Measure & Improve ยท MON-06
User feedback
Feedback from people who use AI and people affected by it is gathered on a regular basis, analysed, and acted on.
Why it matters
Users and affected people see problems that measures miss, such as outputs that are technically correct but unhelpful or confusing. Acting visibly on their feedback also improves trust and take-up.
Levels
| 0 Absent | Confirmed absent: the evidence shows nothing is in place. Not known is recorded as insufficient evidence, not 0. |
|---|---|
| 1 Initial | Feedback is not sought; what arrives comes through general complaints. |
| 2 Developing | Feedback is gathered for some uses, usually at launch, but is not analysed or acted on consistently. |
| 3 Defined | Every live use of AI has a feedback route for users and affected people, with a named owner who reviews it on a set cycle and records action. |
| 4 Managed | Feedback trends are measured and reported alongside performance data, and recurring issues lead to tracked changes. |
| 5 Optimising | Feedback is built into the service, and the way it is gathered and used is improved from evidence of what leads to useful change. |
Level 3 criteria
Level 3 is met when every applicable criterion is met. Criteria marked with an overlay apply only to organisations with that characteristic (overlays).
- MON-06.1 Every live use of AI has a way for users and affected people to give feedback.
- MON-06.2 A named owner reviews the feedback for each use on a set cycle.
- MON-06.3 The action taken on feedback is recorded.
Assessment questions
- How can users and affected people give feedback on a use of AI?
- Who reviews feedback, and how often?
- Can you show a change made because of feedback?
- Feedback channels and a sample of feedback received
- Records of feedback reviews and resulting changes
How to reach Level 3
- Give every live use of AI a simple feedback route for users and affected people, such as a form, a button or a contact point.
- Name an owner for feedback on each use and set how often they review it.
- Record each review and the action taken, including where no action was needed and why.
Common pitfalls
- Collecting feedback only from staff users and not from the people affected by outputs.
- Collecting feedback at launch and then stopping.
- Treating feedback as a route to challenge a decision, which belongs under RSK-08.
Starter artefacts
- Feedback route for each use
- Feedback review and action log
Ownership and sequence
- Accountable
- business owner
- Contributors
- delivery lead
- Depends on
- GOV-05
- Roadmap step
- 5
- Lifecycle
- deploy, operate
Cross-references
| Algorithmic Transparency Recording Standard | Tier 1 Summary information partial |
|---|---|
| ISO/IEC 42001 | A.8.3 partial; A.10.4 partial |
| JSP 936 Part 1 | para 51 partial |
| MITRE AI Maturity Model | P6.2 partial; P6.4 partial |
| NIST AI 600-1 | R7 partial |
| NIST AI RMF 1.0 | GOVERN 5.1 partial; GOVERN 5.2 direct; MAP 5.2 direct; MEASURE 1.3 partial; MEASURE 3.3 partial; MEASURE 4.2 partial; MEASURE 4.3 partial; MANAGE 4.1 partial; MANAGE 4.2 partial |
| UK Data and AI Ethics Framework | Accountability 7 direct; Fairness 6 direct; Societal impact 4 direct; SAT 2.7 partial; SAT 4.3 direct; SAT 4.4 direct |
| UNESCO AI Maturity Framework | 2.4 partial |
direct: this control at Level 3 meets the item. partial: it covers part of it.