liveBook · Manning

Trust

As explained in How Large Language Models Work

Confidence that a system or party is reliable, competent, honest, and aligned with expected interests.

Trustworthiness means confidence that a system is reliable, competent, honest, and aligned with expected interests. Fluent explanations alone cannot establish it: people may trust an AI simply because an explanation appears, even when that explanation is inaccurate. section 8.4

Trustworthy design therefore requires careful evaluation of assumptions about how people and LLMs interact. Transparency, aligned incentives, and feedback cycles matter more than treating explainable AI as a complete answer. section 8.4 section 8.0

Important cautions include:

  • Human review does not automatically prevent errors because automation bias can influence judgment. section 8.0
  • Explanations can increase misplaced confidence rather than reveal whether an output is correct. section 8.4

How How Large Language Models Work covers it

How deep it goes
Distinguishes explainability from trustworthy design and identifies overtrust, automation bias, transparency, incentives, and feedback.
Examples
Uses an LLM-based tech support system and user reactions to apparently explanatory outputs.
What you can do after
Evaluate whether explanations, human checks, transparency, incentives, and feedback genuinely support reliable LLM use.
How it connects
Connects trustworthiness with explainable AI, transparency, automation bias, embeddings, and classical machine learning.
Cover of How Large Language Models Work
liveBookmanning.com
Read How Large Language Models Work on liveBook Get the book at manning.com

Also explained in

Compare how 14 books explain trust →

Related concepts

This explainer was generated by Manning's knowledge graph from the text of How Large Language Models Work. The section links point to the parts of the book it draws on.