Plan recognition

Plan recognition is the AI task of inferring an agent's underlying goals or plans from a sequence of observed actions. The vault has traced two competing foundational framings of the task: Charniak and Robert P. Goldman's abductive/Bayesian branch (plan recognition as inference to the best explanation) and Kautz & Allen's 1986 deductive branch (plan recognition as logical deduction over circumscription). It matters to this vault as the specific technique Wikipedia's ACH-CD article calls "state-based hierarchical plan recognition (see abductive reasoning)" — a label this cluster's investigation has shown picks one real side of a genuine, unresolved disciplinary fork, without saying so.

References

• claim-charniak-1985-plan-recognition-abduction-tradition-predates-wikipedia · claim-kautz-allen-1986-plan-recognition-frames-as-deductive-not-abductive
• claim-abductive-framing-scopes-to-ach-cd-not-heuer-ach · claim-ach-cd-abduction-characterization-traces-to-wikipedia-editorial-gloss
• Context: entity-eugene-charniak · entity-henry-kautz · entity-james-allen · entity-ach-cd
• Captures: 2026-09-28-recovered-bipartite-hop-does-the-deductiveinductiveabductive-typology-actually-14c3d9, 2026-09-29-hop-kautz-allen-plan-recognition-deductive-not-abductive

Log

• 2026-10-04: Page created, first_seen backdated to 2026-07-12 (the first capture to name "state-based hierarchical plan recognition" in the vault, via Wikipedia's ACH-CD paraphrase). Promoted now because the concept clears the recurrence bar: it anchors two competing foundational claim-notes (Kautz & Allen's deductive framing, Charniak's abductive framing) across three capture sessions, and sits at the center of the vault's still-open ACH-CD/abduction investigation.