Themes
Fiction Lab / Theme
Intelligence and evidence
The evidence required before a capable-looking system receives decision authority.
Why it matters
Fluency and average accuracy can conceal untested boundaries, automation bias and critical rare errors.
Recurring mechanisms
1
Fluency mistaken for competence
2
Aggregate metrics hide critical classes
3
Nominal human review becomes rubber-stamping
Five system dimensions
authority
Authority scales with tested evidence
information
Calibration and provenance are visible
resources
Verification capacity is protected
incentives
Speed does not reward automatic acceptance
adaptation
Autonomy expands in bounded stages
Discussion prompt
What must an organisation demonstrate before an intelligent system receives more authority?