Contribution
A critique of Weber and Varela’s (2002) proposal that autopoiesis grounds both intrinsic teleology and sense-making. Di Paolo grants the first and denies the second. Autopoiesis in its original formulation allows no degrees, so the only norm it yields is live-or-die, and that distinction is available to an observer, not to the system. He names the missing property, adaptivity, defines it, and uses it to give new accounts of breakdown, meaning-construction, behaviour, agency, health and the time-structure of acts. The definition of agency here is the one Barandiaran, Di Paolo & Rohde (LIT-566) later formalise.
Key insight
Staying alive is not the same as being able to tell that one is getting worse. A system that merely conserves its organisation is equally well off in every viable state until it dies. Only a system that senses tendencies toward the edge of its viability and acts on them has states that are better or worse for it while all of them are viable, and only then can anything mean anything to it.
Assumptions
- Autopoiesis in its original formulation. Note 1: “throughout this paper the term autopoiesis will be used in its original formulation and interpretation which are given in the primary literature”. Later, looser readings of autopoiesis (Varela’s own included) may already contain adaptivity; Di Paolo says that if so, “it still needs to be made explicit in operational terms”.
- Structural determinism. Changes of state “always operate in the present as a result of its current structure”. This is why the system cannot respond to a future event, only to present tendencies.
- A viability set of “finite measure, bounded, and possibly time-varying”, the ranges of perturbation and structural change the system tolerates, “defined by the organization and current state of the system”. It is not given formally beyond this sentence.
- Physicality. Production is material, in physical space, with a membrane (note 3, which calls the demand “at first ad hoc and arbitrary” and leaves it unresolved).
- Method. “pragmatic circulation” between science and phenomenology: Jonas’s insights are signposts to be explained, not explanations.
Key results
- Conservation, not homeostasis (Autopoiesis as conservation). Maturana
and Varela use both words. Homeostasis implies active mechanisms;
conservation is “a set-theoretic condition that may or may not be
realized in an active manner”. Autopoiesis, “a concept allowing no
degrees, can only entail the more general case of conservation (thus
admitting the possibility of fortuitous self-maintenance)”.
- The norm conservation suggests is all-or-nothing: “organisms live as long as they don’t die”. It lets the observer tell lethal from non-lethal encounters, and so satisfies Weber and Varela’s step A (a unity and a norm), but not step B, which “would require the autopoietic system itself to be able to make the same distinction.”
- The cliff: “balancing at the edge of a cliff is a perfectly viable behaviour, so is falling over the edge – both are interactions that conserve autopoiesis.”
- The bacterium: “Even if the current rate of nutrient intake is lower than the rate of consumption (leading to certain loss of autopoiesis in the near future), bacteria will not seek higher concentrations just because they are autopoietic”.
- Robustness against adaptivity. Robustness is tolerance of a range of perturbations and structural changes; it “implies endurance but not necessarily adaptivity”.
- The definition (Adaptivity). “A system’s capacity, in some
circumstances, to regulate its states and its relation to the
environment with the result that, if the states are sufficiently close to
the boundary of viability, 1. Tendencies are distinguished and acted upon
depending on whether the states will approach or recede from the boundary
and, as a consequence, 2. Tendencies of the first kind are moved closer to
or transformed into tendencies of the second and so future states are
prevented from reaching the boundary with an outward velocity.”
- It may come from dedicated mechanisms or be emergent; it “is not a direct consequence of the definition of autopoiesis”.
- A breakdown is “the severity of a negative tendency distinguished and measured by the amount of regulative resources that it demands to compensate for it”.
- Both self-monitoring and regulation are needed: “Events that provoke the same regulative response are not meaningfully distinguishable.”
- Adaptivity without self-production is cybernetics, and Jonas’s criticism holds: a feedback loop “may instantiate a purpose, but this purpose is external to it”.
- Properties of adaptivity.
- No optimum: “it does not imply any kind of optimal or ideal state for the organism”, and there may be neutral regions. “Nevertheless, adaptivity establishes an inner norm, a distinction between good, neutral, and bad ways of realizing autopoiesis.”
- Graded: “Regulation may function over a large range of conditions or only in selected extreme situations.”
- Recursive in two senses. Adaptive processes may regulate by distance from triggering more basic adaptive responses (preferring nutritious food to avoid drawing on reserves). And “regulative processes may themselves be plastically adjusted or transformed depending on their effectiveness.”
- Pathology becomes definable as failure of adaptivity while still autopoietic: “exhaustion of adaptive resources, malfunction of regulation, loss of adaptive buffering provoking the activation of extreme regulation, disharmonious activation of conflicting adaptive mechanisms”. After Canguilhem and Goldstein, “a successful cure may well re-define rather than simply restore the organism’s own normativity”.
- Two senses of sense-making. Uncovering established meaning corresponds to regulation by existing adaptive processes; constructing new meaning to their “plastic modification, attunement or generation”. Only the second explains why a proxy (the smell of food, snowprints) can mean anything.
