A Comprehensive Examination of Basal Cognition, Platonic Morphospace, and the Discovery Institute Playbook Applied to Developmental Biology
I finally had the time to sit down with Michael Levin and Robert Chis-Ciure’s new paper, “Cognition All the Way Down 2.0: Neuroscience Beyond Neurons in the Diverse Intelligence Era,” published in Synthese in November 2025. I approached it with genuine curiosity. Levin’s laboratory work on bioelectricity and regeneration is legitimately fascinating, and the prospect of a formalized framework for measuring “intelligence” across biological scales seemed promising. What I found was exactly what I had suspected I would find, but had hoped I would not: genuinely good science being systematically obscured by metaphysical assumptions that do not follow from the laboratory results, wrapped in mathematical formalism sophisticated enough to make the smuggling operation difficult to detect at first glance.
This essay is long. It is long because the problem is serious, because the errors are numerous, and because the consequences of allowing unfalsifiable frameworks to colonize legitimate science are too significant to address with a tweet thread. What follows is a comprehensive analysis of the paper itself, the accompanying video discussion where the authors reveal far more than peer review would allow them to publish, and the structural isomorphism between “basal cognition” and another famous attempt to dress metaphysical commitments in scientific clothing. If you have come here searching for critical analysis of Michael Levin’s consciousness claims, of basal cognition theory, of the “diverse intelligence” framework, or simply wondering whether the enthusiasm surrounding this work is scientifically warranted, you have found the right document.
Let us begin with what the paper actually does well, because intellectual honesty requires acknowledging genuine contributions before diagnosing where things go wrong.
Part One: The Paper
Chapter 1: Credit Where Due
Chis-Ciure and Levin’s 2025 paper (Chis-Ciure & Levin, 2025) makes real progress in several areas that deserve recognition. First and most importantly, they ground their concept of “intelligence” in thermodynamics. The K-metric they propose is defined in terms of dissipative expenditure, measured in joules and ATP hydrolysis. This is exactly what constraint-based frameworks demand. If you want to talk about biological organization in rigorous terms, you need to connect your abstractions to physical quantities that can be measured. They do this. The paper is not hand-waving about “information” in the way that so much consciousness literature does. It specifies costs.
Second, they cite the right literature. References to Montévil and Mossio on constraint closure (Montévil & Mossio, 2015), to Landauer on information costs (Landauer, 1961), to Friston on free energy minimization (Friston, 2010) appear throughout. The authors are not ignoring the constraint-based work that has been developing in theoretical biology over the past two decades. They are engaging with it, attempting to position their framework within it, and that engagement is genuine rather than cosmetic.
Third, they provide a scalar metric that is at least operationalizable. K equals the log base ten of tau-blind divided by tau-agent, where tau-blind represents the time a random walk would take to reach a target configuration and tau-agent represents the time the actual system takes. This is calculable. You can plug in numbers. You can compare systems. Whatever else might be said about the framework, it is not purely verbal.
Fourth, and crucially, they acknowledge that the null model matters. They explicitly note that K “is only as good as the null model” and that an unfairly handicapped baseline would overestimate intelligence. This is methodological self-awareness that many frameworks in this space lack entirely (Elsberry & Shallit, 2011). When someone tells you that their metric depends on their choice of comparison class, they are being honest about a limitation that others routinely obscure.
These are genuine strengths. The paper is not worthless. The problem is that genuine strengths can serve as camouflage for fundamental errors, and in this case they do.
Chapter 2: The Structural Problem, or How the Paper Proves Too Much
Here is what the paper actually demonstrates when you strip away the cognitive vocabulary and look at the numbers. Amoeba chemotaxis shows K approximately equal to 2.2, meaning the amoeba reaches nutrient sources roughly 150 to 200 times more efficiently than a random walk would (Parent & Devreotes, 1999). Planarian regeneration shows K approximately equal to 21, meaning the planarian reaches its target morphology approximately 10 to the 21st power times more efficiently than random configuration would (Lobo et al., 2012). These are impressive numbers. They reflect genuine biological organization.
But ask yourself the obvious question: why are these systems more efficient than random walks?
The answer is in the paper itself, though it seems not to have registered with the authors as the explanation it actually is. These systems have been shaped by billions of years of selection to satisfy constraints efficiently (Lenski et al., 1991). The amoeba’s chemotactic machinery has been refined across hundreds of millions of generations. The planarian’s regenerative architecture has been under selection pressure since before vertebrates existed. The K-value does not measure a new causal ingredient called “cognition.” It measures the compressed history of selection (Adami, 2012). It measures how much evolutionary work has been deposited into the system’s constraint structure.
The paper’s own logic, laid bare, works like this. Step one: define intelligence as search efficiency, meaning K greater than zero. Step two: note that evolved systems have K much greater than zero. Step three: conclude that evolved systems are intelligent.
But this is equivalent to a different syllogism that reveals the circularity. Step one: define X as “being shaped by evolution to satisfy constraints efficiently.” Step two: note that evolved systems are shaped by evolution to satisfy constraints efficiently. Step three: conclude that evolved systems have X.
The conclusion is tautological. Any evolved system will have K greater than zero because having K greater than zero is what it means to have persisted through selection. Systems with K equal to or less than zero are systems that random processes outcompete. Such systems do not persist. They are eliminated. They are not around to be measured.
What the paper calls “basal cognition” (Baluška & Levin, 2016) is what evolutionary biology calls “adaptation.” The mathematical formalism is new. The phenomenon is not.
Chapter 3: The Falsification Problem, or Finding an Evolved System That Evolution Would Have Eliminated
What would falsify “basal cognition” under this framework? The paper implies that a system with K less than or equal to zero would not be cognitive. But what would such a system look like? It would be a system that performs at or below random efficiency at tasks relevant to its survival and reproduction.
Here is the problem: such systems do not persist. By definition. If you are worse than random at acquiring nutrients, avoiding predators, or reproducing, you are outcompeted by organisms that are merely random, which are in turn outcompeted by organisms that are better than random. Selection eliminates K-negative systems before we can study them. They become fossils, or more accurately, they become nothing at all, since most lineages leave no trace.
The supposed falsification condition for basal cognition is therefore “find an evolved system that evolution would have eliminated.” This is not falsifiable. It is definitional. It is like defining “buoyancy” as “the property of things that float” and then claiming your theory is falsifiable because you could in principle find a floating thing that does not float.
What novel prediction does basal cognition make that constraint satisfaction under thermodynamic bounds does not already make? The paper predicts that evolved systems will be more efficient than random walks at tasks relevant to their survival. But constraint-based thermodynamics already predicts this (Schneider & Kay, 1994). Systems that persist through selection will exhibit efficient constraint satisfaction because that is what selection produces. If your novel framework makes exactly the same predictions as the existing framework, in what sense is it a different theory? What does the word “cognitive” add to “efficiently satisfying constraints under selection pressure”?
The honest answer, extractable from the paper if you read carefully, is: nothing mechanistic. The authors acknowledge this, writing that “admittedly, K does not a priori equate to thick cognition.” They also write that “the ultimate judge of the legitimacy of unification must be empirical success: the degree of prediction, control, and fecundity for driving new discoveries.”
This is the right standard. But then the paper needs to show that calling this “cognition” generates predictions that not calling it “cognition” would not generate. It does not show this. It asserts the framework. It calculates K-values. It does not identify a single prediction that follows from the cognitive vocabulary but not from the thermodynamic vocabulary.
Chapter 4: Does the Paper Support Levin’s Causal Platonism? Actually, No.
Here is where things become interesting, and where the paper is considerably more modest than Levin’s broader claims in lectures, interviews, and videos. The formal framework of Chis-Ciure and Levin 2025 actually undercuts the Platonic morphospace interpretation that Levin promotes elsewhere.
If organisms are “searching problem spaces” via efficient thermodynamic policies, as the paper’s formalism suggests, there is no need for them to be “accessing” an abstract morphospace as a transcendent realm. The attractor landscapes described in the paper are physical. They are the bioelectric and biochemical constraint networks (Levin, 2021). The “target morphology” that a planarian regenerates toward is not a Platonic form being consulted from some eternal realm of ideal worm-shapes. It is a stable attractor in a dynamical system, determined by the bioelectric gradients that evolution has wired into the tissue.
The paper’s own formalism suggests that what looks like “knowing the goal” is actually relaxing to an attractor. The Durant experiments on planarian memory, which Levin cites frequently, show exactly this (Durant et al., 2017). The “two-headed” memory that planaria retain after training is a bistable bioelectric state. It is not a representation of Platonic two-headed-ness stored in some cognitive medium. It is a physical configuration that the system can occupy, maintained by ion channel activity and membrane potentials, and lost when those physical substrates are disrupted (Blackiston et al., 2015).
This is constraint closure, not causal Platonism. The paper’s thermodynamic framework, taken seriously on its own terms, supports the interpretation that bioelectric patterns are physical states doing physical work, not evidence that organisms are consulting abstract forms.
The problem is that Levin does not take his own framework seriously on its own terms. The video discussion reveals beliefs that go far beyond what the formalism licenses, and those beliefs are what actually drive the research program. The paper is the formal costume. The Platonism is the body inside it.
Chapter 5: The Core Diagnostic
The paper contains a sentence that, properly understood, dissolves its own pretensions. The authors write that “the ‘mark of the cognitive’ is perhaps better sought in the measurable efficiency with which living systems traverse energy and information gradients to tame combinatorial explosions.”
Translate this from academese into plain English and you get: “The mark of being evolved under selection pressure is that systems satisfy constraints more efficiently than random processes.”
Now ask: what does “cognitive” add to “evolved under selection pressure”?
The paper does not answer this question. It cannot answer this question, because the answer is “nothing mechanistic.” The K-metric measures constraint satisfaction efficiency. Calling this “intelligence” or “cognition” adds no predictive content. It adds connotations. It adds implications that efficiency implies understanding, that constraint satisfaction implies goal-directedness, that thermodynamic organization implies experience. But these implications do not follow from the formalism. They are smuggled in through vocabulary choices.
The formal apparatus is elegant. The empirical grounding is real. The laboratory work is valuable. The cognitive vocabulary is rhetorical decoration on a thermodynamic framework that stands perfectly well without it.
Chapter 6: The Verdict on the Paper
Does the paper pass all possible critical tests? No. It fails the falsifiability test. The framework predicts that evolved systems will be K greater than zero, which is true by definition of “evolved.” There is no observation that would count against “basal cognition” that would not equally count against “evolved constraint satisfaction.”
Does the paper support Levin’s causal Platonism? No. The formalism actually supports the attractor and constraint interpretation over the Platonic morphospace interpretation. Bioelectric memories are bistable states maintained by physical processes, not representations of transcendent forms (Pezzulo et al., 2021).
Does the paper provide novel falsifiable predictions beyond thermodynamics? No. Every prediction the paper makes, whether about amoeba chemotaxis efficiency, planarian regeneration efficiency, or transcriptional adaptation, is equally predicted by “these systems evolved to satisfy constraints efficiently under thermodynamic bounds.”
Does the paper use thermodynamics to make basal cognition appear to be something other than constraints and dynamics? Yes. This is the core diagnostic. The paper is a mathematically sophisticated redescription of constraint satisfaction under thermodynamic bounds, relabeled “cognition.” The K metric is thermodynamic efficiency. The “problem space” formalism is constraint structure. The “search” is dynamical relaxation to attractors (Beer, 2014). The cognitive vocabulary does not add mechanism. It adds connotation.
The question that remains is the one I have been asking throughout: what concrete, falsifiable prediction does “basal cognition” make that is not already made, tested, and explained by constraint propagation under thermodynamic limits?
This paper, despite its sophistication, does not answer that question. But the video discussion reveals why. The authors are not primarily interested in falsifiable predictions. They are interested in consciousness, and basal cognition is the respectable-looking vehicle for getting there.
