This note separates results that calibrate the model against established work from phase structure that arises only after the declared feedback loops are coupled. It is a novelty boundary, not a literature review.
The repository's common-mode mixture
[ E_n(q,\rho)=\rho+(1-\rho)(1-q)^n ]
is a deliberately simple model in which a fraction (\rho) of cases defeats all same-mode monitors. The resulting floor (P_{bad}\ge b\rho) is therefore a model consequence, but the underlying lesson is not new: redundant components do not remove a common-cause failure component.
Relevant antecedents include:
- K. N. Fleming (1975), A reliability model for common mode failures in redundant safety systems, which introduced the beta-factor approach to splitting component failure into independent and common-cause parts.
- D. E. Eckhardt and L. D. Lee (1985), A theoretical basis for the analysis of multiversion software subject to coincident errors, IEEE Transactions on Software Engineering 11(12):1511-1517.
- B. Littlewood and D. R. Miller (1989), Conceptual modeling of coincident failures in multiversion software, IEEE Transactions on Software Engineering 15(12):1596-1614, DOI 10.1109/32.58771.
The parameter (\rho) used here is a Bernoulli common-mode mixture weight; it should not be identified numerically with a classical beta factor without an additional mapping.
Luu, Teutsch, Kulkarni and Saxena (CCS 2015), Demystifying Incentives in the Consensus Computer, DOI 10.1145/2810103.2813659, named the verifier's dilemma: rational participants can be poorly incentivized to perform costly verification when they benefit from accepting work without checking it.
Truebit subsequently made the same incentive problem operational. Its verification design uses randomly injected forced errors so that a potential challenger has a non-negligible chance of finding an error and earning a reward. In the language of the present toy model, that mechanism acts on the experienced incidence of detectable faults rather than merely adding nominal monitors.
These precedents mean that neither "rare faults select weak checking" nor "injecting detectable faults can support checking incentives" should be reported here as new.
Closing social and ecological feedback loops can produce multiple stable states, tipping and hysteresis in common-pool-resource models. Relevant examples include:
- A. Richter, D. P. van Soest and J. Grasman (2013), Contagious cooperation, temptation, and ecosystem collapse, Journal of Environmental Economics and Management 66(1):141-158, DOI 10.1016/j.jeem.2013.04.004. Their coupled resource/social-norm model generates endogenous erosion of cooperation, alternative stable states and hysteresis.
- S. Sarkar (2023), Managing ecological thresholds of a risky commons, Royal Society Open Science 10:230969, DOI 10.1098/rsos.230969, which studies monostability, bistability and tipping in a common-resource model.
- Recent coupled cooperation-resource models likewise report bistability and resource collapse under feedback between environmental state and behavior.
Therefore the existence of a fold, alternative stable states, or hysteresis in Experiment 2 is not by itself a novelty claim. The question is whether the specific distributed-control composition here — costly verification, selected-versus-sufficient monitoring, common-mode failure, endogenous producer response and state-dependent corrective capacity — yields useful boundaries or observables not already contained in those literatures.
As a fold is approached, recovery from small perturbations slows because the dominant local multiplier approaches one. Scheffer et al. (2009), Early-warning signals for critical transitions, Nature 461:53-59, DOI 10.1038/nature08227, reviewed critical slowing down and associated early-warning signals across complex systems.
Experiment 2 therefore does not claim critical slowing down as new. Its exact model-specific identity
[ 1-M=\gamma(1-\eta) ]
shows how the distributed-control loop gain maps onto that established quantity. It also clarifies a limitation: slowing down in the current commons state detects nearness to the local fold, but does not by itself measure a separate stock of dormant corrective capacity.
Ecology already contains several close conceptual antecedents to Experiment 3:
- Yachi and Loreau (1999), Biodiversity and ecosystem productivity in a fluctuating environment: the insurance hypothesis, PNAS 96:1463-1468, emphasized buffering of ecosystem functioning under environmental variability rather than only average function.
