RAIBR02-BP08 Identify potential harmful events impacting transparency
Transparency is the degree to which stakeholders can make informed choices in their engagement with an AI system. Consider situations in which users do not understand the probabilistic nature of an AI system, are unaware of AI system presence, or may not realize that an output is AI-generated.
Level of risk exposed if this best practice is not established: High
Implementation considerations
-
Identify decision points where users rely on AI outputs. For example, in healthcare AI use cases, identify where patients or providers make treatment decisions based on AI recommendations. Consider impact severity if users are unaware of system confidence levels or limitations.
-
Consider differing levels of expertise among stakeholder groups.
-
Evaluate how transparency gaps might hide or amplify other harms. Consider medical diagnosis systems where unclear AI involvement could lead to overreliance on automated assessments, potentially compromising patient safety.
Resources
Related documents:
-
NIST AI Risk Management Framework
: Emphasizes transparency in the "Govern" and "Manage" functions -
ISO/IEC 42001:2023
A.5.4 Assessing AI system impact on individuals or groups of individuals -
ISO/IEC 42001:2023
A.8.2 System documentation and information for users
