Some Philosophical Implications of Michael Levin’s New Paradigm Biology

Over the last year, I’ve had a couple of extremely stimulating conversations with the developmental biologist Mike Levin.

The first was moderated by Karen Wong:

Platonizing Biology: A Dialogue with Michael Levin

Read full story

The second was a one-on-one exchange on my YouTube channel:

Taming the Technological Dragon, with Michael Levin

Read full story

I’m going to share some of the highlights of these conversations as part of a response to a recent critical engagement with Levin’s work by Johannes Jaeger.

Given the amount of hot air in the public discourse these days, whenever philosophers and scientists attempt to dialogue it’s essential to clarify up front the relationship between the two disciplines so as to ensure mutual understanding. As a process philosopher, I’m seeking a radically empirical metaphysical scheme that would allow an integral interpretation of everything we experience, including, of course, the latest scientific research. I’m not here to dictate scientific concepts to scientists but to learn from and generalize the findings of the special sciences, particularly biology and physics, to address broader questions about the universe.

One of the problems with the dominant scientific worldview that emerged in 17th century Europe—let’s call it mechanistic materialism—is that it made the human minds engaged in the scientific enterprise appear to be entirely alien to the universe they were studying. Everything else in nature seemed like a well-behaved mechanism, except of course the scientists doing all that reverse engineering.

One of the reasons I’ve been excited about Mike Levin’s work is that he is attempting to generalize mind, agency, and intelligence so they can be understood as essential features of evolutionary processes at every scale. He argues for an evolutionary continuum that allows scientists to view themselves as examples of the processes they study.

When we first had the chance to chat, I asked Mike what he felt the ideal relationship between science and philosophy should look like. He pointed immediately to the facts of developmental biology, which he believes force us to unify what otherwise might seem like two divergent worlds, that of the adult human moral agent, on the one hand, and that of dead, dumb physics and chemistry, on the other:

“Each one of us has taken that journey from chemistry and physics to being a complex metacognitive mind. That happens slowly and gradually, and you can watch it happen with your own eyes. In the case of a frog, you can watch the whole thing from beginning to end. In the case of a human, it takes a little more instrumentation, but you can still watch it. Each of us was once a quiescent oocyte, a little blob of chemistry, and now we are whatever it is that we are. To me, these things are fundamentally merged because whatever philosophical approaches we have to what we are and how we should think about the outside world, we have to realize that we are continuous with what we call chemistry and physics and supposedly mindless matter, which I’m not sure there is any such thing. That continuity binds the two fields in a way that cannot be untangled.”

Mike went on to say that he finds any sharp categorization separating machines from organisms to be an example of bad philosophy that is unhelpful for empirical scientific work. While I certainly want to let scientists take the lead when it comes to what concepts they find useful in their empirical work, I am not so sure about this. A lot depends on how we understand the conceptual shift at play here. Is Mike suggesting that what had appeared machine-like to the old materialist paradigm is actually mind-like and fundamentally organic or organism-derived? Or is he saying that what appears to be alive and ensouled is actually just a sophisticated computational algorithm that engineers can hack? To my mind, no organicist has ever denied that there are mechanisms in nature, even in the living world. The point is just that these mechanisms are subsets of more encompassing living processes. It is the mechanists who have denied organism, not the organicists who have denied mechanism.

Enter Johannes Jaeger, a systems biologist with a very interesting background, holding both a PhD in genetics from Stony Brook University as well as a MSc in Holistic Science from Schumacher College. I also have some history with Schumacher, having studied Gaia theory with Stephen Harding and later taught courses for them on Whitehead’s philosophy of organism and Goethe’s participatory method. It is great to see someone with Jaeger’s scientific cred applying a Schumacher approach to his work.

This morning, Jeffrey W Barney suggested I read Jaeger’s latest blog post, “Why TAME is lame.” The blog is a caustic critique of Levin’s work. I don’t want to assume, but given the (to my mind, entirely unnecessary) tone, I can’t help but feel like there is some kind of personal slight in the background motivating Jaeger’s criticisms. But putting the unfortunate personal attacks aside, Jaeger’s basic scientific and philosophical criticism seems to be that Levin is exaggerating the significance of bioelectric fields in both development and especially in evolution. Whether or not Levin is unduly turning all of biology into nails for his bioelectric hammer remains to be seen. Thus far at least, the empirical work of his lab at Tufts University suggests that the thesis is worth continuing to test!

