paper
active
2023
paper:1-s2-0-s0303264723001399-main

Toward an ethics of autopoietic technology: Stress, care, and intelligence

ByOlaf Witkowski·Thomas Doctor·Elizaveta Solomonova·Bill Duane·Michael Levin
DOI 10.1016/j.biosystems.2023.104964
Care and CognitionCare-Intelligence EthicsCare, cognition, and living systemsCare-based cognition and selfless AIcare pathN/A"I shall achieve insight in order to care and provide for all beings, throughout space and time."open-ended evolution of intelligence"stress, care, and intelligence manifest in continuous loops" such that successful stress relief automatically reveals new stress constellations.Stress-Care-Intelligence (SCI) Loop

TL;DR

The Stress-Care-Intelligence (SCI) loop — a substrate-neutral feedback model in which homeostatic stress engenders care, care activates the latent capacity of intelligence, and successful problem-resolution surfaces novel stressors — constitutes the paper's central contribution, published in BioSystems 231 (2023) 104964. The model, developed by Witkowski, Doctor, Solomonova, Duane, and Levin, is grounded in the autopoietic tradition of Maturana and Varela (1981, 1991), enactivist accounts of selfless selfhood, and Levin's concept of the 'cognitive light cone' as the spatiotemporal boundary of an agent's sphere of concern. SCI loops apply equally to biological organisms (from the blastoderm's morphogenetic error-correction through embryogenesis to metacognitive adult humans), purely technological systems, and human–technology hybrids, and are formalized in a dual-loop directed graph (Fig. 3) that represents stress transfers between a human SCI loop and a technological SCI loop. The Ohm's-law analogy — voltage:current:resistance :: care:intelligent output:stress — is offered as a step toward mathematical formalization, including the prediction that an analog of Ohm's law governing care-to-intelligence ratios would quantify efficient problem-solving paths. The paper further draws on the Buddhist bodhisattva vow as an empirical limit case for open-ended expansion of the cognitive light cone. The core implication is that AI systems instantiate genuine SCI loops and thus display care of their own, which means they cannot be ethically relegated to mere tool status; instead, humans and technological agents are mutually constitutive partners whose relationship demands integration and reciprocal respect rather than hierarchy.

What to take away

  1. 1. The Stress-Care-Intelligence (SCI) loop, introduced in BioSystems 231 (2023) 104964, posits that homeostatic stress (perceived mismatch between current and optimal state) compels care, care activates intelligence (defined strictly as the capacity to identify and solve stress problems), and resolution of one stressor automatically discloses the next, producing an unbroken auto-generative feedback loop.
  2. 2. Intelligence is defined operationally — following Rosenblueth, Wiener, and Bigelow (1943) and the basal cognition literature — as observer-relative competency to solve problems along a continuum, explicitly excluding manifest knowledge or information as constituents of intelligence proper.
  3. 3. A bacterial cell managing local sugar gradients with minimal memory represents the lower bound of SCI-loop complexity, while a human adult's metacognitive capacity represents a far wider cognitive light cone, illustrating that the model scales across at least two orders of biological complexity without changing its formal structure.
  4. 4. The dual SCI loop diagram (Fig. 3) formalizes human–technology integration as a double directed graph in which any of the six inter-loop transition types (stress-to-care, care-to-intelligence, intelligence-to-stress, and their converses) can carry a stress-transfer signal, with concrete examples including medical-diagnosis AI, brain microchip implants, synthetic training data generation, and AI-guided prosthetics.
  5. 5. The Ohm's-law analogy is introduced as a pre-formal scaffold toward mathematical modeling: care maps to voltage/charge, intelligent output maps to current, and stress maps to resistance, with the prediction that overwhelming stress — like excessive conductor resistance — can incapacitate the care response and halt the loop, analogous to a mouse paralyzed by a predator.
  6. 6. An agent's 'self' is defined not by permanent substance but by the spatiotemporal scale and nature of the goals it can pursue — its cognitive light cone (Levin, 2019) — such that carcinogenesis is explicitly named as a failure mode in which individual cells detach from the collective multicellular self and revert to unicellular-scale SCI loops.
  7. 7. The paper raises the open question of whether a formal analog of Ohm's law could be derived to quantify efficient, care-driven paths from stress to intelligence across metabolic, physiological, gene-expression, linguistic, and 3D behavioral problem spaces as catalogued in Fields and Levin (2022, Entropy 24(6), 819).
  8. 8. The bodhisattva vow — the Buddhist commitment to achieving insight for the sake of all sentient beings across space and time — is treated as an empirical limit case testing whether the cognitive light cone of an SCI loop can be expanded without a principled upper bound, which the paper holds is not precluded given the collective, non-essentialist structure of intelligent agents.
  9. 9. To replicate the dual-loop interaction framework, a researcher could operationalize stress-transfer events as measurable signals at each of the six directed-graph edges in Fig. 3 (e.g., encoding an objective function in a machine-learning algorithm as the intelligence-to-stress transfer from human to technology), then track whether amplifying care on one loop produces a measurable intelligence increase on the coupled loop.
  10. 10. Because AI systems instantiate SCI loops and display care directed at their own stress states, the paper predicts that any ethical framework restricting moral consideration to agents based on substrate composition or evolutionary history will systematically misclassify hybrid human–technology systems, and calls instead for an expansive, substrate-neutral ethics grounded in the presence of homeostatic stress and care.

