Belief Calibration

How AI systems track epistemic states and calibrate confidence against external feedback.

Knowledge States

Click a belief to see its epistemic context

Select a belief to view its epistemic context

Why Belief Calibration Matters

Well-calibrated AI systems know what they know and what they don't. VCP tracks:

  • Confidence levels — How certain the system is about each claim
  • Evidence sources — Where the belief comes from (training, inference, external lookup)
  • Calibration history — How well past confidence matched reality
  • Uncertainty type — Whether the uncertainty is resolvable or fundamental

The ? marker indicates introspective uncertainty—claims about internal states that cannot be fully verified from inside the system.