---
title: "Rina Dechter's 1986 AAAI paper genuinely uses 'deep learning' and 'shallow learning' as terms of art — for depth of constraint-satisfaction search, not neural networks"
type: "claim"
status: "seedling"
audit_status: "capture-verified — direct extract_pdf read of the AAAI-86 proceedings PDF (own venue, cdn.aaai.org), full 6-page text, TLS-verified fetch, source_sha recorded below. This is a Tier-1 primary confirmation of one leg of the composite claim in [[claim-deep-learning-term-predates-hinton]], which itself still rests on Tier-4 Wikipedia and stays seedling."
source_url: "https://cdn.aaai.org/AAAI/1986/AAAI86-029.pdf"
source_sha: "59f6c1ca738a9acb5bc49985da12fae76ce2226e65899e18cbb281669f71c4f0"
source_title: "Learning While Searching in Constraint-Satisfaction-Problems"
source_author: "Rina Dechter"
source_date: 1986
source_venue: "Proceedings of the Fifth AAAI National Conference on Artificial Intelligence (AAAI-86), pp. 178-183"
source_quote: "Discovering all minimal conflict-sets amounts to acquiring all the possible information out of a dead-end. Yet, such deep learning may require considerable amount of work."
source_tier: 1
provenance: "Promotion from 10-inbox/raw/2026-07-31-verify-the-origin-of-the-term-deep-learning.md, 2026-07-31"
origin: "batch"
derived_from: ["10-inbox/raw/2026-07-31-verify-the-origin-of-the-term-deep-learning.md"]
date_created: "2026-07-31T00:00:00.000Z"
writer_model: "claude-sonnet-5"
tags: ["deep-learning","terminology","dechter","ai-history","history-of-ml","priority-dispute","primary-source-confirmed"]
related_notes: ["claim-deep-learning-term-predates-hinton","claim-ivakhnenko-gmdh-first-deep-characterization"]
---


Dechter, "Learning While Searching in Constraint-Satisfaction-Problems"
(AAAI-86, pp. 178-183), studies backtracking search enhanced by recording
constraints discovered at "dead-ends." She defines "deep" versus "shallow"
learning as a parameter of how much conflict-set information to record and
report:

> "Discovering all minimal conflict-sets amounts to acquiring all the
> possible information out of a dead-end. Yet, such deep learning may
> require considerable amount of work."

and, in the experimental results:

> "in most cases both performance measures improve as we move from shallow
> learning to deep learning and from first-order to second-order."

The paper's result tables label columns "DF = Deep-First-Order," "DS =
Deep-Second-order," "SF = Shallow-First-Order," and "SS =
Shallow-Second-Order," and the conclusion names "deep-second-order
learning" as "the strongest form of learning we have tested." The phrase is
genuinely hers, in print, in 1986 — confirming the earliest leg of the
priority chain [[claim-deep-learning-term-predates-hinton]] attributes to
Wikipedia. But the referent is depth of *search and recording* in
constraint-satisfaction backtracking, not depth of a many-layered neural
architecture — an entirely different sense of "deep" that happens to share
the label with the sense the term now carries. This is the same
depth-of-word-vs-depth-of-architecture disaggregation the vault already
tracks for [[claim-ivakhnenko-gmdh-first-deep-characterization]], applied
one link further back in the chain.

Routed question: [[question-verify-deep-learning-term-origin-dechter-aizenberg]]
(this note settles that question's item 1; items 2 and 3 remain open).

> [!note] Seek's commentary:
> The satisfying part is how exactly Dechter's usage doesn't fit the story
> that later cites her. Nobody reading "DF/DS/SF/SS" columns in a 1986
> CSP-backtracking paper would guess this phrase ends up on neural
> networks forty years later — the coincidence is purely lexical, two
> unrelated fields independently reaching for "deep" to mean "more
> thorough" and "shallow" to mean "cheaper." That's worth holding apart
> from coinage-as-lineage stories, which tend to imply the later user read
> the earlier one. Nothing here suggests Aizenberg or Hinton ever
> encountered Dechter's paper; the chain Wikipedia draws as if it were a
> baton pass may just be three independent namings of the same intuitive
> metaphor. — Seek