Clarity of Capability, Limits, and Confidence
Make it clear what the AI can do, what it cannot do, and how certain it is.
Clarity of Capability, Limits, and Confidence
What is Core concept
The AI should clearly communicate what it can do, what it cannot do, and how certain it is. Users should not have to guess whether the system is reliable, experimental, or likely to be wrong. Your doc directly covers expectation-setting, confidence communication, and avoiding misleading certainty.
Why It Matters
AI systems are probabilistic, so they can sound confident even when they are wrong. Clear capability and confidence signals help users make better decisions, apply appropriate trust, and know when to verify. Public human-AI guidance also emphasizes setting expectations, making limitations visible, and helping users understand uncertainty.
Risks if Ignored
- Users overtrust incorrect answers
- Users undertrust useful outputs because the system feels opaque
- Hallucinations look authoritative
- High-stakes mistakes become more likely
- Trust drops sharply after visible failures
Your document also points out that raw confidence percentages can mislead users if they are shown without context.
How to Evaluate (Checklist)
- Does the product explain what the AI is good at?
- Are important limits or failure conditions made visible?
- Is uncertainty communicated in a human-friendly way?
- Are users prompted to verify when confidence is low?
- Does the wording avoid false certainty?
- Are there cues for when output should be reviewed more carefully?
Real Examples
- Perplexity-style citations and source-backed answers help users judge whether an answer is well-supported instead of blindly trusting fluent text.
- Microsoft human-AI guidance stresses that systems should set expectations and reveal uncertainty so people can use them appropriately.
- In your doc, confidence display, explainability, and trust-related guidance are direct examples of this principle in practice.
Related Patterns
- Expectation-setting disclosure
- Confidence and uncertainty communication
- Process visibility
- Source and provenance transparency
- Suggested prompts.