Confidence Levels

The library uses confidence levels to preserve that distinction.

proven

proven means the order is backed by explicit causal evidence.

Examples:

• same-node sequence progression
• known parent-child relationships
• explicit dependency edges

This is the strongest kind of claim the library makes.

derived

derived means the order was inferred from metadata that is useful but not fully causal.

Examples:

• HLC order
• ingestion metadata
• sequence presence without a stronger explicit relationship

This is often operationally useful, but it should not be confused with proof.

fallback

fallback means the library had to impose a deterministic order because the input still needed a stable output shape.

This is a stability tool, not a truth claim.

A fallback answer may be helpful for reproducibility, but it should never be mistaken for strong evidence.

unknown

unknown means the library cannot honestly justify a reliable order.

This is not a failure of the project. It is part of the honesty of the model.

Sometimes the correct answer is not “A before B.” Sometimes the correct answer is:

• we do not know
• we cannot prove this
• these should not be collapsed into a fake certainty

Why This Matters

Without confidence levels, many systems silently upgrade weak signals into strong claims.

causal-order exists partly to stop that upgrade from happening invisibly.