- Behaviour and agency (Natural agency). Structural coupling is undergone; behaviour is its regulation: “Unregulated coupling is better described as suffering an exchange while behaviour is the control and selection of what exchanges to suffer.” An agent is “a self-constructed unity that engages the world by actively regulating its exchanges with it for adaptive purposes that are meant to serve its continued viability.” Note 8: agency is not implied by autopoiesis and adaptivity together (internal-only regulation, as in the Chemoton), and comes in degrees.
- Robots. “Being functionally scared is not the same as being scared.” The test is the system’s organisation, its “self-generated structure of sense-making (if any) and its relation to the system’s viability”. He leaves open whether artificial intentionality is possible “short of reproducing a fully metabolizing robot”.
- Temporality. An adaptive act has onset, acceleration, consummation and cadence (after Langer). It has a direction that autopoiesis lacks: run backwards, autopoiesis is still conserved, but adaptivity becomes dysfunctional. There is a minimum grain below which an act decomposes into “raw processes”. Adaptive responses can leave the organism “viable but changed”, and so history begins.
- Non-metabolic values (The road ahead, declared speculative). Self-sustaining forms within metabolic viability but “underdetermined by it” (habits, sensorimotor patterns, social values, human projects) can, “via a process of adaptive closure analogous to metabolism”, generate their own values, so that “different identities” coexist in one organism. Prismatic-goggle adaptation happens only when the disrupted patterns are exercised. Addiction is a self-generated value that “may even work against metabolic viability”. Without such forms “inner conflict would never arise”.
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | Autopoiesis in its original formulation entails conservation, not active homeostasis, and so an all-or-nothing norm | moderate | a reading of the primary texts (Maturana 1975; Maturana & Varela 1980; Maturana 2002) |
| C2 | Sense-making requires adaptivity in addition to autopoiesis | moderate | conceptual argument (the bacterium, the cliff); depends on C1 |
| C3 | Adaptivity is a real, operational property not implied by autopoiesis | moderate | note 6: minimal autopoiesis models and synthetic vesicles lack it; the Chemoton has it |
| C4 | Adaptivity yields a graded inner norm (good, neutral, bad) without an optimum | moderate | follows from the definition |
| C5 | Behaviour is the regulation of structural coupling, and agency is behaviour for the system’s viability | moderate as a proposal | definitional, defended by the asymmetry argument |
| C6 | Adaptivity, unlike autopoiesis, is time-asymmetric | weak | a time-reversal thought experiment, not worked out |
| C7 | Self-sustaining non-metabolic forms generate their own values, which can conflict with metabolism | speculative | goggles, addiction, rehabilitation; declared speculative |
Concepts
- Conservation. Changes of state without loss of organisation; it may happen fortuitously.
- Homeostasis. Conservation by active compensating mechanisms.
- Robustness. Tolerance of a range of perturbations and internal changes before autopoiesis is lost.
- Viability set. Those ranges, bounded, of finite measure, possibly time-varying, fixed by the organisation and current state.
- Adaptivity. The capacity defined above: regulating states and relations to the environment so that tendencies toward the viability boundary are detected and reversed.
- Breakdown. A negative tendency, measured by the regulative resources needed to compensate it.
- Behaviour. Regulation of structural coupling, as distinct from the coupling.
- Agent. A self-constructed unity that regulates its exchanges for adaptive purposes serving its viability.
- Sense-making. A concerned perspective on the world. Split here into uncovering meaning (established regulation) and constructing it (plastic change of the regulation).
Connections
It corrects Weber & Varela (2002; LIT-1018, read the same day in NOTE-720) and builds on Jonas’s philosophy of the organism, Ashby (1960) for how adaptivity might arise, Canguilhem and Goldstein (LIT-167) on health, and Ruiz-Mirazo & Moreno (2000) for agency. Its definition of agency, its asymmetry argument and its four-phase act are taken up by Barandiaran, Di Paolo & Rohde (LIT-566), where adaptive modulation of coupling becomes clause (2) of the definition that THEORY-071 records. De Jaegher & Di Paolo (LIT-571) extend sense-making to social interaction. Its note 6 links adaptive autopoiesis to Boden’s “strongest” sense of metabolism.
Bearing on the record
Is adaptivity a capacity in the sense of CLAIM-025’s transformation capacity? Mostly not, but its second-order part comes close.
- First-order adaptivity, as defined, is regulation against one fixed boundary: tendencies toward the edge of the viability set are detected and turned back. That is closer to what CLAIM-025 calls the operative regime, or to its governing organisation, than to the capacity to transform that organisation. Nothing in the definition requires the system to be able to change what it regulates or how.
- Two places in the paper reach further. The second recursion, in which “regulative processes may themselves be plastically adjusted or transformed depending on their effectiveness”, is a capacity to change the regulation itself; and meaning-construction is “plastic modification, attunement or generation of novel adaptive processes”. The cure that may “re-define rather than simply restore the organism’s own normativity” is the Canguilhem point that CLAIM-018 makes, and here it is stated for organisms by an author the record reads. But Di Paolo does not separate these layers as things that can persist or fail apart. He treats them as aspects of one graded property. CLAIM-025’s thesis is that they come apart, so this paper is not evidence for it. It supplies the vocabulary in which the thesis could be stated for organisms.
