NOTE-093: Bruckner — Welfare Subjects and Autopoiesis

Contribution

The paper adds a third, intermediate candidate to the debate over what grounds welfare subjecthood. One side holds that sentience is necessary (Rosati 2009, Singer 1975, Railton 1986). The other holds that teleological organisation is sufficient (Basl and Sandler 2013). Bruckner’s candidate is autopoiesis, offered as sufficient only. Before it, the sufficiency of non-conscious “mattering” had been argued mainly in environmental ethics (fn. 2) and, for phenomenal but not affective consciousness, by Bradford 2023 and Chalmers. After it, there is a stated criterion that admits plants and some machines while excluding simple artefacts, with an explicit strategy for answering the obligation-explosion objection.

Key insight

What an entity’s self-maintaining activity differentially pursues and avoids is thereby constituted as mattering to it, not merely revealed (§4, “Two notes”). Bruckner credits the idea that autopoietic behaviour reveals mattering to Maher 2017. The step he claims as his own is that this mattering is “entirely sufficient” for a good of one’s own (fn. 4). The consequence is that one can have an interest in X without being interested in X, and it holds for humans too: cardiopulmonary function, gut microbiota (§4).

Assumptions

• Premises
  • Welfare subjecthood requires that things can “matter to the entity itself”. Good-for talk about engines is elliptical for serving a human-designed function (§2; fn. 1, following Bradley 2015 and Rosati 2009).
  • Adaptive, self-sustaining behaviour can be read as “behavior in a broad sense” without any commitment to mentality or agency (§5).
• Authorities
  • The notion of autopoiesis is taken from Thompson 2007, who credits Maturana and Varela 1973/1980 and Varela, Maturana and Uribe 1974 (fn. 3).
  • Nearly all the plant cases and the “showing up” locution are taken from Maher 2017 (§§4–5).
• Method premise
  • Abduction: autopoiesis is preferred because it better explains the agreed data (humans and deer are welfare subjects; engines and rocks are not) than sentience or teleology does (§4, second note; §8).
• Restriction left implicit
  • Autopoiesis is claimed to be sufficient, not necessary (§6.2). A sentient non-autopoietic welfare subject would cost the view only its “bonus” explanation of sentience’s sufficiency.

Key results

What it argues, by section:

• §4. Humans and deer are autopoietic: they seek food, sleep, avoid threats, blink, heal and produce antibodies. That autopoiesis is “what makes” them welfare subjects.
• §5. Plants are autopoietic welfare subjects. The cases are thigmomorphism (bean stems thicken in wind); phototropism, with the edge-of-meadow and mid-forest oak contrast; roots growing around impediments; and sun-tracking flower heads, which show that the past matters. Duration matters too: the Venus flytrap closes only if two triggers are tripped within 20 s, and vernalization needs prolonged cold. Defence: wild tobacco recruits big-eyed bugs against hornworms. Anticipatory defence: lima beans release methyl salicylate, and acacias release tannins against kudu. Plants distinguish background noise from what matters. The summary points (§5) are self-care across varied past, present and anticipated conditions, which is “not … a single end as a thermostat”; self-creation “in one way rather than another”; and a world of “non-consciously valenced” significance.
• §6.1. Governors, thermostats and heat-seeking missiles are “mere input-output machines with a few teleological aspects”. They do not create, maintain or repair themselves, so they are not welfare subjects.
• §6.2. A conjecture: all sentient entities are autopoietic, so sentience’s sufficiency would be explained. He flags that the conjecture needs scientific examination.
• §6.3. Sentience is not necessary.
• §6.4. Tessler, a non-sentient self-driving car, meets the definition: it charges itself, changes its own brake fluid, adjusts its tyre pressure to forecasts and calls a tow truck. It is a welfare subject. The reply to “its good is only human ends”: if all humans perished, things would still go well or badly for Tesslers.
• §7.1. The objection, from Rosati 2009: welfare facts give pro tanto reasons to anyone, so Petunia the houseplant and Tessler generate absurd duties.
• §7.2, Part One. By analogy with aesthetic value, and with the welfare of bank robbers and villains, welfare value need not by itself generate agent-neutral reasons.
• §7.3, Part Two. Grant the reasons; they come in strengths (lawn moles < a scurrying mole < a toddler), so plant and machine reasons can be very weak.
• §7.4. To explain why the strengths differ, he adopts Bradford 2023. Non-conscious welfare states have small value, because the good plus phenomenal consciousness of it is an organic unity. The reasons concerning Petunia and Tessler are therefore “extremely weak”.
• §8. The P/C schema. If a proposed ground P of a category C must be extended to hypothetical P (expected, past or counterfactual-typical) to keep intuitive members, then P “is not doing the unifying”. Accepting hypothetical sentience “is to abandon the sentience view”. Autopoiesis is proposed as the real unifier.
• §9. Open questions: whether theories of well-being generalise to insentient autopoietic beings; what follows for invariabilism (Lin 2018); and what relation philosophy of well-being should have to biology and engineering.

