RAIUC04-BP01 Map the user journey to identify AI interaction requirements
Map the user journey to identify interaction requirements and risks. During pre-interaction, assist users in learning about the system's capabilities and limitations.
During the initial interaction, consider the different accessibility needs of users, and the different sources for key system inputs.
During processing, maintain transparency about AI decision-making and provide appropriate progress indicators. Post-interaction, enable users to understand, challenge, and provide feedback on AI outputs, which keeps the system accountable and improvable.
Consider how different user groups might be affected differently at each stage and implement appropriate safeguards and support mechanisms. If the AI system will be embedded in an existing human-powered workflow, consider what purposes the workflow might address that the AI system does not, and consider the variety of ways in which users might modify system inputs and outputs for other purposes.
Level of risk exposed if this best practice is not established: High
Implementation considerations
-
Identify the expectations users may have and the information they need before engaging with the proposed AI system.
-
Sketch the initial interaction phase. Identify the first touchpoints where users provide input and the guidance they need for effective interaction. Consider how users might alter inputs to influence outputs.
-
Sketch the AI processing phase. Consider how long processing takes, what users see during processing, and what information assists users to understand system activity and confidence levels.
-
Sketch the post-interaction phase. Plan how users receive, interpret, and act on AI outputs, including confidence indicators, explanation features, and guidance for appropriate use of results. Consider the purposes which the outputs might serve.
-
Identify transparency touch points. Mark specific moments in each phase where guidance is most valuable.
Resources
Related documents:
-
ISO/IEC 42001:2023
- User interaction lifecycle implementation -
ISO/IEC 42001:2023
A.5.4 Assessing AI system impact on individuals or groups of individuals -
NIST Artificial Intelligence Risk Management Framework (NIST AI 100-1)
: MAP2.1, MAP2.2, MAP2.3