Skip to content
voiage

voiage.methods.ai_assisted_evidence_triage.value_of_ai_assisted_evidence_triage

value_of_ai_assisted_evidence_triage

Section titled “value_of_ai_assisted_evidence_triage”
value_of_ai_assisted_evidence_triage([positional or keyword] relevance_labels: np.ndarray | list[int] | list[bool] = None, [positional or keyword] triage_scores: np.ndarray | list[float] = None, [positional or keyword] decision_impacts: np.ndarray | list[float] = None, [positional or keyword] reviewer_time_minutes: np.ndarray | list[float] = None, [positional or keyword] extraction_error_rates: np.ndarray | list[float] = None, [keyword-only] triage_threshold: float = 0.5, [keyword-only] audit_sample_rate: float = 0.0, [keyword-only] human_override_rate: float = 0.0, [keyword-only] model_drift: float = 0.0, [keyword-only] automation_cost: float = 0.0, [keyword-only] audit_cost_per_item: float = 0.0, [keyword-only] reviewer_cost_per_minute: float = 1.0) -> AIAssistedEvidenceTriageResult

Estimate the value of automated evidence triage with human oversight.

relevance_labels and triage_scores form a deterministic reference evaluation set. Scores at or above triage_threshold are selected for downstream review. Audit sampling and human overrides recover a bounded fraction of false exclusions and false inclusions, while extraction error, model drift, automation cost, and reviewer time are included in the net decision value. The result remains fixture-backed pending external model validation, licensed evidence corpora, and cross-language parity.

Parameters:

  • relevance_labels np.ndarray | list[int] | list[bool]
  • triage_scores np.ndarray | list[float]
  • decision_impacts np.ndarray | list[float]
  • reviewer_time_minutes np.ndarray | list[float]
  • extraction_error_rates np.ndarray | list[float]
  • triage_threshold float (default: 0.5)
  • audit_sample_rate float (default: 0.0)
  • human_override_rate float (default: 0.0)
  • model_drift float (default: 0.0)
  • automation_cost float (default: 0.0)
  • audit_cost_per_item float (default: 0.0)
  • reviewer_cost_per_minute float (default: 1.0)

Returns: AIAssistedEvidenceTriageResult