h t t p s : / / m e c h a n i s m . u c s d . e d u / b i l l / t e a c h i n g / m o d e l s / c r a i k . p d f

Ingest: Craik, Hypothesis on the Nature of Thought (1943)

Type: kb/sources/types/ingest-report.md

Classification

This book chapter is a conceptual essay: it advances a mechanistic hypothesis through definitions, physical analogies, and possible tests rather than reporting a controlled empirical study. Author: Kenneth J. W. Craik, writing as the named originator of the 1943 hypothesis; the captured chapter supplies no institutional affiliation or systematic evidence review.

Summary

Craik proposes that thought translates an external process into symbols or a working model, transforms that representation through inference, and translates the result back into prediction or action. A useful model need not resemble its target pictorially; it must preserve the relevant relation-structure closely enough to run in parallel with the target. Because the model is cheaper, faster, or safer than the modeled process, an organism can try alternatives involving external reality and its own possible actions before committing. Craik extends this account from mechanical predictors and calculating machines to a tentative hypothesis about neural activity, but he presents that neural claim as a program for testing rather than an established result.

Quotes

No source quotes have been retained yet.

Connections Found

Craik is a historical conceptual anchor for the model-mediated branch of An action model matters only through its consumption path: the model is useful because simulated consequences participate in choosing among actions. His requirement that a model preserve only the relation-structure relevant to the case, together with his warning that every analogy eventually breaks, also supports the transfer boundary in A borrowed pattern transfers only as far as source and target share a mechanism. The chapter is a limitation rather than evidence for the stronger claim that world models assess explanatory-reach through action-conditioned prediction, because it supplies neither a learned predictor nor tests across interventions or shifts. Its reference to an organism's “own possible actions” concerns action on the external world, not a causally connected self-representation, so it does not establish a reflective system. Craik's broad use of “symbolism” also spans words, numbers, neural excitation, and physical models and should not be identified with Commonplace's narrower symbolic representational form.

Extractable Value

  1. Internal trial is a consumption mechanism, not merely a stored description. Craik's translation–inference–retranslation sequence explains how a representation can affect intervention choice by carrying predicted consequences back into action, giving the action-model note a historical conceptual basis. [quick-win]
  2. A model can preserve relations without pictorial resemblance. This sharpens the shared-mechanism test: a proposed transfer should name the target relations that matter, the operations that preserve them, and the aspects deliberately left unmatched. [quick-win]
  3. Cheapness, speed, and safety explain when internal simulation is useful. The comparison with full-scale trials supplies a practical reason to build or retrieve an action model when real interventions are costly, slow, destructive, or irreversible. [just-a-reference]
  4. Model-mediated action is separable from model learning and reflection. The chapter explains using a model to choose among external actions but does not specify evidence-driven revision, retained learning, or a representation of the thinker's own behavior-determining organization. [quick-win]
  5. Craik's “symbolism” is substrate-general historical vocabulary. Keeping it distinct from the KB's operative natural-language, symbolic, and distributed-parametric forms prevents a historical term from collapsing categories that require different inspection methods. [just-a-reference]

Limitations (our opinion)

The central brain-model hypothesis is suggestive but underdetermined. Craik offers analogies, examples, and a small amount of period-specific neurophysiological plausibility, but no controlled comparison that distinguishes internal simulation from simpler associative or reactive accounts. “Similar relation-structure” is explicitly aspect-relative yet lacks a procedure for selecting those aspects, so the hypothesis can become hard to falsify unless a later use names the preserved relations and failure conditions. The chapter establishes neither how models are learned and revised nor that their predictions generalize across shifts. The supplied capture covers only Chapter V, printed pages 51–61, and has substantial two-column OCR interleaving; it is adequate for thematic analysis but not reliable for exact wording or claims about the rest of the book.

Update An action model matters only through its consumption path with a short historical-provenance paragraph citing this ingest as the conceptual anchor for trying action alternatives internally, while preserving the note's stricter requirements for an operative consumption path and behavioral authority.