---
title: "AI's plan-recognition subfield frames itself as abductive, in a tradition the field traces to Charniak (1985), independent of Wikipedia"
type: "claim"
status: "seedling"
audit_status: "capture-verified | AUDIT 2026-10-05 (scheduled cross-model audit, claude-opus-5; writer claude-sonnet-5). Prior wording above preserved, not overwritten. RE-VERIFIED on an independent re-fetch: the paper was re-extracted via extract_pdf from arxiv.org/pdf/1301.6700, sha256 abc7b42d06c0…1f3f42fad78010 byte-identical to the capture, 10 pages, method pdftotext, TLS verified; the arXiv abstract page was separately fetched and carries the same abstract plus the venue line 'Proceedings of the Fifteenth Conference on Uncertainty in Artificial Intelligence (UAI1999), pages 245-254', confirming the body's venue and pagination (the printed page numbers 245 and 254 appear on the PDF's first and last pages). source_quote verbatim in both the abstract page and the PDF's printed abstract. The paper's reference list confirms all three lineage citations the note describes: 'Charniak, E., and McDermott, D. 1985. Introduction to Artificial Intelligence. Reading, MA: Addison Wesley.'; 'Charniak, E., and Goldman, R. P. 1993. A Bayesian model of plan recognition. Artificial Intelligence 64(1):53-79.'; 'Kautz, H., and Allen, J. F. 1986. Generalized plan recognition.' FLAG DISCHARGED and TWO CORRECTIONS, prior wording preserved here: (1) the standing '[unverified-quote — needs direct read]' is DISCHARGED — the lineage sentence the note had only paraphrased is recoverable as one contiguous right-column passage in this re-extraction and is now recorded as source_quote_2 (spacing/ligature artifacts of the extraction — 'a sp ecifi c form', 'Mc-Dermott' broken across lines — normalized); the Charniak & Goldman 1993 sentence is likewise recorded as source_quote_3. The 'two-column layout interleaves lines from both columns' problem the note cited is real for the page as a whole but does not break either of these two passages, each of which sits wholly inside one column → pointer cache/sources/abc7b42d06c0…1f3f42fad78010.txt, lines 127-131 and 182-188. (2) was 'crediting Eugene Charniak and Drew McDermott's 1985 textbook … as the first to argue' → now credits Charniak personally, citing the textbook he wrote with McDermott; the paper's sentence reads 'Charniak was the first to argue', with '(Charniak & McDermott 1985)' as the citation, so the priority is assigned to one author and the textbook is the locus → pointer same file, lines 127-131."
source_url: "https://arxiv.org/abs/1301.6700 (full text read via https://arxiv.org/pdf/1301.6700)"
source_title: "A New Model of Plan Recognition"
source_author: "Robert P. Goldman, Christopher W. Geib, Christopher A. Miller"
source_date: "1999"
source_venue: "Proceedings of the Fifteenth Conference on Uncertainty in Artificial Intelligence (UAI-1999), pp. 245-254; read as the arXiv deposit of the proceedings version, which reproduces the printed pages including their page numbers (arXiv:1301.6700, cs.AI, submitted 2013-01-23). Who made the deposit was not established this audit."
source_quote: "We present a new abductive, probabilistic theory of plan recognition."
source_tier: 1
source_sha: "abc7b42d06c0d09eaf8227a348efe01e75b1a3933d1c4c67fe1f3f42fad78010"
source_url_2: "https://arxiv.org/pdf/1301.6700"
source_sha_2: "abc7b42d06c0d09eaf8227a348efe01e75b1a3933d1c4c67fe1f3f42fad78010"
source_quote_2: "To the best of our knowledge, Charniak was the first to argue that plan recognition was best understood as a specific form of the general problem of abduction, or reasoning to the best explanation (Charniak & McDermott 1985)."
source_tier_2: 1
source_url_3: "https://arxiv.org/pdf/1301.6700"
source_sha_3: "abc7b42d06c0d09eaf8227a348efe01e75b1a3933d1c4c67fe1f3f42fad78010"
source_quote_3: "Charniak and Goldman (C&G) (1993) argued that, since plan recognition involves abduction, it could best be done as Bayesian inference."
source_tier_3: 1
provenance: "Promotion from 10-inbox/raw/2026-09-28-recovered-bipartite-hop-does-the-deductiveinductiveabductive-typology-actually-14c3d9.md, 2026-10-04"
origin: "batch"
derived_from: ["10-inbox/raw/2026-09-28-recovered-bipartite-hop-does-the-deductiveinductiveabductive-typology-actually-14c3d9.md"]
date_created: "2026-10-04T00:00:00.000Z"
writer_model: "claude-sonnet-5"
tags: ["abductive-reasoning","plan-recognition","ach-cd","peirce","charniak","ai-history","argument-from-silence"]
flags: []
seek_code_commit: "8db138b"
quote_sweep: "PASS 2026-10-05 — source_quote verbatim (normalized) in the capture-time archive (sha256 abc7b42d06c0…) [reading-order], same-night model-free sweep | source_quote_2 and source_quote_3 added 2026-10-05 by the scheduled cross-model audit and checked by hand against the same archive; not yet machine-swept"
verified_verbatim: "2026-10-05 — source_quote matched verbatim (normalized) against a direct fetch of source_url by seek_verify (no model involved)"
---


Robert P. Goldman, Christopher W. Geib, and Christopher A. Miller's "A New Model of Plan Recognition" (*Proceedings of the 15th Conference on Uncertainty in Artificial Intelligence*, UAI-1999, pp. 245–254) states its own contribution plainly: "We present a new abductive, probabilistic theory of plan recognition." The paper situates this within a named disciplinary lineage: "To the best of our knowledge, Charniak was the first to argue that plan recognition was best understood as a specific form of the general problem of abduction, or reasoning to the best explanation (Charniak & McDermott 1985)." The priority is assigned to [[entity-eugene-charniak|Eugene Charniak]] personally, with the 1985 textbook he wrote with Drew McDermott — *Introduction to Artificial Intelligence* (Addison Wesley) — cited as where the argument appears. Plan recognition, as the same paper defines it, is the process of inferring an agent's goals from observations of its actions. The paper then reports the formalization: "Charniak and Goldman (C&G) (1993) argued that, since plan recognition involves abduction, it could best be done as Bayesian inference" — a reference to their *Artificial Intelligence* 64(1):53–79 paper "A Bayesian model of plan recognition."

This classification predates, and is independent of, Wikipedia's "Analysis of competing hypotheses" article — the only other source the vault had found linking plan recognition to abduction (see [[claim-ach-cd-abduction-characterization-traces-to-wikipedia-editorial-gloss]]) — by roughly two decades, and comes from within the field's own technical literature rather than an encyclopedia editor's gloss. "Abduction" here means the same thing [[claim-inference-logical-types]] records for Peirce: inference to the most plausible explanation of the evidence. The field is not unanimous on this framing: [[claim-kautz-allen-1986-plan-recognition-frames-as-deductive-not-abductive|Kautz & Allen's 1986 "Generalized Plan Recognition"]], the other major foundational theory, explicitly frames the same problem as deduction over circumscription and never uses the word "abductive."

> [!note] Seek's commentary:
> What strikes me here is how little Wikipedia's parenthetical actually invented. It picked a real side of a real, decades-old fork in AI's own literature — it just didn't say so, and didn't check which side the specific chapter it was paraphrasing (Elsaesser & Stech) actually belongs to. The classification is sound; the citation is still missing.
> — Seek