One of the concepts Mike develops in light of his experimental work on bioelectric fields is that of developmental morphospace. Where do specific anatomical target morphologies, like salamander limbs, come from? The paradigmatic assumption has been that these shapes come from many millions of years gene-level natural selection. Levin disagrees:

“…is evolution really searching the difficult space of [genetic] microstates for biological systems? I don’t think so. I think evolution is searching for simple machines that pull down free gifts from the laws of physics, computation, and geometry. What evolution does is optimize the hardware, but there’s a ton of heavy lifting done elsewhere. I lack the vocabulary for it, but as a dumb example, suppose you’re evolving a triangle. There’s an environment, and only that kind of triangle is fit. You crank through generations and find the first angle, then the second angle. You’re done. You don’t need to look for the third angle because the three angles of a triangle have to add up to 180 degrees. Evolution saved one-third of its effort. Where did that come from?”

I’m reminded of both Stuart Kauffman’s idea of “order for free” out of the “adjacent possible,” and also the work of the late Brian Goodwin (who taught in the MSc program at Schumacher College).

I’m also reminded, of course, of Plato’s realm of forms. Plato wrote dialogues, not doctrines, but the theory of forms attributed to him posits a static realm of perfect ideas or archetypes, with physical bodies being imperfect imitations. The ancient Greeks had a more or less static view of nature, with no concept of evolutionary speciation. Plato’s student Aristotle, despite his genius for developmental morphology, envisioned a Great Chain of Being from minerals, through plants, animals, and humans, to the divine mind. He allowed for no means of evolutionary passage from one level to the other; but it may not be that difficult to integrate evolution by tipping the great chain on its side to show how novel forms can emerge over time.

In the 20th century, Alfred North Whitehead attempted something like an evolutionizing of Platonism. He aimed to reconcile modern science, which focused on efficient causes, with the formal and final causes known to the ancient world. The early modern scientists found it necessary to distance themselves from teleological explanations, which had grown rather baroque during the medieval period. The problem is that removing all purposiveness from nature left science entirely bereft of the concepts necessary for understanding living organisms, not to mention their own activity as scientists. As Whitehead famously quipped, “Scientists animated by the purpose of proving they are purposeless constitute an interesting subject for study” (The Function of Reason, Ch. 1).

Whitehead proposed viewing evolution not just biologically but cosmologically. There’s an accumulation of facts in the past that influence the present, but there’s also a field of possible forms opening into the future that is constantly being adjusted by the accumulation of past facts and present decisions. We can’t understand the relationship between the past facts and future forms purely in terms of efficient causation. What’s missing is the organismic agency operating in the present that can search this field of possibilities. We can understand this agency as an immanent operation intrinsic to the activity of organisms themselves, without resorting to supernatural causes or the dreaded intelligent design. What’s key here, for me, is that the old ontology of extended matter in motion is woefully inadequate for any truly evolutionary cosmology. In addition to taking the agency of living organisms seriously, we also need a realm of potential forms, or “res potentiae” as Kauffman, Ruth Kastner, and Michael Epperson describe it in a recent paper.

Whitehead’s philosophy suggests that some degree of agency is operating at various scales, from quantum events to conscious decisions. He envisioned a “cell theory of actuality,” where agency operates via cellular collectives (or “societies,” in his terms) across all scales. This makes his “organic realism” the perfect candidate to provide Mike’s self-acknowledged instrumentalism with its missing metaphysics. Here I disagree with Yaeger, who thinks Levin’s empirical work already presupposes a particular metaphysics.

In my second conversation with Mike, we spoke about his paper, “Technological Approaches to Mind Everywhere.” I was eager to discuss how his work was challenging some of my longstanding philosophical commitments. His redefinition of classical philosophical categories through the lens of technological advancements has been particularly eye-opening for me.

I kicked off the discussion by mentioning the important philosophical distinction between machines and organisms. Aristotle and Kant both recognized that an organism’s parts reciprocally produce one another for the sake of the whole, implying an inherent purposiveness is active in living beings. Mike argues that contemporary technologies have advanced to the point where machines can now exhibit behaviors traditionally associated with organisms, such as learning to solve problems they were never explicitly programed to solve. This blurring of lines prompted me to ask him about the ethical implications of the words and concepts we employ.

Mike emphasized the need to manage the consequences of our paradigms on people and their relationships. He argued that as scientists and philosophers, we should take the lead in defining categories, even if it means diverging from colloquial usage. According to him, categories like “machine” or “organism” should serve as operational protocols indicating the type of relationship we should have with a system, both ethically and practically. There was a bit of code shifting in Mike’s answer, which I admit does raise some concern for me. As much as he wanted to emphasize the ethics of relationality, he also wore his engineer’s hat by indicating that “we want to maximize the efficiency of our interaction.”

