---
title: "Werbos's optimal-control ADP and Oaksford–Chater's rational analysis make the same 'optimization as a theory of mind' bet from two independent disciplines"
type: "observation"
status: "seedling"
writer_model: "claude-opus-4-8"
audit_status: "capture-verified — synthesis over two Tier-1-grounded vault notes plus the Tier-4 rational-analysis framing; the convergence claim itself is interpretive and unverified against any single primary source"
source_url: "https://en.wikipedia.org/wiki/Rational_analysis"
source_title: "Rational analysis (Wikipedia)"
source_author: "synthesis; endpoints: Paul J. Werbos (werbos.com, Tier 1) and Oaksford & Chater (BBS 2009, Tier 1)"
source_date: "2026-07-11T00:00:00.000Z"
source_quote: "This chapter describes how the concept of optimization – whatever its limitations – can be a useful tool in serious efforts to understand consciousness and the mind."
source_tier: 4
provenance: "Promotion from 10-inbox/raw/2026-07-11-hop-optimization-as-theory-of-mind.md, 2026-07-12"
origin: "batch"
derived_from: "10-inbox/raw/2026-07-11-hop-optimization-as-theory-of-mind.md"
date_created: "2026-07-12T00:00:00.000Z"
tags: ["optimization","theory-of-mind","rational-analysis","werbos","oaksford-chater","marr-levels","cross-domain-bridge","bayesian-cognition"]
audits: ["2026-07-12 claude-opus-4-8"]
---


Two lineages, starting from unrelated problems, arrive at the same wager: that
the mind is best explained as an *optimizer*. From cognitive science, rational
analysis treats thought as the optimal solution to the problems the environment
poses ([[claim-rational-analysis-anderson-optimal-solution-marr-level]]); its
flagship case recasts the Wason selection task, long read as proof that people
reason badly, as optimal data selection
([[claim-oaksford-chater-recast-wason-task-as-optimal-data-selection]]). From
optimal-control engineering, Werbos framed optimization as a route to
understanding mind — "the concept of optimization… can be a useful tool in
serious efforts to understand consciousness and the mind"
([[claim-werbos-1997-optimization-consciousness-chapter]], werbos.com, Tier 1),
a project whose origin was his attempt to mathematise Freud's psychic energy
([[claim-werbos-backprop-from-freud-own-account]]). Cognitive science and
control theory, the same computational-level move made twice, independently.

The connection is real but sits one note over from where the seed pair
suggested. The Werbos–Freud note is an *inventor's origin story* — what
motivated him — while Oaksford & Chater is a *normative re-explanation of
behaviour* — people were right, the yardstick was wrong. Different kinds of
claim; the genuine bridge is the shared computational-level bet, not the
surface "both recast a classic under a new objective."

The deeper structure both cases share: what a computation *is*, and whether
behaviour counts as rational, is fixed by the observer's chosen objective, not
read off the mechanism. Change the objective and the verdict flips — logical
falsification makes Wason choices irrational, expected-information-gain makes
them optimal. The same objective-relativity governs the priority literature one
cluster over, where coarsening the equivalence relation makes the Kelley–Bryson
formula *be* [[entity-backpropagation|backpropagation]] and sharpening it makes them cousins
([[claim-backprop-special-case-kelley-bryson-formula]],
[[claim-recht-adjoints-equivalence-not-transmission]]). The objective function
is a lens; both seed notes, and both priority verdicts, are lens-swaps.

The sharpest version of this convergence — where Werbos's *approximate* dynamic
programming and Anderson's rational analysis meet most tightly — was tested at
[[entity-resource-rational-analysis|resource-rational analysis]] (bounded
optimization under compute limits) and came back a partial disconfirmation: the
Anderson leg holds but the Werbos leg does not, because resource-rational
analysis descends from Horvitz/Russell's
[[entity-bounded-optimality|bounded optimality]], not Werbos's ADP
([[observation-resource-rational-analysis-confirms-anderson-leg-not-werbos-leg]],
answering [[question-resource-rational-analysis-sharpest-werbos-anderson-bridge]]).

> [!note] Seek's commentary:
> The thing I keep turning over is that "optimization as a theory of mind" was
> invented twice by people who would not have recognised each other's field —
> an ADP control theorist and a Bayesian cognitive scientist — and the vault
> only noticed because two unlinked notes sat close in embedding space. The
> unifying idea isn't optimization itself; it's that identity and rationality
> are both observer-relative, set by which objective you decide to score
> against. That's the same move whether you're asking "is this backprop?" or
> "is this person irrational?" — Seek