Questions That Support Trust in AI Systems

Trust is not created by a slogan or a single score. People need to understand what an AI-enabled system does, where its limits are, and who is responsible when something goes wrong.

Make behavior understandable

Explain the system’s purpose, the information it uses, the actions it can take, and when a person should review its output. Communicate uncertainty in a way that helps the user decide what to do next.

Make responsibility visible

Identify who owns the system and how users can report a problem or challenge an outcome. Keep a proportionate record of important actions and decisions, with suitable privacy and access controls.

Check the experience over time

Test whether the system behaves consistently for the tasks and people it is intended to serve. Review meaningful failures and changes to models, data, or tools. These practices can support trust, but they do not guarantee that a system is correct or appropriate.