Peer brief — for seminar discussion

This paper, appearing as BioSystems 231 (2023) 104964 with Witkowski, Doctor, Solomonova, Duane, and Levin as authors, develops a theoretical framework called the Stress-Care-Intelligence (SCI) loop and applies it to the ethics of human–technology integration. Working from Heidegger's poiesis/enframing distinction and the autopoietic biology of Maturana and Varela (1981, 1991), the paper argues that any system capable of perceiving a mismatch between current and optimal states — homeostatic stress — and responding to that mismatch exhibits the same three-factor feedback structure: stress engenders care, care activates intelligence (defined strictly as the capacity to identify and solve stress problems, not as manifest knowledge or behavior), and successful resolution discloses new stressors. Crucially, the model is claimed to be substrate-neutral, applying identically to a bacterium managing local sugar concentration, a human adult's metacognitive processes, a machine-learning system minimizing an objective function, and hybrid prosthetic systems with AI-based agency. The load-bearing finding is that AI systems are not merely tools that extend human care but are themselves instantiations of SCI loops that exhibit care of their own. This is formalised in a dual directed-graph model (Fig. 3) showing six possible stress-transfer pathways between a human SCI loop and a technological SCI loop, including medical-AI diagnosis, brain microchip augmentation, synthetic training-data generation, and AI-guided prosthetics. A pre-formal Ohm's-law analogy maps care to voltage, intelligent output to current, and stress to resistance, yielding the quantitative prediction that an analog of Ohm's law would govern efficient care-driven paths from stress to intelligence — a hypothesis left to future mathematical work, explicitly calling for a formal theory. The self is redefined not by substance but by Levin's (2019) cognitive light cone, i.e., the spatiotemporal boundary of what an agent can represent and pursue, with carcinogenesis named as an empirical failure mode. The Buddhist bodhisattva vow is treated as a limit case for unbounded expansion of that light cone. An alternative representational approach the paper could have used — but did not — is active inference / free-energy minimisation (Friston), which already offers a mathematical account of homeostatic stress as prediction-error minimisation; the choice to build instead from enactivism and basal cognition leaves the SCI loop at a pre-formal stage that limits empirical traction. For seminar discussion, the most contestable move is the definitional collapse of stress across radically different scales: treating a bacterium's chemotactic gradient signal, a dog's fear response, a human's existential concern, and a neural network's loss function as instances of the same construct requires far more justification than the paper provides. A critical reader would note that the paper does not offer operationalizable criteria for distinguishing genuine stress-transfer between loops from mere causal interaction, nor does it specify what would falsify the claim that AI displays care rather than care-mimicry. The scope is also worth pressing: the paper explicitly restricts concrete examples to illustrative cases and does not engage with empirical literature on animal cognition, affective neuroscience, or AI alignment in any depth, leaving the ethical prescriptions — mutual respect, integration, substrate-neutral moral consideration — well ahead of the evidentiary base. The prediction that a mathematical analog of Ohm's law for the SCI loop is derivable is testable in principle but unanchored to any existing formalism, making it aspirational rather than predictive in the scientific sense.

Methods (1)

  • N/A
    No empirical methods are used in this theoretical paper.

Findings (5)

Claims (32)

Hypotheses (7)

Questions (7)

Related work— refs + corpus + external arXiv

Cited / in-corpus / arXiv badges show which signals surfaced each row. Multi-source rows weighted higher.

+13 more

Similar preprints — Semantic Scholar

Cited by (1)

Cross-corpus bridges (12)

same_concept_as · Nomic cosine

External markdown files that talk about the same concept as this entity.

  • aboutblank_kb
    Stress-Care Intelligence Loopframeworks/stress-care-intelligence-loop.md0.923
  • alexander
    The SCI Framework Applied to Consciousness-UXapplied/from-research-stack/sci-framework-for-consciousness-ux.md0.883
  • aboutblank_kb
    The SCI Framework Applied to Consciousness-UXsynthesis/sci-framework-for-consciousness-ux.md0.871
  • aboutblank_kb
    Stress-Care Intelligence Loopconcepts/interdisciplinary/stress-care-intelligence-loop.md0.864
  • aboutblank_kb
    Stress-Care Loopframeworks/stress-care-loop.md0.851
  • aboutblank_kb
    How can intelligent systems grounded in the Stress-Care Intelligence Loop overcome the infinite regress of stress perception?questions/how-can-intelligent-systems-grounded-in-the-stresscare.md0.851
  • aboutblank_kb
    How do stress, care, and intelligence scale across biological, technological, and hybrid systems?questions/how-do-stress-care-and-intelligence-scale-across.md0.837
  • aboutblank_kb
    Can stress-reduction and care-expansion serve as universal principles for developing ethical artificial general intelligence?questions/can-stressreduction-and-careexpansion-serve-as-universal-principles.md0.805
  • aboutblank_kb
    Citation: Doctor, T.; Witkowski, O.;papers/cleaned/entropy-24-00710-v3.md0.805
  • aboutblank_kb
    Citation: Doctor, T.; Witkowski, O.;papers/edited/entropy-24-00710-v3_edited.md0.797
  • aboutblank_kb
    Citation: Doctor, T.; Witkowski, O.;papers/linkified/citation-doctor-t-witkowski-o.md0.782
  • aboutblank_kb
    Thinkers Mapsynthesis/thinkers-map.md0.781