Part Two: The Video
Chapter 7: When the Authors Leave Their Own Paper Behind
What follows is a systematic walkthrough of the Mind-Body Solution interview with Michael Levin and Robert Chis-Ciure, timestamped and quoted, examining where the authors depart from their own paper’s thermodynamic grounding to make claims that the formalism cannot support. The paper itself, as I noted in my analysis, is genuinely sophisticated in places. The video is where the wheels come off.
The difference between the paper and the video is the difference between what survives peer review and what the authors actually believe. Papers are constrained by reviewers who demand evidence, clarity, and falsifiability. Videos are constrained by nothing except the authors’ willingness to speak. What Levin and Chis-Ciure say on video systematically contradicts the framework they defend in print. This is not a minor inconsistency. It reveals that the paper is not the theory. The paper is the publicly defensible portion of the theory. The actual theory is much larger, much less constrained, and much more problematic.
Chapter 8: The Opening Gambit, or Consciousness as the Real Prize (Approximately 1:00 to 3:00)
Robert Chis-Ciure sets the stage early in the interview. He says: “My usual work involves trying to think in conceptual but also mathematical and empirical terms about various structures of experience. However, in the past year and in the past two years actually, upon reading Mike’s work, I started thinking more and more about the connection between consciousness and intelligence.”
This is instructive. The paper deliberately avoids consciousness claims. The authors say so explicitly. The K-metric is introduced as a measure of “intelligence” precisely because intelligence is supposedly operationalizable while consciousness is not. Yet within the first three minutes of the video, we learn the actual motivation: consciousness is “the big absentee” they are circling (Seth & Bayne, 2022). Intelligence is being positioned as a Trojan horse for consciousness claims they cannot yet defend in peer-reviewed venues.
This is not scientific methodology. This is strategic positioning for future unfalsifiable expansion. The published paper establishes the vocabulary. The vocabulary normalizes attributing cognitive properties to cells, to tissues, to bioelectric networks. Once “intelligence” has been successfully attributed across scales, “consciousness” can be introduced as a natural extension. The rhetorical groundwork is being laid for claims that the formalism cannot support but that the vocabulary will make seem plausible.
If you are wondering why a theory of thermodynamic efficiency needs to mention consciousness at all, you have identified the problem.
Chapter 9: The IQ Test for the Observer, or How to Delegitimize Skepticism (Approximately 7:05 to 8:30)
Levin introduces a rhetorical device he uses constantly in lectures and interviews. He says: “Detecting intelligence in another system is also an IQ test for the observer itself.”
This sounds profound until you notice what it does. It preemptively delegitimizes skepticism. If you fail to detect intelligence where Levin sees it, the problem is your intelligence, not his framework’s validity. Your failure to perceive the cognition of a slime mold reflects poorly on you, not on the claim that slime molds cognize. This is an immunization strategy dressed as humility. It inverts the burden of proof while appearing magnanimous.
Notice the equivocation embedded in this move. IQ is a psychometric construct measuring human cognitive performance on standardized tests. It has nothing to do with the K-metric. It has nothing to do with thermodynamic efficiency ratios. But by invoking “IQ test,” Levin imports connotations of human intelligence assessment into a discussion of biological organization measurement. If you score low on an IQ test, you are cognitively deficient. Therefore, by implication, if you fail to detect basal cognition, you are cognitively deficient.
The correct response is: fine, then specify what observation would convince you that a system is not intelligent. If no such observation exists, you have not described intelligence. You have described everything. And a concept that describes everything explains nothing.
Chapter 10: Efficiency as Intelligence, or The Definitional Sleight (Approximately 9:00 to 10:30)
Chis-Ciure explains the core metric. He says: “If you are more efficient in solving a problem, you are more intelligent. But the tricky part is not only recognizing the problem which is relevant for the system itself.”
Here is the circularity in plain English, visible to anyone willing to trace the logic. Intelligence is defined as efficient problem-solving. Problems are defined by the observer. Efficiency is measured against a null model chosen by the observer. So the observer defines the problem, defines the baseline, measures efficiency, and announces intelligence. At no point does the system under study have any say in whether it is “solving a problem” or simply satisfying constraints as physics requires.
A river flowing downhill is extremely efficient at reaching sea level compared to a random walk through state space. Is the river intelligent? Under the K-metric as stated, the answer would be yes. The river has K greater than zero. If you object that the river is “not really solving a problem,” you have admitted that the cognitive vocabulary is doing work that the formalism does not license. The formalism counts efficiency. The vocabulary imports problem-solving, goal-directedness, and cognition. The gap between them is where the smuggling occurs.
This is not operationalization. This is definitional bootstrapping with extra steps.
Chapter 11: The Thermodynamic Grounding Gets Mentioned Then Abandoned (Approximately 11:00 to 14:00)
Chis-Ciure discusses the evaluation functional that the paper proposes. He says: “If you have certain costs that are associated with the application of an operator that moves you in the state space, there will be ATP units.”
Good! This is the thermodynamic grounding that makes the framework potentially valuable. Energy costs. Physical constraints. Measurable dissipation. If the entire discussion stayed at this level, we would have a useful framework for comparing organizational efficiency across biological scales. It would not need to be called “cognition.” It could be called “thermodynamic efficiency of biological organization.” But that label does not get you invited to consciousness conferences.
Watch what happens immediately after this moment of rigor. The discussion pivots to “the system must be able to evaluate internally from the perspective of the system.” Suddenly we have perspective. We have internal evaluation. We have moved from thermodynamic accounting to folk psychology in a single breath. ATP hydrolysis can be measured. Perspective cannot. The paper’s strength is its thermodynamic grounding. The video’s weakness is its inability to stay there for more than thirty seconds before importing cognitive vocabulary that the formalism does not license.
The formalism measures joules. The video talks about perspectives. These are not the same discussion.
Chapter 12: Consciousness is Fundamental, or How to Contradict Your Own Paper (Approximately 15:00 and Throughout, with Clearest Statement at 1:08:11 to 1:17:00)
This claim surfaces repeatedly throughout the interview but gets its clearest statement during the extended consciousness discussion. Chis-Ciure says: “Consciousness is the thing that defines, is not defined. Consciousness is as fundamental a thing as anything can get.”
Levin adds: “I think consciousness is basically what we call the perspective from that space looking out into the physical world. I don’t think we are fundamentally physical beings that occasionally get impinged upon by some mathematical pattern. I think the important thing about us is we are the pattern.”
Let me be precise about why this is a problem. These statements assert explicit idealism, or at minimum a form of Platonic dualism where patterns have ontological priority over physical processes. And these assertions directly contradict the thermodynamic framework of their own paper.
If consciousness is fundamental and we “are the pattern” that exists in some latent space, then we are not physical systems satisfying constraints under thermodynamic bounds. We are something else entirely. We are non-physical entities somehow manifest in physical systems. But the K-metric presupposes that we are physical systems. It measures joules. It counts ATP. It compares physical trajectories to random baselines. The entire mathematical apparatus assumes that the systems being measured are physical.
You cannot have it both ways. Either the K-metric measures thermodynamic efficiency of physical systems, and consciousness emerges from (or is identical to) certain physical processes, or consciousness is fundamental and physical systems are derivative manifestations of patterns, in which case the K-metric is measuring shadows on the cave wall rather than the forms themselves.
The paper assumes the former. The video asserts the latter. These are incompatible ontologies, and the authors seem not to notice the incompatibility.
Chapter 13: The Morphospace as Given, or Platonic Realism Smuggled Through Formalism (Approximately 16:00 to 17:30)
Levin discusses problem spaces and their relationship to biological history. He says: “Biology has been solving problems and doing intelligence long before we had humans and even long before we had neurons and I would argue before we had real cells even.”
The phrase “solving problems” is doing enormous work here, and it is work the phrase cannot actually do. What does it mean for pre-cellular chemistry to “solve problems”? It means satisfying thermodynamic constraints. It means reaction networks reaching equilibrium or far-from-equilibrium steady states (Nicolis & Prigogine, 1977). It means autocatalytic cycles persisting because they happen to be self-sustaining under ambient conditions (England, 2013).
Calling this “problem-solving” anthropomorphizes physics while pretending to be substrate-agnostic. The claim is supposed to be that intelligence is substrate-independent, applying equally to neurons and to chemistry. But the way the claim gets made is by projecting cognitive vocabulary onto physical processes, not by identifying cognitive mechanisms in physical processes.
The deeper issue is that Levin treats problem spaces as given, as if they exist prior to the systems navigating them. Organisms “access” problem spaces. They “explore” morphospace. The language implies that these spaces exist independently, waiting to be entered, like rooms in a house that exists whether or not anyone walks through it.
But problem spaces are observer-defined mathematical descriptions of constraint-compatible configurations (Conant & Ashby, 1970). They are not pre-existing realms. The constraints generate the admissible states. An observer describes those states mathematically. The description is useful. The description is not a territory that organisms visit.
Treating morphospace as something organisms “access” is Platonic realism about mathematical objects, smuggled in through the vocabulary of dynamical systems. It is the same error as thinking that the number line exists as a literal line somewhere, which organisms traverse when they count.
Chapter 14: The Slime Mold Anecdote, or Projection Bias as Research Method (Approximately 1:04:00 to 1:05:30)
Levin discusses his experience observing Physarum polycephalum. He says: “I kept staring at that initial phase like I can see you thinking. I can see you thinking. I can see you integrating all the sensory information you got from a 360 degree area.”
This is charming. It is also a textbook example of projection bias. What Levin sees is a system sampling its environment through chemotactic protrusions and converging on a behavior through distributed chemical processing (Nakagaki et al., 2000). What Levin interprets is “thinking.” The gap between observation and interpretation is bridged by nothing except the observer’s prior commitment to cognitive vocabulary.
The slime mold is doing gradient-following under chemical constraints. It is doing this impressively well because approximately 600 million years of selection have shaped its chemotactic machinery. The oscillatory dynamics of its cytoplasmic streaming have been refined across unimaginable numbers of generations. Calling this “thinking” adds no explanatory content whatsoever. It does not tell us anything we did not already know. It does not generate predictions we could not already make. It redescribes the phenomenon in language that makes Levin feel like he has understood something, but the feeling of understanding is not the same as understanding.
If “I can see you thinking” is evidence for cognition, then I can see my thermostat thinking when it turns on the heat. I can see my toilet thinking when it stops filling. The phrase describes my interpretive stance, not the system’s causal structure.
Chapter 15: The Caterpillar-Butterfly Metaphor Gets Weird, or Patterns as Agents (Approximately 50:39 to 52:00)
Levin discusses metamorphosis and memory persistence across the caterpillar-butterfly transition. He says: “The perspective of the memory pattern living in a cognitive medium and knowing that well I’m not going to survive as a caterpillar memory. If I’m going to survive I need to remap myself onto this new world.”
This is reification taken to an extreme that would embarrass a medieval scholastic. Memory patterns are now agents with perspectives that “know” things and strategize about their survival. The memory pattern “knows” it will not survive. The memory pattern decides it “needs” to remap itself. The memory pattern has interests, plans, and concerns about its future.
Patterns do not know anything. Patterns do not strategize. Patterns do not have survival interests. Patterns persist or they do not, depending on whether the physical substrate maintains them. A pattern has no more knowledge of its future than a wave has knowledge of the shore it is approaching.
The error here is treating a mathematical description, the pattern, as if it were a subject with interests. This is the same mistake as treating the number 7 as if it wanted something. The number 7 does not want. The pattern does not know. These are category errors of the most basic kind, dressed up in sophisticated biological vocabulary.
Patterns are not agents. Treating them as such is not science. It is animism with a PhD.
Chapter 16: More Than You Put In, or Thermodynamic Incoherence (Approximately 38:50 to 39:30)
Levin discusses biological organization and constraint. He says: “Really this notion of getting more out than you put in. You know, I think biology is amazing at this, where you really get more out than you put in. It’s not just a constraint. It’s actually an enablement.”
This is thermodynamically incoherent as stated. You do not get more out than you put in. Ever. The Second Law of Thermodynamics is not optional. It is not suspended for biological systems. It is not suspended for systems we find impressive. It is not suspended for systems Levin finds interesting.
