UK Evaluation Society launch co-produced AI in evaluation: Good practice guidelines for practitioners
The UK Evaluation Society has published professional guidance that takes a scenario-based approach to AI use in evaluation. Rather than abstract principles, it demonstrates decision logic through examples ranging from straightforward literature synthesis to complex interview analysis to situations where AI shouldn’t be used.
What’s different about this guidance:
- Shows how principles scale across different risk profiles
- Includes explicit “do not proceed” scenarios
- Provides a practical six-question checklist for evaluators
- Acknowledges known limitations and unresolved questions up front
The guidance emerged from a working group spanning government, development, private sector, and academic evaluation practice. It reflects real implementation challenges rather than theoretical frameworks.
The guidance will be reviewed regularly, with updated versions reflecting the evolving AI landscape, continued collaboration and feedback from the evaluation community, and technology advancements.
The guidance was co-produced through the collective effort of the UK Evaluation Society’s AI Working Group, with valuable input from evaluators, commissioners, researchers and practitioners across our community.
This presentation has been designed to support the accompanying guidance by showing how the principles can be applied and scaled across different levels of risk. It includes explicit examples of when a project should not proceed, as well as a practical checklist for evaluators to use in their own work.
This is the slide deck from the live launch event, including scenarios, limitations, and checklist.
This includes priorities identified by participants through Q&A and a live poll, as well as links and resources shared in the chat by attendees.