Genuinely Probabilistic (P)
Task nature where no single deterministic correct answer exists
Genuinely probabilistic (P) is a task-nature class: no single deterministic correct answer exists before the fact. The task requires judgment under uncertainty — world-side openness, method-side inference, or both (When Trustworthiness Is Not Enough).
Examples: medical diagnosis on ambiguous presentation, fraud scoring, demand forecasting, content relevance, legal reasoning under underdetermined evidence.
P is not “hard D-F.” Complexity does not make a task probabilistic. Sorting a million records is complex and D-F. “Is this email spam?” sounds simple and is usually P.
Organisational implications:
- Reinforcement mechanisms (Strengthening the Commitment Boundary — blind review, stress testing) apply where outcome severity warrants — not as votes or credentials to discover a hidden deterministic answer. Meritocracy is deferred to Meritocracy — The Sweet Spot.
- LLMs and ML often sit in P cells; trust is set by the epistemic harness, not the naked model (LLMs at the Commitment Boundary).
- Distinguish P from false uncertainty — stakeholders who lack a knowable chain experience openness that is epistemically closed.
Sibling classes: D-F, D-I. Full matrix: Decision Framework map.