From Explainability To Behaviour

How AI Systems Behave Over Time

Description:

This short ebook brings together a set of essays examining how AI systems behave when the same judgement is made repeatedly.

Rather than focusing on individual outputs, explanations, or confidence scores, the book looks at behaviour over time. Each chapter explores a familiar signal – stability, confidence, explanation, cue recognition, and agreement – and considers what becomes visible when those signals are observed across repetition rather than in isolation.

The work is intentionally observational. It does not provide methods, metrics, or evaluation procedures. Its purpose is to support careful attention to behaviour before questions of accuracy, assurance, or governance are asked.

This ebook is also included as a free companion book to the Phase 3 Research Report - Agents At Work available here

© 2026 Imogen Hull. Beyond the Average. Shared with subscribers for personal use. Please do not redistribute without permission.