- What the paper does separate is autopoiesis from adaptivity: a system can remain autopoietic while its adaptivity fails (exhaustion, lost buffering, malfunction of regulation). That is a separation of the same shape as CLAIM-099’s (persisting is not keeping the capacity to govern one’s persistence), one level down: maintaining organisation and maintaining the capacity to regulate it are different achievements. The bacterium that is still autopoietic but whose intake is below consumption, “leading to certain loss of autopoiesis in the near future”, is CLAIM-079’s case in organic dress: currently satisfying the constraints with no trajectory that keeps satisfying them. The mapping is the record’s. Di Paolo does not distinguish the viability set (K) from the viability kernel (THEORY-134). His “viability set” is the set of tolerated states, so states inside it with no viable future are, on his terms, still in it, which is exactly the gap CLAIM-079 names.
- On ARG-007’s mapping premise. ARG-007’s weakest step is choosing the constraint set for an agent’s regulatory capacities. Di Paolo’s speculative last section offers one answer: a self-sustaining behavioural or social form, by “adaptive closure analogous to metabolism”, fixes its own viability conditions, distinct from the organism’s. If that held, K for a regime would be set by the regime’s own closure, not chosen by an observer. He gives it no criterion and calls it speculation, so it is a lead, not an answer.
- Addiction and protective lock-in. His addiction case, a self-sustaining form whose value “may even work against metabolic viability”, resembles TERM-020 in that a regime maintains itself at the cost of the larger system. The resemblance fails on the mechanism. TERM-020 is protective, begins in a disturbance and works through the conditions that justify it; Di Paolo’s form needs none of that.
What norms does it yield, relative to CLAIM-075? Norms for the system only, and on the point CLAIM-075 makes about viability, Di Paolo agrees.
- He argues that viability alone (conservation) yields no norm the system has, only the observer’s lethal/non-lethal distinction, “the rather useless a posteriori realization by the external observer that the organism should have avoided that very last encounter”. That is a stronger version of CLAIM-075’s disclaimer, not a rival to it.
- What adaptivity adds is “an inner norm, a distinction between good, neutral, and bad ways of realizing autopoiesis”, a “normativity within their current set of viability conditions and for themselves”. It is better or worse for the organism. Nothing in the paper says that the organism, or anything, ought to persist, or that its persistence has value for anyone else. CLAIM-075 does not deny norms of this kind.
- One real difference. CLAIM-075’s summary says “Viability licenses conditional functional judgment only”, an observer’s if-then. Di Paolo’s norm is meant to be non-conditional in one respect: it is generated by and accessible to the system, not imputed from outside. If the record’s conditional means conditional on the system’s persisting, Di Paolo’s norm is still conditional in that sense; if it means ascribed by an observer, his is not. CLAIM-075’s wording does not say which, and this paper is a reason to say.
- His own account undercuts any slide from viability to value: health “may well re-define rather than simply restore” norms, there are “many ways of being healthy”, and non-metabolic values can conflict with metabolism. Persistence is not even the system’s only value.
- Enoch (LIT-117). The shmagency objection targets constitutivist arguments that derive reasons for action from what is constitutive of agency. Di Paolo makes no such argument. He does not claim the organism has reasons, or that anyone should care about its norm. So the objection, as the record describes it, does not bite here. LIT-117 is unread; this is the record’s judgement from its summary.
Elsewhere. THEORY-071 takes its agency conditions from LIT-566, and this paper is their source: asymmetry by regulation of coupling, and agency in degrees (note 8). The robot section bears on LIT-135’s biological naturalism: Di Paolo’s test is organisational, not substrate, and he leaves the metabolism question open.
Limitations
- Conceptual. There is no formal model; the only operational example is the Chemoton, in a note. The viability set is defined in one sentence.
- C1 is a reading of autopoietic theory. Di Paolo concedes (note 1) that later formulations may contain adaptivity, and the critique of Weber and Varela is of their reliance on what the original definition does not supply.
- Tendency, “sufficiently close” and “outward velocity” assume a state space with a metric and a boundary; the paper does not supply one.
- The time-reversal argument is a thought experiment; reversing a reaction network’s flows is not shown to be physically meaningful.
- Agency requires regulation that “in general” improves viability; how general is not said.
- The section on non-metabolic values is declared speculative by the author, and nothing there is argued beyond examples.
Open questions
- What counts as a separate failure of the plastic, second-order layer of adaptivity, as distinct from failure of first-order regulation? A case of an organism with intact regulation and lost plasticity would carry CLAIM-025 into biology.
- Can a self-sustaining non-metabolic form’s viability conditions be stated independently of an observer? That would answer ARG-007’s mapping premise.
- Is Di Paolo’s viability set the constraint set K or the viability kernel? A formal restatement in viability theory (THEORY-134) would settle it.