Skip to content

Animism as a Generator Prior (Framing Note)

Status: Working Note

Scope: Animism read as a stance on the inverse problem — a prior over generators that resolves trace→generator ambiguity toward agents. A claim about minds and inference, explicitly not about the cosmos, and explicitly not a one-paragraph theory of indigenous lifeworlds.

Epistemic status: [HYPOTHESIZED] framing over established cognitive-science anchors (agency detection, anthropomorphism theory, the intentional stance). The connection to this repository's measured Occam results is the note's contribution.

Related files:

  • trace-to-generator.md
  • ../identity/psychedelics-as-perturbation.md
  • ../narrative/art-science-one-practice-two-referees.md
  • ../../lab/benchmarks/inverse-reconstruction/README.md

Failure conditions:

  • Sliding from "animism is a prior in minds" to "the universe is animate." The first is testable cognitive science; the second is a narrative-layer statement and must be tagged as such wherever it appears.
  • Treating animism only as error. A prior is not a bug; its value is world-dependent — that is this note's actual point.
  • Colonizing anthropology: lived animist traditions are relational practices and ethics, not merely inference engines. This note addresses one face — the inferential one — and stops there.

The claim

Every observer facing a trace must select from the equivalence class of generators consistent with it. Selection requires a prior. The family-search testbed measured one such prior — simplicity — and found its value is world-dependent: reliable where the world is simple, chance where it is uniform, systematically wrong where it is complex.

Animism is another prior over the same class: the agent prior [HYPOTHESIZED]. Where Occam resolves ambiguity toward the simplest consistent generator, animism resolves it toward an intentional one — a generator with goals, perspective, and the capacity to respond. The rustle in the grass becomes a predator; the storm becomes an actor; the river becomes someone. Animism, on this reading, is not a primitive mistake about the world but a systematic answer to the inverse problem, applied with maximal scope.

The anchors

  • Agency detection (Barrett's HADD, 2004): humans run a hyperactive agent detector because the error costs are asymmetric — a false positive (fleeing a rustle that was wind) is cheap; a false negative (ignoring a rustle that was a leopard) is fatal. The detector is tuned hot by design, not by failure.
  • Anthropomorphism as strategy (Guthrie, Faces in the Clouds, 1993): perceiving agents is the highest-stakes bet available to a perceiver — if right, it captures the most consequential structure in the environment; if wrong, it usually costs little.
  • The intentional stance (Dennett, 1987): treating a system as an agent is a predictive strategy, justified by its compression payoff, not by metaphysics. Animism is the intentional stance with its scope dial turned to maximum.

In the benchmark's terms: the agent prior pays in worlds dense with agents — and the human world is dense with agents (other people are true positives, and they are the most important generators anyone ever reconstructs). It over-extends exactly where the Occam prior under-extends: on traces produced by blind dynamics that look purposive. Emergence is the great generator of such traces — flocks, markets, storms — which is why an emergence-rich world keeps animism permanently plausible: self-organization is what agency looks like from outside, before the inverse problem is solved.

Two mirrors

  • Morelli, reversed. Connoisseurship (Art and Science) is disciplined generator-attribution — querying the incidental traces where the true hand leaks through. Animism is the same inference with the threshold removed: attribution without the discipline of divergence queries. Same machinery, different calibration — which is why the cure for over-attribution is not less inference but better queries.
  • The antenna, internalized. The psychedelics note describes mental events arriving untagged when the self-model's authorship-binding relaxes — and untagged traces trigger exactly this prior: a generator is inferred, and since it isn't me, it must be someone. Brain-as-antenna cosmologies are the agent prior applied to one's own thoughts, displaced outward. Animism begins at home.

The boundary

As a claim about minds — that humans carry a hot agent prior, that it is ecologically rational, that its failure modes are measurable — this is ordinary, well-anchored cognitive science, and the repo's machinery adds a precise frame for it (priors over equivalence classes; world-dependent payoff; divergence queries as the calibration instrument).

As a claim about the world — that rivers, storms, or the universe are in fact agents or conscious — it is a different kind of statement: not falsifiable from traces alone (the equivalence class never collapses without divergence queries, and the cosmos affords none), and therefore at home only in the narrative layer, where it has always done its real work: as relational practice, as ethics, as story. That sorting is not a demotion; it is the assignment of the correct referee.

Related work. Barrett (2004), Why Would Anyone Believe in God? (HADD); Guthrie (1993), Faces in the Clouds; Dennett (1987), The Intentional Stance. Mapping in the Related Work Map.