KLD Institute
Menu
Course mapOpen study notes

Slide 5 of 24

Reliability

Hallucination is a workflow risk

  • AI may produce plausible but false statements.
  • A confident tone is not evidence.
  • Useful collaboration asks AI to name uncertainty and asks humans to verify.
Weak versus strong

Risk

Weak

Trust confident wording.

Strong

Ask what is uncertain and verify key claims.

Source

Weak

Ask from memory and assume accuracy.

Strong

Give reference material when facts matter.

Decision

Weak

Use polish as proof.

Strong

Separate draft quality from truth quality.