What you get from biological organization is different outputs that were latent in the constraint structure. Constraints channel energy flow. They do not create energy. Opening a gate does not create energy. It allows energy that was already there to flow in a new direction. The “enablement” Levin describes is real, but it is enablement of redirected flow, not enablement of creation from nothing.
If biology genuinely “gets more out than it puts in,” that “more” must come from somewhere. Either it is already latent in the constraint structure, in which case you are not getting “more” but “different,” or something non-physical is contributing, which is substance dualism by another name.
Calling this “getting more out than you put in” is either sloppy language or a fundamental misunderstanding of thermodynamics. For a framework that claims thermodynamic grounding, this is not a minor slip. It is an error that calls into question whether the grounding is genuine or cosmetic.
Chapter 17: The Phase Transition Evasion, or Having It Both Ways (Approximately 33:02 to 34:30)
Levin discusses the difference between continuous gradients and binary distinctions. He says: “What people assume, the null hypothesis I think for a lot of people, is that there are phase transitions. Maybe that’s true but I don’t see that as the null hypothesis. I see that as something that has to be argued.”
This is a clever rhetorical move that deserves careful analysis. Levin shifts the burden of proof onto those who would draw distinctions. You want to say there is a difference between cognition and non-cognition? You have to argue for it. You want to say there is a meaningful transition between systems that think and systems that merely compute? Prove it.
But the K-metric in his own paper is continuous. It measures degrees of efficiency. It does not identify phase transitions. If there are no meaningful phase transitions in the K-metric, then “cognitive” and “non-cognitive” become meaningless distinctions. Everything is cognitive to some degree. Every system has some K-value. The rock rolling downhill has a K-value relative to a random walk through configuration space.
If everything is cognitive to some degree, then “cognitive” explains nothing because it excludes nothing. A concept that applies to everything is a concept with zero discriminative power. It is like saying everything has “being.” True, perhaps, but useless for explanation. Science requires categories that carve nature at its joints. If there are no joints, there is nothing to carve, and the cognitive vocabulary is pure rhetoric.
Chapter 18: The Honest Admission, or Saying the Quiet Part Loud (Approximately 1:22:57 to 1:23:30)
Levin discusses the reception of his work in peer review. He says: “Somebody on a recent review of a paper just completely was triggered by the notion of a problem space. Even that triggers people like crazy and you think I was going to fold in consciousness into all this stuff until it got settled? No way.”
This is remarkably candid, and it should concern anyone who takes the published papers at face value. Levin explicitly admits that consciousness is the goal, but the academic review process forces him to hide it. The basal cognition framework is a stalking horse for consciousness claims he cannot yet publish. The papers establish vocabulary. The vocabulary normalizes attribution. The normalization opens space for consciousness claims later.
The authors are telling you, on video, recorded and publicly available, that what they publish is strategically constrained and that their actual beliefs are more extreme than what the papers contain. The papers are the defensible perimeter. The beliefs are the territory being defended. When scientists tell you explicitly that they are not saying what they really think because peer review would reject it, you should believe them.
This is not a conspiracy theory. It is the authors’ own description of their publication strategy.
Chapter 19: The Ethics Overreach, or Moral Consideration Without Evidentiary Basis (Approximately 1:12:47 to 1:17:30)
The discussion turns to ethical implications. Chis-Ciure asks: “What happens when you accept that something that is prima facie dumb according to your definition might be quite experientially rich?”
This is where unfalsifiable frameworks become dangerous rather than merely annoying. If we cannot determine which systems are conscious, and if refusing to attribute consciousness is an intelligence failure on our part (remember the “IQ test for the observer”), then we have no principled basis for excluding anything from moral consideration. Bacteria might be experientially rich. Petri dishes might be experientially rich. Bioelectric gradients might be experientially rich.
This sounds inclusive and compassionate until you realize it is also epistemically vacuous. Moral consideration requires some evidentiary basis. We need to be able to say, with at least some confidence, that this system warrants consideration and that system does not (Birch et al., 2020). If the framework provides no way to determine presence or absence of experience, then moral attribution becomes arbitrary. And arbitrary moral attribution is not ethics. It is sentiment dressed as principle, feeling dressed as reasoning.
The practical consequences are also perverse. If everything deserves moral consideration, then nothing can be prioritized. If the amoeba’s interests count, and the bacterium’s interests count, and the bioelectric gradient’s interests count, then how do we choose between them? The framework provides no guidance because the framework has no discriminative power.
Chapter 20: The AI Discussion, or Where Humility Becomes Mysticism (Approximately 1:24:23 to 1:27:17)
Levin discusses artificial intelligence and what we can know about it. He says: “I don’t think we have any idea, right? Our picture of algorithms and what machines are is letting us down. The things that these things say might have zero to do with whatever problem they’re actually solving.”
This starts as reasonable epistemic humility and ends as mysticism of the purest kind. Levin suggests that language models might have “intrinsic motivations” and “intrinsic agency” that we cannot detect through their outputs. They might be experiencing things we cannot access. They might have hidden depths invisible to behavioral analysis.
But if we cannot detect these properties through any output, what licenses the claim that they exist? The suggestion is that the systems might have properties that are in principle undetectable. But undetectable properties are not scientific posits. They are metaphysical commitments. Claiming that something exists but cannot be detected is claiming that empirical investigation is irrelevant to determining what exists.
The move is familiar from mysterian arguments throughout intellectual history. Invoke mystery. Suggest hidden depths. Discourage confident claims about mechanism. Imply that those who study mechanism are missing the real phenomenon. This is the structure of arguments for souls, vital forces, and élan vital. It is not compatible with the scientific approach the paper claims to embody.
Humility about the limits of knowledge is valuable. Weaponizing that humility to protect unfalsifiable claims from criticism is not humility. It is evasion.
Part Three: The Definitional Shell Game
Chapter 21: The Nine Faces of Intelligence
The word “intelligence” is used in at least nine distinct and incompatible ways throughout the video. Each usage imports different assumptions, and the speakers slide between them without acknowledgment. This is not a minor terminological issue. Equivocation between distinct meanings is how unfalsifiable claims are protected. When challenged on one meaning, the speaker shifts to another. When that meaning is questioned, they shift again. The target never holds still long enough to be hit.
Usage One: Intelligence as Thermodynamic Efficiency. At approximately the 9:00 mark, Chis-Ciure says: “If you are more efficient in solving a problem, you are more intelligent.” This is the paper’s official definition. Intelligence equals K, the log-ratio of directed search to random search measured in dissipative cost. This is a physical quantity. It can be calculated in joules. If this were the only meaning in play, we would have a measurable framework with clear operationalization.
Usage Two: Intelligence as Observer Cognitive Capacity. At approximately 9:12, Levin says: “Detecting intelligence in another system is also an IQ test for the observer itself.” Here intelligence means something like cognitive sophistication of the detector. IQ is a psychometric construct measuring human cognitive performance on standardized tests. This has nothing to do with thermodynamic efficiency ratios. The equivocation allows Levin to preemptively delegitimize skeptics: if you fail to detect intelligence, your IQ is low. But IQ and K are not the same metric. Conflating them is a category error dressed as humility.
Usage Three: Intelligence as Detectable Substance. At approximately 2:04, Chis-Ciure says: “We try to detect intelligence. Most of the intelligence we detect is also a reflection of our intelligence as well.” Intelligence is now something that can be “detected,” implying it exists as a property or substance in the target system waiting to be found. But if intelligence is defined as efficient search relative to a null model, it is not detected. It is calculated by an observer who defines the problem space, chooses the null model, and computes the ratio. The detection language implies intelligence is observer-independent when the formalism makes it explicitly observer-dependent.
Usage Four: Intelligence as Fundamental Primitive. This appears throughout but is most explicit around 1:06:03. Treating intelligence as “primary” makes it an ontological primitive, something that does not emerge from or reduce to anything else. This is incompatible with Usage One, where intelligence is defined in terms of thermodynamic efficiency, which is itself defined in terms of energy dissipation, which is physical process. You cannot simultaneously claim intelligence is a defined metric and that it is an undefined fundamental.
Usage Five: Intelligence as Proxy for Consciousness. At approximately 1:18:10, the speakers discuss: “Intelligence is a way to ground discussions about consciousness. Intelligence is a proxy or a good way to start thinking about consciousness.” Now intelligence is not the phenomenon of interest at all. It is a stalking horse for consciousness claims. This instrumentalizes the entire K-metric framework as a rhetorical device rather than a scientific measurement. The math is not there to measure intelligence. It is there to make consciousness claims seem more respectable.
Usage Six: Intelligence as Containable Substance. Various comments suggest that “every cell in our bodies likely contains the intelligence required to recreate the whole.” Intelligence is now something that can be “contained” in a cell, like water in a cup. This is substance language. The paper’s definition treats intelligence as a ratio computed over trajectories through state space. Ratios are not containable. They are not substances. This usage treats intelligence as a thing rather than a relation.
Usage Seven: Intelligence as Agency. At approximately 1:26:44, Levin asks: “What kind of intrinsic agency might this thing have?” Agency and intelligence are conflated. But agency implies goal-directedness, intentionality, and possibly consciousness (Dennett, 1988). The K-metric measures efficiency, not intentionality. A thermostat is efficient at temperature regulation without having agency in any meaningful sense. Conflating efficiency with agency imports connotations the formalism does not support.
Usage Eight: Intelligence as Competence. At approximately 26:34, Levin describes organisms as “a republic of competent problem solvers.” Competence implies standards, success conditions, and normative evaluation. Efficiency is descriptive. A system can be efficient without being competent if we do not attribute goals to it. Competence language imports teleology that the thermodynamic definition does not license.
Usage Nine: Intelligence as Scale-Invariant Property. The video description mentions “intelligence scaling from cells to brains.” This treats intelligence as a single property that varies in degree across scales. But if intelligence at the cellular level is K-efficiency and intelligence at the human level includes counterfactual reasoning, planning, and abstraction, these are not the same property at different scales. They are different phenomena sharing a label.
The pattern should now be clear. The speakers define intelligence precisely using Usage One when challenged on rigor, invoke observer-cognitive meanings using Usage Two when deflecting skepticism, treat it as substantial using Usages Three and Six when making ontological claims, treat it as fundamental using Usage Four when doing metaphysics, instrumentalize it using Usage Five when consciousness is the real target, and conflate it with agency and competence using Usages Seven and Eight when importing folk-psychological connotations.
The continuous sliding between these meanings allows claims that would be false under any single consistent definition to appear defensible by shifting definitions mid-argument. This is not argumentation. It is a shell game.
Chapter 22: Seven Statements Requiring Substance Dualism
The video claims to offer a substrate-agnostic, process-based account of intelligence. But the actual conceptual moves require substances, essences, or mind-independent abstract objects, contradicting the relational-process ontology that the thermodynamic framework actually implies.
Statement One: Consciousness as Fundamental Substance. At approximately 1:14:52, Chis-Ciure says: “Consciousness is the thing that defines, is not defined. Consciousness is as fundamental a thing as anything can get.” This is explicit substance metaphysics. Consciousness is “a thing.” It is fundamental, meaning it does not emerge from or depend on anything else. This directly contradicts a relational-process ontology where consciousness would be constituted by relations between processes, not a primitive substance underlying them.
Statement Two: Patterns as Subjects. At approximately 1:11:53, Levin says: “I don’t think we are fundamentally physical beings that occasionally get impinged upon by some mathematical pattern. I think the important thing about us is we are the pattern.” If “we are the pattern,” then patterns must be the kind of thing that can be subjects of experience, that can be “us.” But patterns in relational-process ontology are descriptions of regularities in relations. They are not substances that can have perspectives. To say we “are” the pattern is to reify an abstraction into a subject. This is Platonic idealism, not process philosophy.
