nucleation

CLAIM-497: A system's causes are found at the coarsest description whose interventions are unambiguous: a macro description is causally proportionate where it merges only micro states that intervention cannot tell apart, and no coarse-graining carries more causal information than the micro description it supervenes on

3 min read

Properties10
StatusActive
Version2
Rolethesis
Warrantinterpretive
Tagscausality, information-theory, philosophy-of-science
Lineformal-apparatus, distributed-agency
Bears onQUESTION-204 — Does a macro description's effective-information advantage survive when the same intervention distribution over micro states is used at both levels, and does the EI-optimal level agree with an independent interventionist test of which level the effects track?
GroundsTHEORY-1167 — When one intervention distribution over micro states is used at both levels, the effective information of a coarse-graining of a discrete mechanism, with each macro intervention realized by that distribution within its block, cannot exceed the micro effective information; so a positive causal emergence of effective information, scored with a uniform distribution at each level, equals the micro gain from changing the distribution minus a non-negative loss, and is a gain from the change of distribution, THEORY-1169 — For a discrete mechanism scored with one full-support intervention distribution over micro causes at both levels, a coarse-graining of cause and effect keeps all of the micro effective information if and only if it merges only causes with identical rows P(Y | do(x)) and only effects with proportional columns, whatever the distribution; so the coarsest coarse-graining that ties the micro is the causal partition of causal feature learning, with effects grouped by normalized likelihood, and any merge of causes with different effects loses information under every such distribution, LIT-1833 — Multi-Level Cause-Effect Systems, LIT-1834 — Causal feature learning: an overview
Cited as “concludes” byARG-041 — From the effective-information bound and its ties to proportionality: a gain that depends on the intervention distribution is not a property of the system, the coarse-grainings that lose nothing are, and the coarsest of them is where the causes are
Rendered intodocs/arguments-by-conclusion.md

Graph View

The claim

There is a level at which a system’s causes are found, and it can be a macro level. It is the coarsest description in which every macro intervention is unambiguous: every micro state merged into one macro state has the same distribution of effects under intervention. At that level a coarser description loses causal information, and a finer one adds none. This is proportionality, the condition that a cause be neither broader nor narrower than its effect requires.

What the claim gives up is THEORY-787’s further thesis, that a macro description can carry more causal information than the micro and so be where causation emerges. A macro can be the proportionate description. It is never a richer one.

What it rests on

  • The bound. THEORY-1167 (Active) shows that under one intervention distribution over micro states at both levels, a coarse-graining of one transition has at most the micro effective information. This is data processing.
  • The ties. THEORY-1169 (Active) shows that a coarse-graining keeps all of the micro effective information if and only if it merges only causes with identical rows P(Y | do(x)) and only effects with proportional columns. That holds for every full-support distribution. Those are the causal partitions of Chalupka, Perona and Eberhardt (LIT-1833, LIT-1834), with effects grouped by normalized likelihood.
  • The interpretive step. Of the descriptions that lose nothing, the coarsest is the one at which the causes are found. It is the record’s own step, which is why this is a CLAIM and its warrant is interpretive. It takes the coarsest description because a finer one distinguishes micro states that no intervention tells apart. Those distinctions are realization detail, not causal structure.

What it does not say

  • Not that the macro beats the micro. No coarse-graining gains under a common distribution. A positive causal emergence of effective information, scored with a separate uniform distribution at each level, measures the change of distribution (THEORY-1167, THEORY-222).
  • Not that the micro description is wrong. It is true, and it is finer than the causes require.
  • Not how to merge estimated rows. Exact equality is the condition. With data, rows are merged by a threshold, and whether real systems have a stable proportionate level is what defeated_if tests.
  • Not about temporal or spatio-temporal macros, where a macro step spans several micro steps. QUESTION-335 holds that case.
  • Not that only causal partitions are useful. A description chosen for prediction or compression for a declared target can be coarser or finer (QUESTION-252), and normalized measures can rise at a macro (LIT-1843). This claim concerns where causes are, not which description serves a purpose.

Backlinks