Part VIII · Chapter 80Version 1.0 · 5 November 202631 min read
Explainability in Generative and Agentic AI
Key message
Chains of thought, self-explanations and citations from generative and agentic systems are often plausible without being faithful; trustworthy explanation rests on behavioral tests, logged trajectories and measured source support, not on what the model says about itself.
Five chapters or files are free. This one would be your sixth.
Leave an email address and every chapter and every PDF of the compendium opens, in this browser, for a year. Nothing else is asked, and whether we may contact you is your choice.