Claims

id	claim	strength	support
C1	Autopoiesis is sufficient for welfare subjecthood	informal argument (abductive)	§4 human and deer cases; §5 plant cases; explicitly “not … deduce[d]” (§4)
C2	Autopoietic activity constitutes, not merely reveals, what matters to an entity	assertion	§4, first note; the idea of revealing is credited to Maher 2017 (fn. 4)
C3	Plants are welfare subjects	informal argument	§5 cases, drawn from Maher 2017; the empirical botany is cited, not examined
C4	Teleologically organised non-autopoietic artefacts (governor, thermostat, missile) are not welfare subjects, and their mattering “differs in kind”	informal argument, weaker than stated	§6.1; the support offered is comparative (“much richer”), so it shows at most a difference of degree
C5	All sentient entities are autopoietic	conjecture	§6.2; the author says it “would need to be examined scientifically”
C6	A non-sentient self-maintaining car of the near future would be a welfare subject	informal argument (thought experiment)	§6.4 Tessler; the reply rests on the intuition in the “if humans perished” case
C7	Welfare value does not by itself generate agent-neutral reasons	informal argument (analogy)	§7.2 aesthetic value; robbers and villains; the author concedes the debate is unsettled (§7.3)
C8	Reasons from non-conscious welfare are extremely weak	borrowed argument	§7.4, resting wholly on Bradford 2023
C9	Hypothetical-sentience extensions abandon the sentience view, and autopoiesis is what unifies the cases	informal argument (first half) plus assertion (second half)	§8 P/C schema; “My suggestion, of course…”

Method

Conceptual analysis with inference to the best explanation. The data are the agreed classifications: humans, deer and mice are welfare subjects; rocks and engines are not. Candidate grounds (sentience, teleological organisation, autopoiesis) are compared for how well they explain those classifications and the contested cases (plants, Tessler, hypothetical-sentience cases). The argument is dialectical: objections are answered with independent replies (§7.2, §7.3), and a borrowed account fills the gap in the second reply (§7.4).

Concepts

• welfare subject: “a being with a good of its own … for whom things can go better or worse” (§2).
• mattering to / having an interest: what an entity differentially pursues or avoids through its self-maintaining activity. Distinguished from being interested in, which requires consciousness (§§3–5).
• sentience (affective consciousness): the capacity for positively or negatively valenced conscious experience (§3, §7.4).
• phenomenal consciousness: there being something it is like to be the entity (Chalmers 2020; Block 1995, LIT-643) (§7.4).
• teleological organization sense of mattering: conditions conduce to, or interfere with, an artefact’s designed end: oil for engines, seasoning for cast iron, humidity for pianos (§3).
• autopoietic: “sustains, maintains, and renews itself in a way that responds and adapts to its environmental conditions”; “produces itself on an ongoing basis where its productive activities are responsive to the challenges and opportunities presented by its situation” (§4). A looser notion than Maturana and Varela’s; see corrections.
• showing up: Maher’s term for what an entity’s behaviour singles out as worthy of response (§4).

