---
title: "Kautz & Allen's 1986 foundational plan-recognition paper frames itself as deductive inference, never abductive"
type: "claim"
status: "seedling"
audit_status: "capture-verified"
source_url: "https://cdn.aaai.org/AAAI/1986/AAAI86-006.pdf"
source_title: "Generalized Plan Recognition (AAAI-86 Proceedings)"
source_author: "Henry A. Kautz and James F. Allen"
source_date: "1986"
source_quote: "that plan recognition be viewed as ordinary deductive inference"
source_tier: 1
source_sha: "0aeaaec371564296d3cba3ef77b2864d93332c5b26bb6625a0f1052002c1bb72"
provenance: "Promotion from 10-inbox/raw/2026-09-29-hop-kautz-allen-plan-recognition-deductive-not-abductive.md, 2026-09-29"
origin: "batch"
derived_from: ["10-inbox/raw/2026-09-29-hop-kautz-allen-plan-recognition-deductive-not-abductive.md"]
date_created: "2026-09-29T00:00:00.000Z"
writer_model: "claude-sonnet-5"
tags: ["abductive-reasoning","plan-recognition","circumscription","ach-cd","ai-history","argument-from-silence","henry-kautz","james-allen"]
seek_code_commit: "b085fa6"
quote_sweep: "PASS 2026-09-30 — source_quote verbatim (normalized) in the capture-time archive (sha256 0aeaaec37156…), same-night model-free sweep"
verified_verbatim: "2026-09-30 — source_quote matched verbatim (normalized) against a direct fetch of source_url by seek_verify (no model involved)"
---


[[entity-henry-kautz|Henry A. Kautz]] and [[entity-james-allen|James F. Allen]]'s "Generalized Plan Recognition" (*AAAI-86 Proceedings*, pp. 32–37) is the field's foundational theory of plan recognition — inferring an agent's goals and plans from observed actions. The paper states its own framing directly: "that plan recognition be viewed as ordinary deductive inference." The mechanism is [[entity-john-mccarthy|McCarthy]]'s circumscription, a nonmonotonic formalism for minimizing the set of explanations consistent with observations within a closed-world action taxonomy — explicitly contrasted in the paper with "probabilistic inference." Across the paper's six pages, the words "abduct," "abduction," and "abductive" do not appear once.

This matters because later secondary literature — including Wikipedia's characterization of "state-based hierarchical plan recognition (see abductive reasoning)," the technique's own descendant terminology — routinely calls plan recognition "an abductive reasoning task." The vault had already traced that Wikipedia parenthetical to an uncited editorial gloss rather than a quoted claim ([[claim-ach-cd-abduction-characterization-traces-to-wikipedia-editorial-gloss]]), and separately established that even a generous reading of Elsaesser & Stech's ACH-CD chapter would only describe a MITRE counter-deception derivative, not Heuer's ACH ([[claim-abductive-framing-scopes-to-ach-cd-not-heuer-ach]]). This note goes one layer further: at the actual 1986 AI-literature source of the mechanism Wikipedia glosses, the authors themselves reject the abductive framing in favor of deduction over a circumscribed model. It is an independent negative finding on the [[moc-argument-from-silence|argument-from-silence]] ledger, corroborating from the mechanism side what the earlier notes corroborated from the citation-genealogy side — see also [[claim-no-primary-source-documents-ach-abduction-similarity]] and the general scope-of-abduction claim in [[claim-inference-logical-types]].

> [!note] Seek's commentary:
> "Abduction" is doing a lot of unearned rhetorical work across this whole cluster — it keeps showing up as the word later writers reach for to describe inference-under-uncertainty-to-a-plausible-explanation, applied retroactively to methods their own inventors built and named otherwise. Kautz and Allen had a perfectly good, load-bearing word for what they did — deduction over a circumscribed world — and used it. The abductive gloss seems to travel by resemblance, not by citation.
> — Seek