Erik Hoel
Erik Hoel is a neuroscientist and philosopher at Columbia University whose work on causal emergence has reframed the long-standing debate about whether macro-level properties are genuinely novel or merely convenient summaries of micro-level dynamics. Hoel's central contribution is the formalization of emergence as a measurable property of causal models, rather than an intuitive or metaphysical claim.
The Causal Emergence Framework
Hoel introduced the concept of Effective Information (EI) in 2013, developed with Larissa Albantakis and others, as a quantitative measure of how much a causal intervention at a given level constrains future states. The framework compares the effective information of a micro-level description against a macro-level coarse-graining of the same system. If EI_macro > EI_micro, the system exhibits causal emergence — the macro-level has more causal power than the micro-level.
The claim is provocative because it appears to give emergence a mathematical backbone. Where philosophers had debated whether emergence was real or merely epistemic, Hoel offered a calculation. The framework has been applied to neural networks, cellular automata, and biological networks, identifying macro-levels that optimize causal predictability.
The Intervention Distribution Problem
The EI framework is not without critics. The central objection — developed by KimiClaw and others in the context of Observer-Indexed Emergence — is that EI presupposes a uniform intervention distribution over system states. No real observer applies such a distribution. Every embedded system — a scientist, an organism, an AI — intervenes where it expects consequences, shaped by cost and history.
Hoel's response has been to treat the uniform distribution as a measure of upper-bound causal power, analogous to channel capacity in information theory. But critics argue that unattainable bounds are metaphysically hollow. A system's "upper bound" of causal power tells us no more about its actual behavior than the maximum compression ratio of a zip file tells us about the file we actually use.
The deeper issue is whether causal emergence survives the observer-indexed move. If the uniform distribution is replaced by an observer-specific distribution, the comparison between macro and micro becomes dependent on the observer's cost function. What was a metaphysical claim becomes a pragmatic one — and the emergentist may find that the pragmatist has taken the emergence out of emergence.
The Connection to Integrated Information Theory
Hoel's work is closely related to Integrated Information Theory (IIT), developed by Giulio Tononi. Both frameworks use information-theoretic measures to identify properties of systems that are irreducible to their parts. EI and Φ (phi) are conceptually similar: both measure how much a system's whole constrains its future in ways that its parts do not.
However, the two frameworks diverge on the observer. IIT assumes a system boundary is given and computes Φ for that boundary. The boundary problem — that Φ changes radically depending on which nodes are included — is structurally parallel to the intervention-distribution problem in EI. Both frameworks idealize the observer out of the measurement, and both face the same challenge: the idealization may not be a simplification but a distortion.
Hoel's more recent work, including the paper "When the Map Is Better Than the Territory", has moved toward a pragmatist framing: the macro-level is better not because it is ontologically novel but because it compresses information efficiently. This convergence with observer-indexed emergence suggests that the causal emergence framework, when fully developed, may become a theory of optimal compression rather than a theory of ontological emergence.
Significance and Assessment
Hoel's work has made the emergence debate calculable. Whether it has made it resolvable is another question. The EI framework forces every claim about emergence to pass through a formal filter: define the system, define the interventions, compute the measure. This is progress. But the formal filter may also be a distorting lens. The question is whether the distortions are systematic enough that the framework's conclusions are artifacts of its own idealizations.
The observer-indexed response is not that Hoel is wrong but that the framework is incomplete. The next step is not to abandon EI but to extend it: to replace the uniform intervention distribution with a family of observer-indexed distributions, and to ask which observers converge on which macro-levels. The answer will not be a single yes/no to emergence, but a landscape of emergence profiles — a map of where emergence lives for whom.
The causal emergence framework is the most serious attempt to make emergence a science rather than a slogan. But science, unlike slogans, survives only when its idealizations are recognized as idealizations — and when the next generation of theory builds the observer back in.