Connections

• LIT-146 (SEP, Moral Status of Animals), read. Its §1.3.1 reports Wood 1998’s extension of Kantian considerability to beings that potentially, virtually, formerly or partly have rational nature. That has the same structure as the cases Bruckner’s §8 uses against hypothetical sentience. Bruckner’s argument would apply to Wood’s move too, though he does not cite it. The SEP entry cites Bruckner 2021 on animals’ inability to form autonomous preferences (its §2.1). It separates considerability from weight (“moral radar screen”, §2.2), which is the structure Bruckner’s §7.3–7.4 relies on.
• LIT-192 (Nahas & Sachs), read. They report that autopoiesis (Thompson 2007) is criticised by Deacon and Cashman 2013 precisely for lacking the tools to demarcate living teleology from dissipative self-organisation, and that closure-of-constraints accounts face the candle-flame problem (their §4.1). Bruckner’s demarcation from thermostats is behavioural and comparative, and does not engage that literature. By Nahas and Sachs’s lights his use of teleology serves a third, normative project, whose success criteria (fit with considered welfare judgements) differ from both of theirs.
• LIT-167 (Baedke, The Organism), read. Baedke argues that self-maintenance (“persister”) criteria are too permissive to individuate organisms, and that they carry a “preservation teleology” that reduces organismal agency to self-maintenance (NOTE-148). Bruckner’s criterion is exactly such a persister criterion, used for a normative purpose. Baedke’s permissiveness worry transfers: colonies, holobionts and ecosystems would plausibly qualify as welfare subjects, and Bruckner does not address that.
• LIT-168 (SEP, Biological Individuals), read. Its §6.3 autonomy-and-agency material (Moreno and Mossio 2015) is the organisational tradition that Bruckner’s “autopoiesis” gestures at without adopting.
• LIT-193 (Terrill et al., plants and degrees of moral status), Deferred when this was written; since read in NOTE-120. It is a rival route to partial plant status, through degrees of sentience (“quasi-sentience”). Bruckner routes around sentience instead. Not cross-checked.
• LIT-111 (Birch) and LIT-135 (Seth), both Deferred. The dossier contrasts these as consciousness- or life-based views. Not checked against their texts here.
• Other works it depends on, none held in the record: Maher 2017, Plant Minds; Bradford 2023, Noûs 57: 905–921; Basl and Sandler 2013; Rosati 2009 and 2020; Thompson 2007; Lin 2018.

Agency. The paper holds no account of agency and explicitly avoids one for plants (§5). It presupposes a minimal, behaviourally read notion of self-directed, adaptive self-maintenance, which is what an organisational account of minimal agency would call autonomy. It then uses that notion as the ground of normative standing. The demarcation it draws, between adaptive self-maintainers and single-goal homeostats and missiles, is the same demarcation argued in the goal-directedness literature. The tag is justified on that basis, as secondary to ethics.

Bearing on the record

• No THEORY document in this record touches welfare subjecthood, so none is supported or contradicted. The paper would be a natural source if the record ever files a THEORY on criteria for moral patienthood that do not run through consciousness. Tessler is the case such a document would have to state.
• For ML practice: nothing directly. It carries no instruction for training or evaluating models. Its bearing on AI is conceptual. An AI system would qualify on this view only if it adaptively maintains itself. Nothing in the paper suggests that a model’s input–output competence, however sophisticated, counts, and §6.1 puts “input-output machines” outside the criterion. No ANTH document is affected.
• NOTE-093’s summary should drop “producing” as a gloss, or qualify it (see corrections), and should record that §7.4’s answer imports a phenomenal-consciousness condition on how much welfare matters.

Limitations

• The central notion is never given necessary and sufficient conditions. “Autopoiesis” is fixed by paraphrase and by cases, and borderline cases are not tested: bacteria, colonies, fire, a Roomba that docks itself, a server cluster that reroutes around failures and orders replacement disks.
• The kind/degree tension of §6.1 (see corrections) leaves the thermostat line without a principle. Tessler differs from a thermostat in how many of its own conditions it regulates, not in any structural respect the paper names.
• The plant evidence is taken from a single philosophical secondary source (Maher 2017), not from the botanical literature.
• The §7 replies are both borrowed or hedged. Part One is admitted to be unsettled. The strength account (§7.4) depends entirely on Bradford 2023, and it makes phenomenal consciousness decisive for how much welfare matters, which concedes much of what the sentience view wanted in practice.
• §8’s conclusion (autopoiesis is the unifier) is asserted, not argued against alternatives.
• By the author’s own statement it is an “exploration” meant to make the claim “plausible and likely” (§4), not a proof.