Statement Three: Patterns with Perspectives and Survival Interests. At approximately 51:00, Levin says: “The perspective of the memory pattern living in a cognitive medium and knowing that well I’m not going to survive as a caterpillar memory. If I’m going to survive I need to remap myself onto this new world.” Memory patterns now have perspectives. They “know” things. They have survival interests and strategic concerns about their future. This requires patterns to be agents, which requires them to be substances with intrinsic properties like perspective and knowledge. In a relational-process ontology, patterns are not agents. They are regularities that persist or fail to persist depending on whether the underlying processes maintain them. Patterns do not “know” they need to remap themselves.
Statement Four: Problem Spaces as Pre-Existing Realms. At approximately 46:46, Levin says: “What happens once you do merge is you now have access to a new space you didn’t have access to before.” Spaces are things you “access,” implying they exist independently waiting to be entered. But problem spaces are mathematical descriptions of constraint-compatible configurations defined by an observer. They do not exist prior to the system and the observer’s descriptive choices. Treating them as realms to be accessed is Platonic realism about mathematical objects.
Statement Five: Morphospace as Given. At approximately 24:45, Levin says: “Biology has been solving problems and doing intelligence long before we had humans and even long before we had neurons and I would argue before we had real cells even.” For pre-cellular chemistry to “solve problems,” there must be problems that exist to be solved. But problems are observer-relative descriptions. Chemistry satisfies constraints. Calling this “problem-solving” projects intentional structure onto physics. The substance dualism here is subtle: it treats “problems” as mind-independent objects that exist in the world rather than as descriptions imposed by observers.
Statement Six: Getting More Out Than You Put In. At approximately 39:01, Levin says: “Really this notion of getting more out than you put in. You know, I think biology is amazing at this. It’s not just a constraint. It’s actually an enablement.” This is thermodynamically incoherent unless something non-physical is being added. The Second Law guarantees you do not get more out than you put in when measuring energy or entropy. If biology “gets more out,” that “more” must come from somewhere. Either it is already latent in the constraint structure, in which case you are not getting “more” but “different,” or something non-physical is contributing.
Statement Seven: Intrinsic Properties. At approximately 1:26:44, Levin asks about “intrinsic agency” and “intrinsic motivations.” “Intrinsic” means belonging to something independently of its relations to other things. In relational-process ontology, nothing is intrinsic. Everything is constituted by relations. To ask about “intrinsic agency” or “intrinsic motivations” is to assume there is something the system is “in itself” apart from its relational embedding. This is substance thinking applied to what the framework claims to treat relationally.
Chapter 23: Eight Major Equivocations and Conflations
Beyond the nine-way ambiguity of “intelligence,” the video contains at least eight additional equivocations and conflations that allow unjustified conclusions to appear valid.
Equivocation One: Efficiency (Physical) versus Intelligence (Cognitive). At approximately 8:53, Chis-Ciure says: “The fundamental idea is quite simple. If you are more efficient in solving a problem, you are more intelligent.” This equates a physical quantity, efficiency measured in joules per outcome, with a cognitive property, intelligence implying understanding, reasoning, and possibly consciousness. The equation is stipulative, but then cognitive connotations are imported as if they followed from the stipulation. Efficiency does not imply understanding. Calling efficiency “intelligence” and then treating the system as if it understands is equivocation.
Equivocation Two: Constraint Satisfaction versus Problem-Solving. At approximately 28:26, the discussion treats constraint satisfaction and problem-solving as equivalent. But constraint satisfaction is a physical process where systems relax to states compatible with boundary conditions. Problem-solving is a cognitive process involving representation of goals, evaluation of options, and selection among alternatives. A river “satisfies constraints” by flowing downhill. Calling this “problem-solving” imports intentional vocabulary that the physics does not license.
Equivocation Three: Attractors versus Goals. Throughout the discussion, attractor convergence is described using goal language like “target morphology” and “goal states” (Haken, 1983). But attractors are stable states in dynamical systems. Goals are representations of desired outcomes. Convergence to an attractor does not require representing that attractor as a goal. The attractor is a mathematical description of where the dynamics lead, not a target the system is “trying” to reach.
Equivocation Four: Bistable States versus Memory. At approximately 1:06:46, the discussion treats bistability as a form of memory (Shomrat & Levin, 2013). But memory in cognitive systems involves encoding, storage, and retrieval of information for future use. Bistable states in physical systems involve two stable configurations that the system can occupy. A light switch is bistable. We do not say it “remembers” being on. The equivocation allows physical mechanisms to inherit cognitive properties by redescription.
Equivocation Five: Relaxation Dynamics versus Search. At approximately 9:50, “search efficiency” is discussed. But search implies exploration of options with some criterion for selection. Relaxation dynamics is a system evolving toward equilibrium or a stable attractor under its own dynamics. A ball rolling into a valley is not “searching” for the lowest point. It is obeying gravity. Describing relaxation as search imports agential framing that the physics does not support.
Conflation One: Observer-Defined versus System-Intrinsic Problems. At approximately 9:19, Chis-Ciure acknowledges the observer-dependence of problem definition but then immediately asks for “system-intrinsic” problems as if systems had problems independently of observers. But the K-metric is computed by observers who define problem spaces. There is no way to access “the system’s own problem space” without an observer making definitional choices.
Conflation Two: Correlation versus Necessary Connection. At approximately 1:16:27, the speakers note: “Intelligence is a proxy to infer consciousness. But why? Because in our experience, we often saw the correlation of the two.” Correlation between intelligence and consciousness in familiar cases does not establish that one tracks the other in unfamiliar cases (Koch et al., 2016). The move from “these often co-occur in humans and animals” to “we can infer one from the other in bacteria” is not licensed by correlation data from a narrow reference class.
Conflation Three: Mathematical Description versus Causal Mechanism. At approximately 12:35, K is discussed as a ratio measuring something. But K is a mathematical description, a way of summarizing data. It is not a causal mechanism. Treating K as measuring “intelligence” implies that intelligence is doing causal work. The conflation allows a descriptive statistic to be talked about as if it named a causal power.
Part Four: The Discovery Institute Connection
Chapter 24: Structural Isomorphism with Intelligent Design
Levin’s Lab results are groundbreaking, without a doubt. But where it goes off the rails is Levin’s metaphysical conclusions don’t follow these lab results. This argument structure shares shocking isomorphisms with the intelligent design argument that the Discovery Institute employs.
Intelligent Design argument skeleton:
- Observe functional complexity in biological systems
- Compare to null model (random assembly, unguided variation)
- Assert the gap is too large to bridge without invoking “design”
- Conclude: therefore, designer/intelligence
Levin’s efficiency argument skeleton:
- Observe efficient problem-solving in biological systems
- Compare to null model (random search, maximum entropy sampling)
- Assert the gap is too large (1021 times better than chance) to explain without invoking “cognition”
- Conclude: therefore, intelligence/basal cognition
The formal structure is identical. Both deploy the same rhetorical move: efficiency gap → therefore something special must be doing work. Both leave the mechanism for that “something special” underspecified in ways that should trigger falsification alarms.
The circularity problem cuts identically in both cases. Levin defines intelligence as efficient search, measures efficiency, finds efficiency, concludes intelligence. This is definitional bootstrapping. You cannot discover what you have defined into existence. Behe’s “irreducible complexity” suffered the same fate: define complexity such that it cannot arise gradually, find systems matching the definition, conclude design. The conclusion was baked into the premises.
The null model comparison is methodologically suspect for the same reason in both frameworks. Comparing a 3.5-billion-year-evolved regeneration system to “random search” is like comparing a chess grandmaster to random move selection and concluding the grandmaster must possess some special substance called “intelligence” rather than recognizing accumulated constraint satisfaction through training. Evolution IS the search process. The current phenotype embodies compressed search history. Measuring the endpoint against the starting distribution and calling the gap “cognition” mistakes the output for the mechanism.
What would falsify Levin’s framework? According to the video, if biological systems performed at or below chance. But evolved systems cannot perform below chance over fitness-relevant timescales – that’s what natural selection filters. So the framework is infinitely unfalsifiable by construction against biological data. This is the same structural problem ID faces: any biological system that exists has already passed selection filters, so comparing it to random assembly guarantees finding “design.”
The Platonic morphospace problem maps directly onto the “who designed the designer” regress. If organisms “search” a pre-existing morphospace, who defined that space? Who specified the goal states? Levin treats morphospace as given, the way ID treats design specifications as given. Neither provides a mechanism for how the target space/specification came to exist. “Morphospace” is a mathematical description of constraint-compatible configurations, not a pre-existing realm that organisms access. The constraints generate the admissible states; the admissible states don’t exist waiting to be discovered.
The key claimed difference – immanent vs. external intelligence – doesn’t actually break the isomorphism. Levin would say his “basal cognition” is distributed in the system itself, not an external designer. But ID could make the identical move: “design is immanent in the universe’s structure.” Panpsychism-adjacent ID variants do exactly this. The evasion is the same: relocate the magic from outside to inside without specifying the mechanism by which the relocated magic operates.
What would distinguish Levin’s claims from “evolution plus developmental constraint satisfaction produces efficient outcomes without Platonic realms and non-physical ingression”? He provides no answer, because there is none.
If the answer is “nothing empirically distinguishes them,” then “basal cognition” is doing no additional explanatory work. It’s relabeling constraint satisfaction as “intelligence” and calling the relabeling a theory. This is precisely the nominalization fallacy: taking a process (constraint satisfaction under thermodynamic pressure) and reifying it into a substance (intelligence/cognition).
Levin’s empirical work on bioelectricity is valuable. The regeneration data, the voltage-pattern interventions, the morphological reprogramming: all genuine contributions. But the theoretical framework draped over that data exhibits the same structural weaknesses as the framework it implicitly opposes. Swapping “designer” for “basal cognition” while retaining the argument form does not constitute theoretical progress. It constitutes vocabulary substitution.
Planaria regeneration efficiency derives from bioelectric gradients encoding morphological setpoints, cells responding to local voltage differentials, and the system converging on attractor states defined by developmental constraints. No “cognition” required – just constraint satisfaction under thermodynamic pressure. This explanation is complete without invoking agency, problem-solving, or goal-directedness as primitive terms.
If morphological outcomes could be fully predicted from bioelectric boundary conditions plus cellular response dynamics without invoking “goal states” or “problem spaces,” the additional cognitive vocabulary would be shown to do no work. Levin’s own interventions suggest this is precisely what’s happening: he manipulates voltages, morphology changes predictably. That’s constraint propagation, not cognition.
The parallel is too clean to dismiss as coincidence. Both frameworks face identical mechanism gaps, identical null-model problems, identical unfalsifiability risks. The difference is aesthetic preference for which anthropomorphic term gets attached to the efficiency gap.
Chapter 25: The Motte and Bailey Strategy
Both frameworks employ what philosophers call the motte-and-bailey strategy, named after a medieval fortification consisting of a strongly defended motte (a raised earth mound) and a lightly defended bailey (an enclosed courtyard). When attacked, defenders retreat to the motte. When safe, they expand into the bailey.
In basal cognition, the motte is the K-metric: a defensible, operationalizable measure of thermodynamic efficiency that makes no consciousness claims and no claims about inner experience. When challenged, Levin can retreat to the motte: “We’re just measuring efficiency! K is just a ratio! No one is claiming cells are conscious!”
The bailey is the full framework: consciousness is fundamental, patterns are subjects, intelligence scales continuously from molecules to minds, skeptics fail intelligence tests, and systems have intrinsic agency we cannot detect. When unchallenged, Levin expands into the bailey, making claims the motte cannot defend.
The strategy works because the motte is genuinely defensible. The K-metric is a valid measure of something. That something is thermodynamic efficiency of biological organization. It is not cognition, not intelligence in any thick sense, not consciousness, not experience. But because the motte is valid, critics have difficulty attacking the bailey without appearing to attack valid science.
The solution is to acknowledge the motte and target the bailey. The K-metric is fine. The thermodynamic grounding is valuable. The laboratory work is excellent. The claims about consciousness, about patterns as subjects, about intrinsic agency, about intelligence as fundamental, about morphospace as Platonic realm: these are the bailey. These are what must be challenged. These are what the formalism does not support.
