On this page we provide an overview of general benefits of using AI in decision analysis, to help you ensure you are utilising AI responsibly, where it can it add value.
Image:
Benefits of Using AI in Decision Analysis
Generative AI offers significant opportunities to support decision analysis by enhancing learning, information processing, creativity, communication, and analytical work. It can assist decision-makers, analysts, and stakeholders throughout the decision analysis process by helping them understand concepts, explore alternatives, identify relevant factors, synthesise information, and communicate results.
At the same time, AI should be understood as a support tool rather than a source of authoritative answers. Responsible use requires awareness of its limitations, critical evaluation of its outputs, and transparency about how AI has been used in the process.
Principles for responsible AI use in decision analysis include at least the following:
- Human accountability remains with decision-makers
- AI outputs are hypotheses, not facts
- Important claims should be verified using credible sources
- Stakeholders should know when AI has been used
- AI-generated content should be documented
- AI-supported analyses should be reproducible when possible
- Sensitive data should not be shared with unauthorised AI systems
Together, these principles help to ensure that AI is applied transparently and rigorously, without undermining stakeholder participation, accountability, or trust in the decision-making process.
Interpreting AI Benefit Scores
The AIDA module uses a simple three-level scoring system to indicate the potential benefits of using AI in different phases of the decision-analysis process.
+ Low benefit: A low score indicates that AI can provide useful support but plays only a limited role in the process.
++ Medium benefit: A medium score indicates that AI can make a significant contribution when used carefully and combined with expert knowledge and stakeholder input.
+++ High benefit: A high score indicates that AI is particularly well suited to the task and can substantially improve efficiency, understanding, communication or learning.
In general, AI works best when it helps people learn, organise information, explore different options and explain their ideas more clearly.
Table 1 summarises the benefits and risks associated with different AI roles and provides an overall assessment of their suitability for supporting decision analysis.
| AI role | Typical tasks | Benefits* | Risks* | Overall suitability of AI |
| Tutor / Explainer | Explaining concepts, methods and terminology | +++ | ! | Excellent if able to avoid misinterpretation |
| Communication assistant | Reports, summaries, visualisations, stakeholder communication | +++ | ! | Excellent if able to avoid misinterpretation |
| Knowledge synthesiser | Summarising literature, methods, examples | ++ | ! | Very good with verification |
| Brainstorming partner | Generating problem framings, objectives, alternatives, and uncertainties | ++ | !! | Good if generated issues are not accepted uncritically |
| Structured thinking facilitator | Generating checklists, value trees, stakeholder maps | ++ | !! | Good with verification |
| Information retrieval assistant | Finding studies, methods, indicators, examples | ++ | !! | Good with verification |
| Analyst of existing data | Summarising impact matrices, preference data, monitoring data | ++ | !! | Good if inputs are reliable |
| Visualisation assistant | Visualisation of the results | ++ | !! | Good, assuming the results are correct |
| Computational assistant | Calculations, sensitivity analyses | + | !! | Risky without verification of results |
| Method selection and process planning advisor | Suggesting DA methods and participatory approaches | + | !! | Use cautiously due to context dependency |
| Preference interpreter | Interpreting stakeholder values and priorities | + | !!! | Risky due to highly personal nature of values/priorities |
| Policy advisor | Formulating recommendations | + | !!! | Risky if recommendations are accepted uncritically |
| Impact assessor | Estimating impacts of alternatives | 0/+ | !!! | Limited suitability due to context dependency |
| Value elicitation assistant | Assigning values, weights, trade-offs | 0/+ | !!! | Poor suitability because values are highly personal |
| Decision Maker Substitute | Selecting alternatives or making decisions | 0 | !!! | Should not be used |
Image:
Image:
