Files
blackroad-os-research/papers/agent-architecture/trinary-logic.md
2025-11-23 15:05:09 -06:00

1.4 KiB

Trinary Logic in BlackRoad OS

BlackRoad OS agents reason with three states: +1 (true/affirmed), 0 (unknown/undecided), and -1 (negated/contradictory). This representation keeps uncertainty explicit and allows contradictions to be managed rather than hidden.

Belief Representation

  • Vectors of trinary values: Each proposition maps to {+1, 0, -1} to capture stance and uncertainty.
  • Confidence weights: Optional weights can scale the impact of each value when aggregating across agents.
  • Temporal context: Beliefs can be journaled over time, enabling a PS-SHA∞-backed history of shifts.

Combination Rules

  • AND: Minimum of contributing values, keeping contradictions (-1) dominant over unknowns (0) and truths (+1).
  • OR: Maximum of contributing values, preferring truths (+1) but preserving contradictions when present.
  • Consensus: Weighted sums normalized to the trinary domain to find majority positions while surfacing disagreement.

Operational Patterns

  • Decision gating: Actions require +1 consensus or explicit override when -1 appears.
  • Escalation triggers: Presence of balanced +1 and -1 across critical propositions triggers contradiction events.
  • Logging: Every aggregation result is written as a PS-SHA∞ entry for auditability.

TODOs

  • Extend combination rules with temporal decay to emphasize recent evidence.
  • Define canonical proposition sets for finance, compliance, and platform operations.