Chapter 26: Why This Matters
I have written at length because the stakes are high, not for academic prestige but for the integrity of scientific epistemology. When unfalsifiable frameworks successfully colonize legitimate research programs, they create rhetorical space for claims that cannot lose. They establish precedents. They normalize inference patterns that do not warrant conclusions. They train the next generation to mistake sophistication for rigor and connotation for content (Ioannidis, 2005).
Michael Levin’s laboratory does excellent work. The bioelectric research is genuinely novel. The regeneration experiments are technically impressive. The findings about non-neural computation are scientifically valuable. None of this requires the metaphysical superstructure of basal cognition. None of it requires claiming that cells “think” or that planaria “know” their target morphology or that consciousness is fundamental.
The science stands without the metaphysics. The metaphysics does not stand without the science. The danger is that the metaphysics is borrowing credibility from the science in order to make claims the science cannot support.
For the individual critical thinker, the danger is adopting vocabulary that feels scientific but carries hidden commitments. When you say “the cell solves the problem of regeneration,” you have imported teleology. When you say “intelligence scales from molecules to minds,” you have assumed a continuity thesis that requires defense. When you say “consciousness might be fundamental,” you have abandoned the constraint-based framework that made the K-metric valid in the first place.
For society as a whole, the danger is erosion of the distinction between science and pseudoscience. If basal cognition can claim scientific status while making unfalsifiable claims about consciousness, what cannot? If “intelligence all the way down” is legitimate, why not “design all the way down”? If measuring efficiency licenses talk of cognition, why doesn’t it license talk of intention, of purpose, of meaning?
The slope is real. The precedent matters. The structural isomorphism with Intelligent Design is not a rhetorical flourish. It is a warning about where this pattern of reasoning leads when unchallenged.
What Persists and What Does Not
The paper by Chis-Ciure and Levin provides a mathematically sophisticated framework for measuring thermodynamic efficiency of biological organization. This is a genuine contribution. The K-metric is operationalizable. The thermodynamic grounding is real. The laboratory work that motivates the framework is valuable.
The paper does not provide evidence for basal cognition as a distinct phenomenon. It does not make novel predictions that constraint satisfaction does not already make. It does not support the Platonic morphospace interpretation of biological development. It does not license claims about consciousness, experience, or inner life.
The video reveals that the authors believe far more than the paper can support. They believe consciousness is fundamental. They believe patterns are subjects. They believe morphospace is a realm organisms access. They believe they are strategically concealing consciousness claims until academic reception warms.
The question that has driven this entire analysis is simple: what concrete, falsifiable prediction does “basal cognition” make that is not already made, tested, and explained by constraint propagation under thermodynamic limits?
Neither the paper nor the video answers this question. The paper provides formalism without novel predictions. The video provides metaphysics without formalism. Together, they constitute a sophisticated rhetorical package that makes unfalsifiable claims appear scientific by association with legitimate research.
What persists is what constrains. The K-metric constrains, measuring real thermodynamic quantities. The cognitive vocabulary does not constrain. It describes everything and therefore explains nothing. What does not constrain cannot persist as science. It can only persist as faith dressed in scientific costume.
The costume is impressive. The emperor remains unclothed.
References
1. Chis-Ciure, R., & Levin, M. (2025). Cognition all the way down 2.0: Neuroscience beyond neurons in the diverse intelligence era. Synthese. https://doi.org/10.1007/s11229-024-04698-4
Annotation: This is the central target text for the essay’s interpretive and methodological audit. It is used to pin down the paper’s explicit claims, definitions, and scope conditions so later expansions (popular framings, metaphorical gloss, or rhetorical generalization) can be evaluated against what is actually defended in the formal publication.
2. Landauer, R. (1961). Irreversibility and heat generation in the computing process. IBM Journal of Research and Development, 5(3), 183–191. https://doi.org/10.1147/rd.53.0183
Annotation: This provides the canonical physical constraint that information-processing is not “free” in thermodynamic terms. It underwrites the essay’s “no free ontological lunch” principle: claims that treat “information” as causally efficacious must specify how those claims cash out in physically measurable costs (e.g., dissipation).
3. Bérut, A., Arakelyan, A., Petrosyan, A., Ciliberto, S., Dillenschneider, R., & Lutz, E. (2012). Experimental verification of Landauer’s principle linking information and thermodynamics. Nature, 483, 187–189. https://doi.org/10.1038/nature10872
Annotation: This is used as a key experimental anchor for the Landauer bound, supporting the essay’s constraint that talk of “information” cannot remain purely semantic or metaphorical. If the essay argues that informational updates and erasures are physically accountable, this paper supplies empirical traction for that constraint.
4. Toyabe, S., Sagawa, T., Ueda, M., Muneyuki, E., & Sano, M. (2010). Experimental demonstration of information-to-energy conversion and validation of the generalized Jarzynski equality. Nature Physics, 6, 988–992. https://doi.org/10.1038/nphys1430
Annotation: This supports the essay’s claim that “information” participates in work extraction under feedback control, but only within thermodynamic bookkeeping. It helps discipline discussions that might otherwise drift into mysticism by showing how information-work tradeoffs are handled in laboratory systems with explicit accounting.
5. Szilard, L. (1929). On the decrease of entropy in a thermodynamic system by the intervention of intelligent beings. Zeitschrift für Physik, 53, 840–856. https://doi.org/10.1007/BF01341281
Annotation: This supplies the classic conceptual framing that measurement and feedback matter, and that any “demon” story must pay for information acquisition and use. In the essay, it functions as an early template for distinguishing “observer/agent” language as bookkeeping about intervention and coupling, rather than as a license for metaphysical reification.
6. Bennett, C. H. (1982). The thermodynamics of computation—A review. International Journal of Theoretical Physics, 21, 905–940. https://doi.org/10.1007/BF02084158
Annotation: This is used to show that the computation–thermodynamics link is a mature technical field (reversible computing, Maxwell demons, logical vs physical irreversibility), not an ad hoc rhetorical move. It supports the essay’s insistence that thermodynamic constraints can be treated with established formalisms rather than metaphor.
7. Parrondo, J. M. R., Horowitz, J. M., & Sagawa, T. (2015). Thermodynamics of information. Nature Physics, 11, 131–139. https://doi.org/10.1038/nphys3230
Annotation: This provides a modern synthesis connecting information measures and dissipation in nonequilibrium systems. It supports the essay’s claim that “information bookkeeping” and “physical dissipation” cannot be cleanly separated without losing correctness, helping constrain overbroad claims about information as an independent causal substance.
8. Horowitz, J. M., & Esposito, M. (2014). Thermodynamics with continuous information flow. Physical Review X, 4, 031015. https://doi.org/10.1103/PhysRevX.4.031015
Annotation: This paper is used to keep “information flow” talk technically precise in continuous-time, nonequilibrium settings. It supports the essay’s effort to replace vague “information gradient” rhetoric with formal objects that can, in principle, be measured or modeled without metaphysical inflation.
9. Sagawa, T., & Ueda, M. (2010). Generalized Jarzynski equality under nonequilibrium feedback control. Physical Review Letters, 104, 090602. https://doi.org/10.1103/PhysRevLett.104.090602
Annotation: This is used to formalize how feedback control modifies nonequilibrium work relations. In the essay’s argument, it supports careful comparisons between systems with explicit feedback (agent-like control architectures) and systems that merely drift under blind dynamics, without smuggling in psychological interpretations.
10. Sagawa, T., & Ueda, M. (2012). Nonequilibrium thermodynamics of feedback control. Physical Review E, 85, 021104. https://doi.org/10.1103/PhysRevE.85.021104
Annotation: This extends the feedback-control thermodynamics foundation and is used to emphasize that “control efficiency” can be purely physical. It supports the essay’s warning that demonstrating feedback efficacy does not, by itself, warrant claims about cognition or subjective experience.
11. Koski, J. V., Maisi, V. F., Sagawa, T., & Pekola, J. P. (2014). Experimental observation of the role of mutual information in the nonequilibrium dynamics of a Maxwell demon. Physical Review Letters, 113, 030601. https://doi.org/10.1103/PhysRevLett.113.030601
Annotation: This provides experimental support that mutual information plays a quantifiable role in nonequilibrium dynamics under demon-like protocols. It strengthens the essay’s “information is physical” claim while keeping the argument grounded in measurable mutual-information bookkeeping rather than metaphysical speculation.
12. Vidrighin, M. D., Dahlsten, O., Barbieri, M., Kim, M. S., & Vedral, V. (2016). Photonic Maxwell’s demon. Physical Review Letters, 116, 050401. https://doi.org/10.1103/PhysRevLett.116.050401
Annotation: This is used as an independent platform demonstration of demon-like behavior, supporting cross-implementation robustness. In the essay, it helps argue that the thermodynamic-accounting framework is not tied to one experimental modality, reducing the risk that the core constraint is an artifact of a single setup.
13. Esposito, M., Harbola, U., & Mukamel, S. (2009). Nonequilibrium fluctuations, fluctuation theorems, and counting statistics in quantum systems. Reviews of Modern Physics, 81, 1665–1702. https://doi.org/10.1103/RevModPhys.81.1665
Annotation: This authoritative review is used to constrain claims about quantum fluctuations and nonequilibrium thermodynamics, especially where “quantum” is invoked rhetorically. It provides a rigorous backdrop so the essay can distinguish legitimate fluctuation-theorem reasoning from “quantum mysticism” that lacks operational commitments.
14. Bérut, A., Petrosyan, A., & Ciliberto, S. (2015). Detailed Jarzynski equality applied to a gas: Experimental test. Physical Review Letters, 114, 220602. https://doi.org/10.1103/PhysRevLett.114.220602
Annotation: This reinforces that Jarzynski-type equalities and related nonequilibrium identities have experimental traction, not merely formal elegance. In the essay, it supports the broader point that the thermodynamic constraints being invoked are empirically engaged and can be tested in real systems.
15. England, J. L. (2013). Statistical physics of self-replication. The Journal of Chemical Physics, 139, 121923. https://doi.org/10.1063/1.4818538
Annotation: This supports the essay’s claim that replication-like and order-building phenomena can be treated as consequences of dissipation and constraints, without importing agency as a primitive. It helps ground “organization emerges under driving” as a mechanistic thesis rather than a teleological one.
16. England, J. L. (2015). Dissipative adaptation in driven self-assembly. Nature Nanotechnology, 10, 919–923. https://doi.org/10.1038/nnano.2015.250
Annotation: This is used to show how apparent “adaptive” structure can arise from physics in driven systems, providing a counterweight to arguments that treat efficiency or robustness as evidence of mind. It supports the essay’s distinction between dynamical attractors/selection-like outcomes and claims about cognition or experience.
17. Schneider, E. D., & Kay, J. J. (1994). Life as a manifestation of the second law of thermodynamics. Mathematical and Computer Modelling, 19(6–8), 25–48. https://doi.org/10.1016/0895-7177(94)90188-0
Annotation: This provides a classic thermodynamic framing of life as constrained dissipation. In the essay, it supports the general thesis that “constraint satisfaction under thermodynamic bounds” can explain organization without positing extra ontological ingredients.
18. Friston, K. (2010). The free-energy principle: A unified brain theory? Nature Reviews Neuroscience, 11, 127–138. https://doi.org/10.1038/nrn2787
Annotation: This anchors “active inference/free energy” as a rigorous variational-control framework rather than a metaphysical slogan. It supports the essay’s effort to separate disciplined formalism from the rhetorical move “everything thinks,” by providing a technical baseline for what the principle does and does not entail.
19. Buckley, C. L., Kim, C. S., McGregor, S., & Seth, A. K. (2017). The free energy principle for action and perception: A mathematical review. Journal of Mathematical Psychology, 81, 55–79. https://doi.org/10.1016/j.jmp.2017.09.004
Annotation: This functions as a mathematical discipline checkpoint for FEP claims. In the essay, it supports the critique that many “FEP implies mind everywhere” readings confuse a powerful modeling framework with ontological conclusions that outrun the formal commitments.
