# Mind as Model — Cognitive Science Map (Phase II)

**Workplan:** SOUL-WP-0003-T01  
**Date:** 2026-08-12  
**Purpose:** Map predictive processing, active inference, internal models, and
metacognition against Soul Frame **Mind = Model**.

Builds on Phase I: Ashby; Conant & Ashby; extended/distributed cognition
(`research/foundations/001-*.md`).

---

## 1. Soul Frame target definition

> Mind is the adaptive modeling and regulatory dimension of a Being: models of
> self, others, environment, and relationships guide action and are revised by
> interaction.

Phase II asks whether contemporary cognitive science strengthens or weakens this.

---

## 2. Source crosswalk

| Source | Core idea | Agreement with Soul Frame | Tension |
|---|---|---|---|
| **Ashby 1956; Conant & Ashby 1970** | Regulation requires a model of the regulated system | Direct support for Mind = Model (H3) | Classical cybernetics is not social/personhood theory |
| **Friston / Active Inference** | Agents minimize free energy via generative models of sensory causes; action and perception as inference | Strong formalization of adaptive modeling + control as one package | Philosophical over-extension debates; Markov blanket ≠ full Being (RN-002) |
| **Clark, *Surfing Uncertainty* (2016) and related PP** | Brain as prediction machine; perception as controlled hallucination; action as prediction error minimization | Mind as hierarchical generative modeling fits “model-based adaptation” | Risk of reducing Mind to a single neural story; Soul Frame needs multi-substrate models |
| **Hohwy, *The Predictive Mind* (2013)** | Systematic PP account of mind | Supports internal generative models as central to cognition | Stronger internalism than extended-mind / distributed views |
| **Metacognition literature** (overview) | Models of own cognitive processes, uncertainty, confidence | Supports Self-Model and calibration of Other-Models | Not sufficient alone for Social Mind |
| **Baker, Saxe & Tenenbaum 2009** | Inverse planning: infer goals/beliefs from action | Other-Model as inverse inference | Rational-agent assumptions |
| **Enactive critiques** (e.g. participatory sense-making; radical enactivism) | Interaction can constitute cognition; caution on representation talk | Interaction process is first-class (RN-006) | May reject “model” vocabulary; Soul Frame must keep “model” inclusive |

**Done criterion (≥4 rigorous sources with agreements/tensions):** met.

---

## 3. Working synthesis for Mind

```text
Mind  ⊇  generative / regulatory models
      +  update from evidence (including social evidence)
      +  action selection guided by those models
      +  (for Social Mind) coupled self/other/relation models
```

**Not required for Mind in Soul Frame:**

- solved consciousness,
- biological brain,
- single substrate location,
- perfect accuracy of models.

**Quality of Mind** (from RN-006, generalized):

- calibration (uncertainty),
- revisability,
- differentiation (distinct counterparts),
- perspective separation (mine vs yours vs yours-of-mine),
- not accuracy alone when models are causal in interaction.

---

## 4. Metacognition placement

Metacognition maps primarily to:

- **SM2 Self-Model** (including confidence about own knowledge),
- **SM4 Reflected-Self Model** (what others model about me),
- model-quality monitoring (calibration).

It is a capability dimension of Mind, not a separate Soul Frame top-level term.

---

## 5. Implications

1. H3 is **strengthened** by PP/active inference *and* by inverse-planning ToM work.
2. “Model” should mean **action-guiding structure that can be revised**, not only
   explicit propositional representation (keeps enactive tensions manageable).
3. Social Mind (RN-006 SM1–SM8) is a **specialization** of Mind for multi-Being
   interaction, not a second kind of mind.

## 6. Bibliography pointers

See `sources/bibliography.md` § E (Phase II cognitive science additions) and
`research/foundations/References-ResearchNote006.md`.