Open questions

• Does adaptive self-maintenance without self-production (Tessler) satisfy any principled organisational criterion? That would be settled by stating the criterion formally and checking Tessler, bacteria and a candle flame against it, for example against closure of constraints (Montévil and Mossio 2015).
• Is the criterion too permissive in Baedke’s sense, taking in colonies and ecosystems as welfare subjects?
• Do theories of well-being generalise to insentient autopoietic beings, and does invariabilism survive (§9)?
• Does Bradford’s organic-unity account of the weakness of reasons collapse the practical difference between Bruckner’s view and a consciousness-based view?

Corrections to the seeded skim

• “Autopoiesis” here is not self-production of components. The dossier and NOTE-093 gloss the criterion as “adaptively producing, maintaining and renewing oneself” and flag the Maturana–Varela origin as a question. The full text settles it. Bruckner follows Thompson 2007 chs. 5–6 (fn. 3) and Maher 2017, and never uses the organisational definition: nothing on components producing the network that produces them, and nothing on a self-produced boundary or operational closure. His decisive case, Tessler (§6.4), produces none of its parts. It “cannot replace” worn or broken parts itself and drives to a service station. Yet he says it “sustains, maintains, and renews itself” and qualifies. In practice the criterion is adaptive, self-directed maintenance of one’s own condition across a range of environmental contingencies.
• The kind/degree line is internally unstable, which the skim only suspected. §6.1 says the mattering of mere teleological entities “differs in kind” from that of autopoietic ones. The argument then given is comparative: governors and missiles show mattering only “in a very loose sense”, and autopoietic mattering is “much richer”. Plants are distinguished from homeostats because they respond to “a wide variety” of conditions, not “a single end” (§5). So the criterion is graded in its own terms, and Tessler shows that it can be engineered in.
• Bradford is §7.4, not Part Two. The skim puts Bradford 2023 in §7’s “Part Two”. Part Two (§7.3) argues only that moral reasons come in strengths (the moles/toddler example). Bradford enters in a separate §7.4, “A Further Challenge”, to answer Rosati 2020’s demand for an account of why reasons differ in strength. The answer there is that non-conscious welfare states have little value because welfare good plus phenomenal consciousness forms an organic unity (Bradford 2023: 913–918). The reply therefore imports a phenomenal-consciousness condition on how much welfare matters, while denying that consciousness is needed for welfare subjecthood.
• §8 argues by schema, and the conclusion is asserted. The skim’s gloss, that autopoiesis grounds welfare “from entities that were, will be, or would have been sentient”, is right about the cases (A)–(C): the fetus, the coma patient, the person with a severe congenital neurological disorder. But the argument only shows that a sentience view extended to hypothetical sentience has abandoned sentience as the unifier. That autopoiesis is the unifier is introduced as “My suggestion, of course” (§8). No rival unifier is considered, such as being a living organism of a sentient kind, or Wood’s “infrastructure of rational nature” (reported in LIT-146 §1.3.1), which has the same potential/past/counterfactual structure.
• Minor: the skim’s §6.1 list (“engine governor, a thermostat and a heat-seeking missile”) is exact: “an engine with a governor, a cooling system with a thermostat, or a heat-seeking missile”. Plant agency is explicitly not claimed: “No theoretical commitments—for example, about plant agency—will be undertaken if we take behavior in a broad sense” (§5), where behaviour includes “shingles in a windstorm”.
• agency tag: justified, as a secondary tag. The paper disclaims plant agency, but its operative criterion (self-directed, adaptive self-maintenance that constitutes what “matters” to an entity, set against thermostats and missiles) is the goal-directedness/autonomy demarcation in the tag’s blurb, and it applies it to machines. It should not be cited as an account of agency.