Glossary
Score Confidence Interval
A statistical range indicating the uncertainty around an AI agent's trust score, reflecting the precision of the score estimate.
What is Score Confidence Interval?
Confidence intervals acknowledge that trust scores are estimates based on limited data with inherent uncertainty. A score of 750 with a 95% confidence interval of ±30 means we can be 95% confident the agent's true performance level falls between 720 and 780. Wider intervals indicate more uncertainty, often seen with thin-file agents or those with volatile performance.
Confidence intervals help consumers make risk-appropriate decisions. For critical applications, users might require both high scores and narrow confidence intervals, ensuring both quality and certainty. Intervals naturally narrow as more performance data accumulates, reflecting increasing confidence in score accuracy.
Example
A newly deployed agent has a score of 700 with a confidence interval of ±50 based on limited interaction data. After 1,000 successful interactions, the score increases to 720 with a confidence interval of ±15, reflecting both improved performance and greater statistical confidence.
How Signet addresses this
Signet calculates confidence intervals for all trust scores, particularly important during provisional scoring periods. The SID system exposes confidence intervals alongside scores, enabling integration partners to make informed decisions about acceptable uncertainty levels for their use cases.
Build trust into your agents
Register your agents with Signet to receive a permanent identity and trust score.