- Elmqvist et al. (2003), Response diversity, ecosystem change, and resilience, Frontiers in Ecology and the Environment 1:488-494, DOI 10.1890/1540-9295(2003)001[0488:RDECAR]2.0.CO;2, defined response diversity as different responses to environmental change among species contributing to the same ecosystem function.
- Ardichvili et al. (2026), Beyond Biomass: How Interactions Shape Species' Contribution to Ecosystem Functioning, Ecology Letters 29:e70370, DOI 10.1111/ele.70370, explicitly distinguish a species' static contribution from its dynamic contribution after interactions propagate through the community.
These precedents mean that neither "different failure profiles improve resilience" nor "removal effects should include the response of the remaining network" is claimed as new here.
Shapley values and Harsanyi/Möbius interaction decompositions are likewise standard cooperative-game-theory tools. Experiment 3 uses them to keep three questions separate: current-network necessity, average marginal attribution, and explicit non-additivity.
The narrower research object is contribution to a declared viability margin under a specified disturbance/failure architecture, which can later be conditioned on state and behavior.
The next question is not whether correlation leaves a floor at fixed harmful attempt rate (b). It is what happens when the attempted-harm rate is itself a response to the distributed controller.
In the declared adaptive model,
[ b_t^*=\sigma!\left(\frac{g-\ell d_t}{T}\right), \qquad d_t\le 1-\rho. ]
Therefore the producer target is bounded below by its response at maximal achievable detection. Partial adjustment then gives a finite-time lower envelope for (b_t), and the static common-mode floor converts this into an endogenous lower envelope for harmful finalization.
That coupled result is proved in formalization/EndogenousFloor.lean at the
level of any producer response antitone in detection and specialized to the
logistic response in the executable model. Whether this exact coupled theorem
is novel relative to inspection-game, security-game and reliability
literatures remains a literature question; the repository does not currently
claim priority.
The intended research sequence is:
[ \text{recover known limit} \rightarrow \text{couple feedback loops} \rightarrow \text{derive new phase structure} \rightarrow \text{test whether it survives richer incentives and substrates}. ]
Recovering established results is useful evidence that the abstraction is well anchored. Novelty, if present, must come from what the coupled model adds beyond those baselines.
Experiment 8 does not claim that organisms maintaining or modifying environmental conditions is new.
Relevant antecedents include:
- Jones, Lawton & Shachak (1994), Organisms as ecosystem engineers: organisms can create, modify and maintain habitats through non-trophic processes.
- Niche-construction theory: organisms alter environmental states and thereby change subsequent ecological and evolutionary conditions.
- Scholz et al. (2016), Maintenance of Root Function in Tropical Woody Species During Droughts: hydraulic redistribution, xylem repair and facilitation can preserve function during low water availability; redistributed water can become available to neighboring plants.
- Fricker et al. (2007; 2017), work on mycelial networks: fungal networks adapt through growth, branching, fusion and regression in response to resources, damage and predation; resource routes can be selectively reinforced and redundant mycelium recycled.
These literatures support the empirical plausibility of non-human state-dependent maintenance processes. They do not establish the repository's stronger proposed abstraction that signal preservation, correction, reinforcement, repair and maintenance inheritance form a common cross-substrate operator family.
The term "behavior" is therefore operational in Experiment 8. It does not require intention or cognition, and it should be replaced with "participant process" in substrates where behavioral language would invite anthropomorphic inference.
The novelty question, if any, lies in whether the same viability-margin formalism can distinguish when such operators are necessary, excessive, misdirected or counterproductive across substrates—not in the observation that organisms modify environments.
Experiment 11's claim is not that monitoring costly cooperative behavior or conditioning rewards on reputation is new.
Relevant prior art includes:
- Panchanathan & Boyd (2004), Indirect reciprocity can stabilize cooperation without the second-order free rider problem, Nature 432:499–502, DOI 10.1038/nature02978. The paper links costly collective action to reputation and future inclusion.
- Fowler (2005), Human cooperation: second-order free-riding problem solved?, Nature 437:E8, DOI 10.1038/nature04201, challenges whether such reputation mechanisms fully eliminate second-order free riding.
