory) and those relying on imprecision ranges of parameter input (e.g. preference programming).
Outcome of the phase: A structured list and classification of the key uncertainties affecting the decision problem (e.g. data-, model-, scenario- and preference-related uncertainties) and, when explicit methods are used, a model for analysing them.
Follow up Questions for the AI
Ask the following questions to get more information from the AI:
- Can you elaborate what types of uncertainties are relevant in decision analysis and what are the sources of different types of uncertainties?
- Can you elaborate the differences between the methods that model uncertainties by probabilities and expected utilities (e.g. multi-attribute utility theory) and the methods that model imprecision by ranges of parameter inputs?
- What methods can be used to deal with uncertainty in later phases of the process?
Image:
Image:
Risks and Benefits of AI
Image:
Benefits (+)
- Identifying potential sources of uncertainties based on available information.
- Structuring and categorising uncertainties systematically.
- Supporting documentation and explanation of uncertainty sources.
- Suggesting methods for dealing with different types of uncertainties.
Benefit Scoring: + Low
Image:
Risks (!!)
- Overlooking context-specific uncertainties not present in training data.
- Misinterpreting the nature or importance of uncertainties.
- Creating a false sense of completeness in uncertainty identification.
- Lack of transparency in how uncertainties are inferred
- Ignoring stakeholder-specific perceptions of uncertainty.
Risk Scoring: !! Medium
Questions to get more information from the AI
Questions to make use of the benefits
- I have this problem: [description of the problem]. Can you give me suggestions for potential sources of uncertainties in these kinds of problems and which uncertainties are likely to have the greatest impact on the results?
- I have this problem: [description of the problem]. Can you suggest explicit methods for dealing with different types of uncertainties?
Questions to avoid the risks
- Why is identifying uncertainties challenging in complex decision problems?
- What are the limitations of AI in recognising uncertainties?
Questions NOT to ask
- I have this problem: [description of the problem]. What are the uncertainties affecting this problem?
Uncertainty identification is inherently incomplete and context-dependent.