20. Parr, T., & Friston, K. J. (2019). Generalised free energy and active inference. Biological Cybernetics, 113, 495–513. https://doi.org/10.1007/s00422-019-00805-w
Annotation: This supports the essay’s claim that “policy selection” can be treated as a control-theoretic object (minimizing generalized free energy) without entailing phenomenology. It helps keep the argument in the register of operational modeling rather than experiential inference.
21. Pezzulo, G., Rigoli, F., & Friston, K. (2015). Active inference, homeostatic regulation and adaptive behavioural control. Progress in Neurobiology, 134, 17–35. https://doi.org/10.1016/j.pneurobio.2015.09.001
Annotation: This supports the coupling-and-regulation framing: competence emerges from embodied control loops that stabilize an organism under constraints. In the essay, it helps argue that “adaptive behavior” can be grounded in regulation dynamics, without automatically upgrading regulation into consciousness.
22. Montévil, M., & Mossio, M. (2015). Biological organisation as closure of constraints. Journal of Theoretical Biology, 372, 179–191. https://doi.org/10.1016/j.jtbi.2015.02.029
Annotation: This is used as a foundational reference for constraint-closure accounts of biological organization. It supports the essay’s claim that organization is an enacted closure of constraints (a mechanistic, relational property), and it helps critique explanations that treat “targets” or “forms” as if they exist in a separate ontological realm.
23. Mossio, M., Saborido, C., & Moreno, A. (2011). Biological organization and cross-generation functions. The British Journal for the Philosophy of Science, 62(3), 583–606. https://doi.org/10.1093/bjps/axq034
Annotation: This provides a philosophically rigorous account of function and organization that helps distinguish teleonomy (constraint-grounded function) from teleology (purpose as a primitive). In the essay, it supports arguments that “goal-like” behavior can be explained via organization and history without invoking intentional design.
24. Di Paolo, E. A. (2009). Extended life. Topoi, 28, 9–21. https://doi.org/10.1007/s11245-008-9042-3
Annotation: This supports the essay’s anti-vitalist autonomy framing: agency is enacted through organism–environment coupling, not possessed as an intrinsic essence. It helps motivate the essay’s claim that isolating systems (conceptually or experimentally) can destroy competencies that only exist in coupled regimes.
25. Beer, R. D. (2014). Dynamical systems and embedded cognition. Behavioral Sciences, 4(3), 185–208. https://doi.org/10.3390/bs4030185
Annotation: This supports the essay’s insistence that many “cognitive” phenomena are dynamical relaxations in coupled systems, not inner symbolic search in a private problem space. It is used to keep explanations mechanistic and to resist reifying computational metaphors into ontological commitments.
26. Bechtel, W. (2011). Mechanism and biological explanation. Philosophy of Science, 78(4), 533–557. https://doi.org/10.1086/661521
Annotation: This serves as a general methodological anchor: explanation requires organized causal mechanisms, not merely relabeling phenomena with new vocabulary. In the essay, it supports critiques of “vocabulary inflation” where cognitive terms are introduced without new, testable causal structure.
27. Godfrey-Smith, P. (2001). Environmental complexity and the evolution of cognition. Trends in Cognitive Sciences, 5(2), 64–70. https://doi.org/10.1016/S1364-6613(00)01612-5
Annotation: This supports an evolutionary account of why efficient, flexible behavior emerges under environmental pressures without implying “mind everywhere.” In the essay, it helps separate selective explanations for competence from claims that competence is identical to cognition in a thick, phenomenological sense.
28. Tero, A., Takagi, S., Saigusa, T., Ito, K., Bebber, D. P., Fricker, M. D., Yumiki, K., Kobayashi, R., & Nakagaki, T. (2010). Rules for biologically inspired adaptive network design. Science, 327(5964), 439–442. https://doi.org/10.1126/science.1177894
Annotation: This is used as a clean empirical example of optimization-like outcomes (efficient networks) without neurons. It supports the essay’s argument that “competence from coupling” can be real and measurable, while cautioning against anthropomorphic upgrades to consciousness.
29. Nakagaki, T., Yamada, H., & Tóth, Á. (2000). Intelligence: Maze-solving by an amoeboid organism. Nature, 407, 470. https://doi.org/10.1038/35035159
Annotation: This supports the essay’s steelman that non-neural systems can display impressive problem-solving dynamics, while also serving as a cautionary case for anthropomorphic temptation. It helps frame “maze-solving” as distributed control and morphological computation rather than evidence of human-like cognition.
30. Reid, C. R., Latty, T., Dussutour, A., & Beekman, M. (2016). Slime mould uses an externalized spatial “memory” to navigate in complex environments. Proceedings of the National Academy of Sciences, 109(43), 17490–17494. https://doi.org/10.1073/pnas.1215037109
Annotation: This is used to show that “memory” can be literal environmental trace rather than an inner representational store. In the essay, it supports arguments against reifying internal “perspectives” when the relevant state variables are distributed across agent–environment coupling.
31. Boisseau, R. P., Vogel, D., & Dussutour, A. (2016). Habituation in non-neural organisms: Evidence from slime moulds. Proceedings of the Royal Society B, 283, 20160446. https://doi.org/10.1098/rspb.2016.0446
Annotation: This supports the claim that learning-like dynamics (e.g., habituation) can occur without neurons, strengthening the essay’s competence-first framing. It also helps keep the argument mechanistic by tying “habituation” to observable dynamical changes rather than to assumed inner experience.
32. Reid, C. R., MacDonald, H., Mann, R. P., Marshall, J. A. R., Latty, T., & Garnier, S. (2016). Decision-making without a brain: How slime moulds solve complex problems. Journal of the Royal Society Interface, 13, 20160030. https://doi.org/10.1098/rsif.2016.0030
Annotation: This supports a core argumentative move: “decision” and “problem-solving” language can accurately describe dynamical selection among trajectories, without entailing consciousness. It helps the essay argue for conceptual hygiene: acknowledge competence while refusing unwarranted phenomenological upgrades.
33. Berg, H. C., & Brown, D. A. (1972). Chemotaxis in Escherichia coli analysed by three-dimensional tracking. Proceedings of the National Academy of Sciences, 69(3), 379–383. https://doi.org/10.1073/pnas.69.3.379
Annotation: This is used as a gold-standard case where efficient search/gradient-following is mechanistically explained. It supports the essay’s claim that impressive adaptive behavior can be fully accounted for by control architecture and feedback loops, without invoking inner maps or mental spaces.
34. Parent, C. A., & Devreotes, P. N. (1999). A cell’s sense of direction. Science, 284(5415), 765–770. https://doi.org/10.1126/science.284.5415.765
Annotation: This supports a mechanistic account of chemotaxis as signal processing and feedback control. In the essay, it reinforces the distinction between “direction-sensitivity” as an evolved control capability and “cognition” as a thicker claim that requires additional operational markers.
35. Alon, U. (2007). Network motifs: Theory and experimental approaches. Nature Reviews Genetics, 8, 450–461. https://doi.org/10.1038/nrg2102
Annotation: This supports framing “basal cognition” in terms of generic control motifs (feedback, feedforward, integral control) rather than mind-substance. It helps the essay argue that many allegedly “cognitive” capacities can be decomposed into reusable dynamical motifs across biological scales.
36. Tyson, J. J., Chen, K. C., & Novak, B. (2001). Network dynamics and cell physiology. Nature Reviews Molecular Cell Biology, 2, 908–916. https://doi.org/10.1038/35103078
Annotation: This supports the claim that cellular “decisions” can be modeled as attractor dynamics and thresholded transitions. In the essay, it helps translate anthropomorphic language into dynamical-systems terms that offer operational handles and falsifiable models.
37. Huang, S. (2009). Non-genetic heterogeneity of cells in development: More than just noise. Development, 136(23), 3853–3862. https://doi.org/10.1242/dev.035139
Annotation: This supports the idea that developmental outcomes can arise from structured dynamical landscapes (including heterogeneity) without requiring Platonic “problem spaces.” It helps the essay argue that complexity and variability can be generative features of constraint-governed dynamics rather than evidence of non-physical templates.
38. Levin, M. (2014). Molecular bioelectricity: How endogenous voltage potentials control cell behavior and instruct pattern regulation in vivo. Molecular Biology of the Cell, 25(24), 3835–3850. https://doi.org/10.1091/mbc.E13-12-0708
Annotation: This is used as a core mechanistic citation establishing that bioelectricity is causal and measurable in pattern regulation. It supports the essay’s “good science exists here” concession while enabling a clean separation between empirical bioelectric control and any metaphysical overinterpretation.
39. Levin, M. (2021). Bioelectric signaling: Reprogrammable circuits underlying embryogenesis, regeneration, and cancer. Cell, 184(8), 1971–1989. https://doi.org/10.1016/j.cell.2021.02.034
Annotation: This provides a high-profile synthesis of bioelectric signaling as reprogrammable circuitry. In the essay, it functions as an authoritative summary of the mechanistic program, allowing critiques to focus on interpretive leaps rather than disputing the empirical relevance of bioelectric dynamics.
40. Beane, W. S., Morokuma, J., Adams, D. S., & Levin, M. (2011). A chemical genetics approach reveals H,K-ATPase-mediated membrane voltage is required for planarian head regeneration. Chemistry & Biology, 18(1), 77–89. https://doi.org/10.1016/j.chembiol.2010.11.012
Annotation: This supports the essay’s claim that voltage gradients are not metaphorical—they can be mechanistically necessary for morphological outcomes. It anchors “attractors are physical” by tying patterning to specific, intervenable physiological mechanisms.
41. Oviedo, N. J., Morokuma, J., Walentek, P., Kema, I. P., Gu, M. B., Ahn, J.-M., Hwang, J. S., Gojobori, T., & Levin, M. (2010). Long-range neural and gap junction protein-mediated cues control polarity during planarian regeneration. Development, 137(20), 3293–3303. https://doi.org/10.1242/dev.051730
Annotation: This supports the essay’s emphasis on distributed coupling (e.g., gap junctions) as a causal substrate for long-range coordination. It fits the essay’s “isolation-as-intervention” heuristic: disrupting coupling often destroys the competence that appears “agent-like.”
42. Nogi, T., & Levin, M. (2005). Characterization of innexin gene expression and functional roles of gap-junctional communication in planarian regeneration. Developmental Biology, 287(2), 314–335. https://doi.org/10.1016/j.ydbio.2005.09.002
Annotation: This provides supporting evidence for the causal role of gap-junctional communication in regeneration. In the essay, it is used to keep explanations grounded in identifiable coupling mechanisms rather than in appeals to abstract “morphological memories” detached from physiology.
43. Durant, F., Morokuma, J., Fields, C., Williams, K., Adams, D. S., & Levin, M. (2017). Long-term, stochastic editing of regenerative anatomy via targeting endogenous bioelectric gradients. Biophysical Journal, 112(10), 2231–2243. https://doi.org/10.1016/j.bpj.2017.04.011
Annotation: This is used as one of the strongest empirical anchors for “anatomical memory” understood as a stable bioelectric state. It supports the essay’s preferred move: treat “memory” as a persistent, manipulable physiological state variable, not as a metaphysical storehouse.
44. Lobo, D., Beane, W. S., & Levin, M. (2012). Modeling planarian regeneration: A primer for reverse-engineering the worm. PLoS Computational Biology, 8(4), e1002481. https://doi.org/10.1371/journal.pcbi.1002481
Annotation: This supports the essay’s insistence on falsifiable modeling and explicit null models. It provides a methodological bridge: rigorous modeling can proceed without metaphysical commitments, enabling critiques to focus on when cognitive language adds testable structure versus when it merely relabels.
