Designing transparent, predictable AI experiences following Microsoft HAX guidelines
Adjust AI behavior and feedback to user goals, skills, and situations.
Avoid human-like cues that inflate expectations or suggest intent or emotions.
Make responsibility, provenance, and decision ownership transparent.
Expose uncertainty so users understand confidence, risk, and limitations of AI outputs.
Let users modify, override, pause, or undo AI actions easily.
Communicate AI errors clearly and offer safe recovery paths.
Clearly show when AI is involved and what it does within the interaction flow.
Ensure users remain the final decision-makers, with AI acting as support.
Explain results in user-facing terms that support decisions, not model internals.
Reduce mental effort with clear, progressive, and concise AI output.
Align messaging with real capabilities to prevent overtrust or misuse.
Design AI workflows that allow continuous human supervision and input.