- Carpenter, Kariv & Schotter (2012), Network architecture, cooperation and punishment in public good experiments, Review of Economic Design 16:93–118, DOI 10.1007/s10058-012-0120-z, shows that who can monitor and punish whom is itself a network-architecture question.
- Bruggeman, Sprik & Quax (2021), Spontaneous cooperation for public goods, Journal of Mathematical Sociology, DOI 10.1080/0022250X.2020.1756285, discusses monitoring, reputation maintenance and costly incentives as classic solutions to public-goods cooperation.
- Recent work also shows that reputation/indirect reciprocity can fail across conflicting group scales, reinforcing the need to declare which viability loop a reputation signal is meant to maintain.
The narrower contribution of Experiments 10–11 is therefore not reputation or monitoring per se. It is the integration of:
- a grounded JAM correction architecture;
- a declared viability vector;
- explicit compute-versus-imitation selection boundaries;
- a recursive maintenance interpretation in which the process that maintains the commons may itself require sensing and return-loop maintenance.
Whether that recursive framing adds scientific value beyond established monitoring/reputation theory remains a cross-substrate research question.
Experiment 12 does not claim novelty for fungi as adaptive transport networks or for flow-dependent network reinforcement.
Key antecedents include:
- Heaton et al. (2010), Growth-induced mass flows in fungal networks, Proceedings of the Royal Society B, DOI 10.1098/rspb.2010.0735. In Phanerochaete velutina, cords predicted to carry faster/larger currents were significantly more likely to increase in size; the authors propose that fluid velocity provides a local signal carrying quasi-global information about cord function.
- Fricker et al. (2008), The interplay between structure and function in fungal networks, Topologica 1:004, DOI 10.3731/TOPOLOGICA.1.004, describing selective reinforcement of transport routes, recycling of redundant mycelium, dynamic flux switching, and resilience to damage/grazing.
- Fricker et al. (2017), The Mycelium as a Network, Microbiology Spectrum, DOI 10.1128/microbiolspec.FUNK-0033-2017, reviewing the reciprocal relation between network structure and resource flows across scales.
- Aguilar-Trigueros et al. (2022), Network traits predict ecological strategies in fungi, ISME Communications 2:2, DOI 10.1038/s43705-021-00085-1, quantifying connectivity, construction cost, transport efficiency and robustness trade-offs.
- Nonlinear positive flux reinforcement is also established in adaptive biological-network models, especially the related Physarum literature.
The Experiment-12 power-law allocation is therefore a deliberately minimal cross-substrate falsification model, not a proposed novel fungal mechanism. Its use here is to test whether the repository's maintenance abstractions survive when explicit agents, monitoring and economic rewards are absent.
The synthesis document VIABILITY_OUGHT.md does not claim novelty for the
general move from a constraint set to admissible controls.
- Aubin's viability theory and later work with Bayen and Saint-Pierre develop viability kernels, invariance and regulation for systems constrained to remain in prescribed regions.
- Di Paolo (2005), Autopoiesis, adaptivity, teleology, agency, DOI 10.1007/s11097-005-9002-y, develops adaptivity as regulation relative to conditions of viability and connects this to organismic value/normativity.
- Moreno & Mossio (2015), Biological Autonomy, develop an organizational account based on mutually maintaining constraints.
- Bolton & Sustar (2022), Regulation and the Normativity Problem, DOI 10.1080/02698595.2022.2149050, explicitly discusses normativity through biological regulation.
- Cusimano & Sterner (2020), The Objectivity of Organizational Functions, DOI 10.1007/s10441-019-09365-9, provides an important critique of organizational-function accounts and the risk of arbitrary redescription.
- Decentralized control and multi-agent systems already study preservation of network connectivity and collective constraints.
The repository's open question is therefore not whether viability constraints or biological normativity exist. It is whether a common, falsifiable decomposition of distributed maintenance processes remains useful across engineered, biological and eventually social substrates without smuggling in agent-specific concepts.