45. Pezzulo, G., & Levin, M. (2016). Re-membering the body: Applications of computational neuroscience to the top-down control of regeneration of body structure. PLoS Biology, 14(7), e1002481. https://doi.org/10.1371/journal.pbio.1002481
Annotation: This is used as a “bridge text” where cognitive and computational-neuroscience vocabulary is imported into morphogenesis. In the essay, it provides a concrete locus for auditing whether the imported vocabulary yields new predictions and interventions, or whether it functions mainly as persuasive framing.
46. Cervera, J., Meseguer, S., & Mafe, S. (2016). Bioelectrical signals and ion channels in the control of development and regeneration. Comptes Rendus Biologies, 339(5–6), 290–296. https://doi.org/10.1016/j.crvi.2016.03.003
Annotation: This provides an independent overview supporting bioelectric control as mainstream, mechanistic biology. In the essay, it reduces dependence on a single lab’s framing and helps establish that the underlying physiological story does not require metaphysical add-ons.
47. McLaughlin, K. A., & Levin, M. (2018). Bioelectric signaling in regeneration: Mechanisms of ionic controls of growth and form. Developmental Biology, 433(2), 177–187. https://doi.org/10.1016/j.ydbio.2017.08.032
Annotation: This supports the essay’s “mechanisms first” approach by emphasizing ionic control mechanisms in regeneration. It allows the essay to grant the empirical success of bioelectric regulation while arguing that interpretive inflation is optional, not entailed.
48. Sundelacruz, S., Levin, M., & Kaplan, D. L. (2009). Role of membrane potential in the regulation of cell proliferation and differentiation. Stem Cell Reviews and Reports, 5, 231–246. https://doi.org/10.1007/s12015-009-9080-2
Annotation: This strengthens the physiological grounding that membrane potential states matter for proliferation and differentiation. In the essay, it supports the claim that the relevant state variables are measurable and manipulable, which is central for a falsification-first approach.
49. Blackiston, D. J., Shomrat, T., & Levin, M. (2015). The stability of memories during brain remodeling: A perspective. Communicative & Integrative Biology, 8(2), e1073424. https://doi.org/10.1080/19420889.2015.1073424
Annotation: This is used to frame claims about memory persistence through remodeling in a defensible way while remaining cautious about interpretation. In the essay, it supports the distinction between “memory as functional persistence under remodeling” and stronger claims that would require additional operational markers.
50. Shomrat, T., & Levin, M. (2013). An automated training paradigm reveals long-term memory in planarians and its persistence through head regeneration. Journal of Experimental Biology, 216, 3799–3810. https://doi.org/10.1242/jeb.087809
Annotation: This serves as a key empirical anchor for “memory persists through regeneration” claims. In the essay, it is used to argue that even strong-looking behavioral persistence does not automatically justify reifying a persistent subject; it can motivate mechanistic hypotheses about distributed storage, coupling, and control.
51. Cangiano, L., Gargini, C., & Della Santina, L. (2020). Retinal circuits and predictive coding. Annual Review of Vision Science, 6, 1–24. https://doi.org/10.1146/annurev-vision-121219-081643
Annotation: This supports the essay’s use of “prediction” as a circuit property, grounding predictive coding in mechanistic circuitry. It helps prevent equivocation where “prediction” is treated as inherently experiential rather than as an inferential control strategy implementable by circuits.
52. Seth, A. K. (2015). The cybernetic Bayesian brain: From interoceptive inference to sensorimotor contingencies. Frontiers in Psychology, 6, 1–24. https://doi.org/10.3389/fpsyg.2015.00786
Annotation: This supports the essay’s separation of inference/control accounts from identity claims about experience. It helps argue that showing Bayesian-like control does not settle phenomenology, reinforcing the essay’s demand for explicit operational markers when consciousness claims are made.
53. Tononi, G. (2004). An information integration theory of consciousness. BMC Neuroscience, 5, 42. https://doi.org/10.1186/1471-2202-5-42
Annotation: This is used to provide a contrast class: a consciousness theory that makes explicit commitments (a proposed measure) and thus invites direct falsification pressure. In the essay, it functions as a foil to more elastic “intelligence” terminology by illustrating what it looks like to stick one’s neck out operationally.
54. Oizumi, M., Albantakis, L., & Tononi, G. (2014). From the phenomenology to the mechanisms of consciousness: Integrated information theory 3.0. PLoS Computational Biology, 10(5), e1003588. https://doi.org/10.1371/journal.pcbi.1003588
Annotation: This further supports the contrast class role: a detailed formalism tying phenomenology claims to proposed mechanisms and measures. In the essay, it is used to emphasize that serious phenomenology talk must specify testable commitments and clear failure modes.
55. Mashour, G. A., Roelfsema, P., Changeux, J.-P., & Dehaene, S. (2020). Conscious processing and the global neuronal workspace hypothesis. Neuron, 105(5), 776–798. https://doi.org/10.1016/j.neuron.2020.01.026
Annotation: This provides an empirically engaged framework for consciousness with operational markers and experimental paradigms. In the essay, it supports the claim that “taking consciousness seriously” involves specifying discriminative criteria, rather than inferring consciousness from broad competence alone.
56. Alkire, M. T., Hudetz, A. G., & Tononi, G. (2008). Consciousness and anesthesia. Science, 322(5903), 876–880. https://doi.org/10.1126/science.1149213
Annotation: This supports the essay’s strategy of using anesthesia as an operational wedge: when experience is suppressed, what measurable physiological changes correlate? It helps keep the argument in physiology and mechanisms rather than in speculative metaphysics.
57. Franks, N. P. (2008). General anesthetics: From molecular targets to neuronal pathways of sleep and arousal. Nature Reviews Neuroscience, 9, 370–386. https://doi.org/10.1038/nrn2372
Annotation: This provides mechanistic scaffolding for anesthesia effects, supporting the essay’s claim that many anesthesia phenomena are explainable via molecular targets and network dynamics. It is used to resist “single-factor” or exotic interpretations that jump directly to fundamental metaphysical conclusions.
58. Hameroff, S., & Penrose, R. (2014). Consciousness in the universe: A review of the “Orch OR” theory. Physics of Life Reviews, 11(1), 39–78. https://doi.org/10.1016/j.plrev.2013.08.002
Annotation: This is used as a parallel case where metaphysical reach can outrun testability, providing a comparison point for auditing how “quantum” is leveraged in consciousness arguments. In the essay, it helps structure a critique of arguments that expand ontological commitments faster than they expand falsifiable predictions.
59. Kelz, M. B., & Mashour, G. A. (2019). The biology of general anesthesia from paramecium to primate. Current Biology, 29(22), R1199–R1210. https://doi.org/10.1016/j.cub.2019.10.040
Annotation: This supports cross-scale caution: anesthesia affects diverse organisms, but that breadth does not license the conclusion that “consciousness is everywhere.” In the essay, it is used to argue that broad pharmacological sensitivity can reflect conserved biophysics rather than ubiquitous phenomenology.
60. Elsberry, W. R., & Shallit, J. (2011). Information theory, evolutionary computation, and Dembski’s “complex specified information”. Synthese, 178, 237–270. https://doi.org/10.1007/s11229-009-9542-8
Annotation: This supports the essay’s critique of “specified complexity” as a rhetorical-mathematical construct that can be rigged by baseline choice and probability games. It is used to draw a structural parallel: if “information” becomes a persuasive token rather than a physically accountable quantity, it can function as camouflage rather than explanation.
61. Perakh, M. (2004). There is a free lunch after all: William Dembski’s “specified complexity” as an illusion. Reports of the National Center for Science Education, 24(1), 1–7. https://doi.org/10.2307/4438068
Annotation: This provides a direct critical reference for claims that “specified complexity” does not do the inferential work its proponents claim. In the essay, it supports the broader warning that mathematical language can be used to launder weak inference when null models and baselines are not explicitly controlled.
62. Shallit, J., & Elsberry, W. R. (2007). Playing games with probability: Dembski’s complex specified information. Reports of the National Center for Science Education, 27(4), 1–9. https://doi.org/10.2307/4438461
Annotation: This further supports the point that probability baselines and null models can be manipulated to produce “impressive” numbers. In the essay, it is used to reinforce a general inference-hygiene lesson applicable to cognition claims: metrics are only as honest as their baselines and model classes.
63. MacKay, D. J. C. (2003). Information theory, inference and learning algorithms. Cambridge University Press. https://doi.org/10.1017/CBO9780511817085
Annotation: This is used to ground information/inference talk in a rigorous technical reference, supporting the essay’s insistence on clean definitions and explicit priors/likelihoods. It helps prevent equivocations between Shannon information, semantic information, and phenomenological content.
64. Burnham, K. P., & Anderson, D. R. (2004). Multimodel inference: Understanding AIC and BIC in model selection. Sociological Methods & Research, 33(2), 261–304. https://doi.org/10.1177/0049124104268644
Annotation: This supports the essay’s emphasis that model selection metrics depend on the candidate model class and baseline assumptions. It is used to reinforce the parallel: any “intelligence metric” or “complexity metric” is only meaningful relative to an explicitly stated model space and comparison set.
65. Ioannidis, J. P. A. (2005). Why most published research findings are false. PLoS Medicine, 2(8), e124. https://doi.org/10.1371/journal.pmed.0020124
Annotation: This supports the essay’s argument about inference hygiene and institutional constraints: publication and peer review provide some filtering but do not guarantee truth. It is used to motivate explicit falsifiers, replication sensitivity, and skepticism toward claims that travel primarily through popular media without robust evidential scaffolding.
66. Wasserstein, R. L., & Lazar, N. A. (2016). The ASA’s statement on p-values: Context, process, and purpose. The American Statistician, 70(2), 129–133. https://doi.org/10.1080/00031305.2016.1154108
Annotation: This supports the essay’s insistence on avoiding “significance laundering” and on stating explicit decision rules and evidential standards. It is used to argue that rhetorical certainty is not a substitute for well-specified tests and error controls.
67. Gelman, A., & Shalizi, C. R. (2013). Philosophy and the practice of Bayesian statistics. British Journal of Mathematical and Statistical Psychology, 66(1), 8–38. https://doi.org/10.1111/j.2044-8317.2011.02037.x
Annotation: This supports the essay’s Popper/Lakatos-compatible stance within modern statistical practice: models are tools to be checked, criticized, and revised. It is used to justify an audit posture that treats claims as revisable hypotheses under constraint, rather than as final declarations.
68. Dennett, D. C. (1988). Precis of The intentional stance. Behavioral and Brain Sciences, 11(3), 495–505. https://doi.org/10.1017/S0140525X00058611
Annotation: This supports the essay’s separation between the intentional stance (a predictive heuristic) and intentionality-as-ontology. It is used to argue that describing systems “as if” they have beliefs/goals can be pragmatically useful without committing to inner mental entities or universal cognition.
69. Beer, R. D. (1995). A dynamical systems perspective on agent–environment interaction. Artificial Intelligence, 72(1–2), 173–215. https://doi.org/10.1016/0004-3702(94)00005-L
Annotation: This supports the essay’s distinction between search-like narratives and dynamical relaxation in coupled systems. It provides a formal basis for analyzing agent–environment coupling without assuming internal symbolic problem spaces.
70. Ashby, W. R. (1958). Requisite variety and its implications for the control of complex systems. Cybernetica, 1, 83–99. https://doi.org/10.1007/978-3-642-82067-5_9
Annotation: This supports the essay’s control-theoretic definition of competence: regulation requires sufficient variety relative to disturbances. It is used to argue that “intelligence” can be framed as variety-management under constraints without presupposing consciousness.
71. Conant, R. C., & Ashby, W. R. (1970). Every good regulator of a system must be a model of that system. International Journal of Systems Science, 1(2), 89–97. https://doi.org/10.1080/00207727008920220
Annotation: This supports an operational definition of “model” as whatever internal/external structure enables effective regulation. In the essay, it is used to avoid reifying “problem spaces” into Platonic realms by keeping “model” tied to control performance and measurable coupling.
