Contribution
Autopoiesis and Rosen’s closure to efficient causation had said organisms are organisationally closed, but not which entities close, at what level, or how closure differs from any causal cycle. The paper gives closure an explicit definition in terms of constraints, picked out by time-scale symmetries, and states conditions (direct dependence; strictness) that separate it from a cycle of processes and from looser mutual benefit.
Key insight
Split the causes acting in an organism into what is consumed or transformed (processes) and what shapes the transformation while staying the same at that time scale (constraints). Organisms are the systems whose constraints are themselves produced, at other time scales, by processes other constraints of the same system shape. That mutual production among constraints is closure, and it is a different causal regime from the thermodynamic flow it rides on.
Assumptions
- Symmetry as the criterion of a constraint. Conditions I–II of Definition 1 can be met “after (properly justified) quantitative approximations” (footnote 9): a constraint is only approximately conserved.
- Time scale is decisive; spatial scale and level of description are not, “on a first approximation” (footnotes 15–16 allow exceptions, including strong emergence, left for another paper).
- Constraints are mostly limiting at long time scales and generative at biological ones; this is how the paper reconciles itself with Juarrero.
- Direct dependence is required for closure, justified by the degree of integration of biological systems; “we do not provide in this paper a theoretical justification for this claim” (§4).
- Closure is context-dependent: realised relative to boundary conditions, including external constraints (C5 in the §5 diagram).
Key results
- Definition 1 (constraint), §3. C constrains A → B at τ iff (I) A → B and A_C → B_C are asymmetric with respect to B at τ, and (II) the aspects of C responsible are temporally symmetric with respect to A_C → B_C at τ. Hence no flow of matter or free energy from C to the process at τ; the heart’s “pumping” is re-described as its architecture constraining the conversion of cardiomyocyte chemical energy into blood kinetic energy.
- Definitions 2–3 (dependence), §4. C1 depends on C2 if C2 constrains, at τ2, a process producing aspects of C1 relevant at τ1; directly if no later process at τ2 maintains that aspect. Slow (τ2 > τ1) and fast (τ1 > τ2) dependence both occur; alkaline phosphatase is the fast example.
- Definition 4 (closure), §5. Overall closure: every member depends directly on another member and is generative for another. Strict: not splittable into two closed sets. A closed set must contain at least one constraint with τs − τd > 0 and one with τs − τd < 0. Constraints subject to closure “constitute biological functions”.
- Closure versus cycles (footnote 25). In the hydrologic cycle entities transform into each other but “do not act as constraints on each other”.
- Boundaries (§6). Lower bound: ascribe closure only to maximally closed systems in the available description. Indirect mutual dependence (the hunting/cooking group) gives overall but not strict closure.
- Tendency to closure (§6.1). K(V) counts dependencies between constraints subject to closure located in volume V; δK(V, l) by spatial scale l peaks within each organism and collapses at its edge (Figure 1, two toy cells sharing two functions at a larger scale). Conjectured to be higher for a unicellular eukaryote than a metazoan cell, for a bacterium than a biofilm.
- Levels (§6.2). Individual cells fail both conditions of Definition 1 with respect to the organism’s processes (negligible effect; symmetries at larger scales), so cell and organism closures separate while nesting. Two closures that are separated and hierarchically nested are two levels of organisation.
- Invariance and variation (§7). Closure is invariant over any sufficiently long interval while the constraints realising it vary.
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | Constraints and processes are distinct causal regimes, separated by time-scale symmetry | moderate | Definition 1 with enzyme, vasculature and heart examples |
| C2 | Biological organisation is closure of constraints | moderate as a characterisation | Definitions 2–4; schematic diagrams; no organism fully described |
| C3 | Closure differs from cycles of processes | moderate | footnote 25 (hydrologic cycle) |
| C4 | Constraints subject to closure are biological functions | moderate | §5, citing Mossio et al. 2009 |
| C5 | Tendency to closure can draw boundaries between interdependent organisms | weak | toy Figure 1; called “preliminary” |
| C6 | Separated and nested closures define levels of organisation | weak | conjecture, §6.2 |
Concepts
- Constraint. An entity that changes a process while its relevant aspects are conserved through that process at the process’s time scale.
- Dependence / generative. C1 depends on C2 if C2 constrains the production of C1’s relevant aspects; C2 is then generative for C1.
- Closure (overall, strict). As Definition 4.
- Maximally closed system. The set of all mutually dependent constraints in the available description.
- Tendency to closure. δK(V, l), the gain in constraint dependencies per increment of volume at spatial scale l.
Connections
It sits between Rosen (closure to efficient causation) and Kauffman (the work–constraint cycle), and against autopoiesis for being under-specified. It narrows the closure of Mossio, Saborido & Moreno (LIT-570), which a flame satisfied, and carries over that paper’s function thesis. The time-scale treatment of symmetries comes from Longo & Montévil. It cites Nicolis & Prigogine (in the record as LIT-524, via LIT-516) for open systems.
Bearing on the record
- LIT-536 (Biological Autonomy). This reading stands in for the book’s chapter 1, with the caveat in the LIT note: the paper refers to the book for whether closure defines organisms.
- The flame. The paper does not mention flames. My reading, not the paper’s: Definition 4 needs every member to be both dependent and generative within the set, so a system whose only constraint is one macroscopic pattern (the 2009 paper’s description of a flame) cannot realise closure; whether a flame has more than one constraint at distinct time scales is an empirical question the paper does not take up. This is the check NOTE-093 asked for (Tessler, bacteria, candle flame), still undone.
- Baedke (LIT-167). §6.1–6.2 is the paper’s answer to the holobiont objection; NOTE-148 records that Baedke draws his overcomer contrast with “Montévil et al. 2016”, not with this paper.
- Markov blankets (LIT-526; LIT-603; LIT-598). A rival boundary criterion. Raja et al. argue blankets cannot represent a boundary the system produces; closure is defined by production among constraints. No paper in the record compares the two on the same system.
- THEORY filed as THEORY-072 (Proposed): organisms are separated from dissipative structures by closure among multiple constraints at distinct time scales (with differentiation, per LIT-570), not by far-from-equilibrium self-maintenance alone.
- No instruction for machine-learning practice.
Limitations
- No real organism, cell or protocell is described as a closed network; the examples are fragments and toy diagrams.
- Direct dependence is required without justification, by the authors’ admission.
- The tendency-to-closure measure is a count with no stated way to identify or weight dependencies, applied only to Figure 1.
- Closure is not claimed to be necessary and sufficient for life.
Open questions
- Does any non-living system (a flame, an autocatalytic set, a Belousov–Zhabotinsky medium, Friston’s simulated soup) realise strict closure under Definitions 1–4?
- Can δK(V, l) be computed for a real symbiosis, and does it agree with the boundaries biologists already draw?