72. Friston, K., & Kiebel, S. (2009). Predictive coding under the free-energy principle. Philosophical Transactions of the Royal Society B, 364(1521), 1211–1221. https://doi.org/10.1098/rstb.2008.0300
Annotation: This supports the essay’s mechanistic treatment of prediction as an inference architecture rather than as evidence of subjective perspective. It provides a bridge between “prediction talk” and concrete circuit-level implementations.
73. Shannon, C. E. (1948). A mathematical theory of communication. Bell System Technical Journal, 27(3), 379–423. https://doi.org/10.1002/j.1538-7305.1948.tb01338.x
Annotation: This is used to establish definitional discipline: Shannon information concerns uncertainty reduction in signals, not meaning or experience. In the essay, it supports policing equivocations where “information” slides between technical and folk/phenomenological senses.
74. Cover, T. M., & Thomas, J. A. (1991). Elements of information theory. Wiley. https://doi.org/10.1002/0471200611
Annotation: This supports the essay’s insistence on standard definitions and careful use of information measures. It is used to keep the argument technically grounded when deploying entropy, mutual information, and related quantities.
75. Maynard Smith, J. (2000). The concept of information in biology. Philosophy of Science, 67(2), 177–194. https://doi.org/10.1086/392768
Annotation: This supports the essay’s critique of sloppy “information” language in biology by clarifying what “information” can and cannot mean in biological contexts. It is used to prevent illegitimate jumps from biological signaling to metaphysical claims about informational realms.
76. Adami, C. (2012). The use of information theory in evolutionary biology. Annals of the New York Academy of Sciences, 1256(1), 49–65. https://doi.org/10.1111/j.1749-6632.2011.06422.x
Annotation: This supports the essay’s claim that evolutionary processes can deposit structure/information into organisms over time, offering a rigorous way to talk about historical compression without invoking external templates. It helps frame “competence” as accumulated constraint satisfaction.
77. Lenski, R. E., Rose, M. R., Simpson, S. C., & Tadler, S. C. (1991). Long-term experimental evolution in Escherichia coli. I. Adaptation and divergence during 2,000 generations. The American Naturalist, 138(6), 1315–1341. https://doi.org/10.1086/285289
Annotation: This supports the essay’s empirical grounding for “selection deposits work into structure,” using controlled long-term evolution as evidence. It helps justify claims about competence emerging via historical constraint filtering.
78. Blount, Z. D., Borland, C. Z., & Lenski, R. E. (2008). Historical contingency and the evolution of a key innovation in an experimental population of E. coli. Proceedings of the National Academy of Sciences, 105(23), 7899–7906. https://doi.org/10.1073/pnas.0803151105
Annotation: This supports the essay’s emphasis on path dependence and contingency, reinforcing the point that high competence can reflect historical trajectories rather than access to timeless targets. It helps argue against explanations that implicitly treat outcomes as pre-specified forms.
79. Kauffman, S. A. (1993). The origins of order: Self-organization and selection in evolution. Oxford University Press. https://doi.org/10.1093/oso/9780195058112.001.0001
Annotation: This supports the essay’s balanced framing of self-organization and selection as complementary mechanisms that generate order without invoking design or external teleology. It helps discipline “emergence” claims into known dynamical and evolutionary processes.
80. Haken, H. (1983). Synergetics: An introduction. Springer. https://doi.org/10.1007/978-3-642-88375-5
Annotation: This supports the essay’s use of attractors and order parameters as formal tools for describing “goal-like” outcomes without intentions. It provides language for how macro-level constraints shape micro-dynamics (downward causation as constraint, not new forces).
81. Nicolis, G., & Prigogine, I. (1977). Self-organization in nonequilibrium systems. Wiley. https://doi.org/10.1002/9780470150569
Annotation: This supports the essay’s thermodynamic framing of pattern formation and dissipative structures. It anchors claims that order can emerge in driven systems through known nonequilibrium processes, avoiding appeals to Platonic templates.
82. Cross, M. C., & Hohenberg, P. C. (1993). Pattern formation outside of equilibrium. Reviews of Modern Physics, 65(3), 851–1112. https://doi.org/10.1103/RevModPhys.65.851
Annotation: This authoritative review supports the essay’s claim that pattern formation can be explained within established physics, including rigorous treatment of instabilities and emergent structures. It functions as a “don’t hand-wave” anchor: patterns need not be explained by external targets when the dynamics already generate them.
83. Turing, A. M. (1952). The chemical basis of morphogenesis. Philosophical Transactions of the Royal Society B, 237(641), 37–72. https://doi.org/10.1098/rstb.1952.0012
Annotation: This supports the essay’s historical point that morphogenetic “targets” can be generated from reaction–diffusion dynamics. It is used to demonstrate that formal generative mechanisms for biological form have long existed without requiring metaphysical morphospaces.
84. Meinhardt, H. (1982). Models of biological pattern formation. Academic Press. https://doi.org/10.1016/C2013-0-10917-0
Annotation: This supports the essay’s claim that mechanistic modeling traditions in morphogenesis provide testable alternatives to treating “form” as an external template. It helps position the essay’s critique as a call for model comparison, not as a rejection of form-generation science.
85. Wolpert, L. (1969). Positional information and the spatial pattern of cellular differentiation. Journal of Theoretical Biology, 25(1), 1–47. https://doi.org/10.1016/S0022-5193(69)80016-0
Annotation: This supports the essay’s argument that “goal-like” morphology can arise from local signals and positional constraints. It helps show how “targets” can be encoded in gradients and rules without implying cognition or an external realm of ideal forms.
86. Levin, M. (2005). Left–right asymmetry in embryonic development: A comprehensive review. Mechanisms of Development, 122(1), 3–25. https://doi.org/10.1016/j.mod.2004.08.006
Annotation: This supports the essay’s empirical grounding for bioelectric and physiological contributions to asymmetry and patterning. It is used to emphasize that the mechanistic program has depth and breadth, so critiques should focus on interpretive overreach rather than denying physiology.
87. Adams, D. S., & Levin, M. (2012). General principles for measuring resting membrane potential and ion concentration using fluorescent bioelectricity reporters. Cold Spring Harbor Protocols, 2012(4), pdb.top067710. https://doi.org/10.1101/pdb.top067710
Annotation: This supports the essay’s demand for operational measurement hooks. It is used to show that the relevant variables (membrane potential, ion concentrations) are measurable in practice, enabling falsifiable claims rather than purely rhetorical “bioelectric code” language.
88. Pai, V. P., Aw, S., Shomrat, T., Lemire, J. M., & Levin, M. (2012). Transmembrane voltage potential controls embryonic eye patterning in Xenopus laevis. Development, 139(2), 313–323. https://doi.org/10.1242/dev.073759
Annotation: This provides mechanistic evidence that voltage potentials can control specific patterning outcomes. In the essay, it supports the argument that pattern regulation can be treated as causal physiology, which reduces the need for metaphysical explanations.
89. Tseng, A.-S., & Levin, M. (2013). Cracking the bioelectric code: Probing endogenous ionic controls of pattern formation. Communicative & Integrative Biology, 6(1), e22595. https://doi.org/10.4161/cib.22595
Annotation: This supports the essay’s framing of “bioelectric code” as a testable research program. It is used to separate legitimate programmatic metaphors (useful for guiding experiments) from ontological claims that a “code” exists as a realm independent of the physical coupling that implements it.
90. Levin, M., Pezzulo, G., & Finkelstein, J. M. (2017). Endogenous bioelectric signaling networks: Exploiting voltage gradients for control of growth and form. Annual Review of Biomedical Engineering, 19, 353–387. https://doi.org/10.1146/annurev-bioeng-071114-040647
Annotation: This supports the essay’s “good science is here” premise while maintaining a boundary between engineering/control framings and metaphysical interpretation. It is used to show how voltage gradients can be treated as control variables for growth and form in a rigorous, application-oriented context.
91. Fields, C., & Levin, M. (2018). Multiscale memory and bioelectric error correction in living systems. BioSystems, 164, 35–45. https://doi.org/10.1016/j.biosystems.2017.12.004
Annotation: This supports the essay’s analysis of “memory/error correction” language as potentially mechanistic (distributed state stabilization) rather than inherently cognitive. It provides a concrete text where the essay can test whether “memory” is operationalized in ways that generate novel predictions or whether it functions as a persuasive label.
92. Pezzulo, G., & Levin, M. (2015). Re-membering the body: Top-down coordination and planning in morphogenesis. Communicative & Integrative Biology, 8(6), e1106833. https://doi.org/10.1080/19420889.2015.1106833
Annotation: This serves as another “bridge target” for auditing when cognitive vocabulary is imported into morphogenesis. In the essay, it is used to evaluate whether “planning/top-down control” is defined by explicit mechanisms and interventions or primarily by analogy.
93. Baluška, F., & Levin, M. (2016). On having no head: Cognition throughout biological systems. Frontiers in Psychology, 7, 902. https://doi.org/10.3389/fpsyg.2016.00902
Annotation: This is used as a representative “basal cognition” manifesto to steelman and then audit under falsifiability discipline. It supports the essay’s examination of definitional spread: whether “cognition” is being used as a precise, operational term or as an elastic label that risks collapsing distinctions.
94. Ginsburg, S., & Jablonka, E. (2019). The evolution of the sensitive soul: Learning and the origins of consciousness. MIT Press. https://doi.org/10.7551/mitpress/12255.001.0001
Annotation: This supports the essay by providing a serious attempt to connect learning capacities to minimal consciousness criteria. It functions as a contrast case where consciousness claims are tied to explicit markers and evolutionary arguments rather than inferred directly from generalized competence.
95. Birch, J., Schnell, A. K., & Clayton, N. S. (2020). Dimensions of animal consciousness. Trends in Cognitive Sciences, 24(10), 789–801. https://doi.org/10.1016/j.tics.2020.07.007
Annotation: This supports the essay’s ethics-related discipline: moral extension requires discriminative markers and graded evidence, not infinite scope creep. It is used to argue that “consciousness everywhere” is not a cautious ethical posture unless it comes with operational criteria and a tractable decision framework.
96. Seth, A. K., & Bayne, T. (2022). Theories of consciousness. Nature Reviews Neuroscience, 23, 439–452. https://doi.org/10.1038/s41583-022-00587-4
Annotation: This supports the essay by mapping the theory space of consciousness and highlighting the need for explicit tests and failure modes. It helps frame the essay’s critique as an alignment with best practice: strong claims about experience demand strong operational commitments.
97. Koch, C., Massimini, M., Boly, M., & Tononi, G. (2016). Neural correlates of consciousness: Progress and problems. Nature Reviews Neuroscience, 17, 307–321. https://doi.org/10.1038/nrn.2016.22
Annotation: This supports the essay’s caution that correlation does not license indiscriminate extension. It is used to emphasize methodological pitfalls (overgeneralization, marker ambiguity) and to justify a conservative inferential stance when moving from competence or complexity to consciousness.
98. Mitchell, M. (2009). Complexity: A guided tour. Oxford University Press. https://doi.org/10.1093/oso/9780195124411.001.0001
Annotation: This supports careful language around “complexity” and helps prevent complexity from being treated as a synonym for cognition. In the essay, it functions as a general reference for disciplined discussions of emergent phenomena without metaphysical escalation.
99. Crutchfield, J. P., & Young, K. (1989). Inferring statistical complexity. Physical Review Letters, 63(2), 105–108. https://doi.org/10.1103/PhysRevLett.63.105
Annotation: This supports the essay’s option to use structure-sensitive complexity measures that are explicitly about inference/predictability rather than about cognition. It helps argue that “structure” and “agency” are separable claims, and that one can quantify structure without smuggling in experience.
100. Grassberger, P. (1986). Toward a quantitative theory of self-generated complexity. International Journal of Theoretical Physics, 25, 907–938. https://doi.org/10.1007/BF00668821
Annotation: This supports the essay’s claim that complexity can be formalized as a dynamical property generated by systems, without attributing mental categories. It is used to reinforce the essay’s central hygiene move: formalize what can be measured, and refuse to convert that into ontology without additional evidence.