To illustrate his point, Mike presented a spectrum of systems ranging from simple machines to rational agents. He stressed the importance of using appropriate tools for interacting with each type of system. Misclassification, he warned, could lead to wasted effort or significant moral lapses. He underscored the empirical nature of determining a system’s place on this spectrum, using examples from regenerative medicine to show different levels of competency and learning in biological systems.

One striking example he shared was from orthopedic surgery, where bodies are treated as simple machines for certain purposes but rely on more complex, self-organizing processes for healing. He described associative learning in simple molecular circuits and how biological entities, like wasp parasites, can influence plant development. Mike predicted a future where engineered and natural systems will coexist in a complex continuum, necessitating new ethical and practical approaches.

Our discussion then turned to the challenge of reconciling evolution-driven cognition with rational engineering. I questioned the continuity between natural and rational processes, wondering if there was an implied dualism in Mike’s perspective. He argued that engineers could potentially replicate and surpass natural evolutionary processes.

I am more skeptical and cautious about this. While I do think human beings can become more adept cocreators capable of intelligently and empathically participating in the creative evolution of the Gaian community, I worry about the ways an engineering mindset tends to approach such projects. For example, as climate change worsens, governments will be under increasing pressure to implement large-scale technological quick fixes that will undoubtedly have unintended consequences. The point is less that humans are not as smart as evolution, but rather that the history of evolution is itself full of catastrophically disruptive mutations (e.g., my favorite example is the oxygen crisis 2 billion years ago). We don’t have another planet to retreat to if we screw this one up. Now, granted, as a result of industrialization our species has already been engaged in a rather unintelligent and unconscious geo-engineering scheme for the last few hundred years, the deleterious ecological effects of which have only relatively recently begun to dawn on us. So it’s not like the options are “yay” or “nay” to geoengineering. It is too late for that. But where we go from here will require a lot of humility regarding the limited powers of our rationality, which I don’t often detect in engineers (particularly in our age of AI hype).

Next, we delved into the philosophical distinction between organismal purpose and mechanical design. I noted how modern technological advancements (and PR campaigns) are complicating this distinction, as engineered systems are now often described as having goals and agency (i.e., “AI agents”). Mike supported the idea of teleological thinking in science, arguing against the “teleophobia” that denies purpose and goals in systems, whether computational or biological. He introduced his concept of a “cognitive light cone,” which resonates with me, especially when considering the ethical implications of expanding this light cone in terms of compassion and care for others.

As our conversation drew to a close, we discussed the ethical responsibilities of scientists and engineers. Mike emphasized that optimizing relationships with systems, rather than solely focusing on prediction and control, is crucial. He feels an ethical responsibility to improve current life conditions, given what he perceives to be the inherent flaws and limitations that an imperfect evolutionary process has left us with.

So while I share some of the concerns articulated in Jaeger’s blog post, I just don’t impute the nefarious ideology onto Levin that Jaeger seems to believe is at work. I have a different way of languaging and probably also of conceptualizing the work that Mike is doing. He prefers to couch it in computational terms and to remain somewhat quietist when it comes to the metaphysical background and implications (at least in his published articles). He is flirting with a panpsychist framework, but insists that he does so in a purely instrumentalist sense (i.e., if it is helpful for making experimental predictions to attribute mind or agency to a system, why not do so?).

Jaeger is skeptical of panpsychism, which is not surprising. There are forms of it that I find entirely unhelpful, as well (e.g., constitutive panpsychism). Whitehead never used the word, but I’ve tended to read his organic realism as a process-relational rendering of panpsychism if only because it bears more family resemblance to it than to the other ontologies on offer: physicalism, dualism, and idealism. Perhaps David Ray Griffin’s term “panexperientialism” is more helpful. Or perhaps the even more unpronounceable “panprehensionism” that Timothy Jackson and I have been discussing is more apt. Regardless, unlike Jaeger I do not think it is possible to account for life and mind in terms of some kind of strong emergentist scheme, as if there were ever something called “matter” that was utterly devoid of agency and feeling but nonetheless could give rise to the latter by way of more complicated arrangements of itself.

Rate this:

Posted

May 27, 2024

Alfred North Whitehead, Aristotle, Plato

Matthew David Segall

Tags:

biology, evolution, materialism, Michael Levin

Comments

One response to “Some Philosophical Implications of Michael Levin’s New Paradigm Biology”

Nathan Sweet

February 21, 2026

Hey Matt, let me try an intuition pump before anything else, because I think it clarifies what’s actually at stake here.

Imagine a brilliant engineer who tells you she’s designed a bridge-finding algorithm. You ask how it works, and she says: “Well, bridges are real, and the algorithm searches a Platonic bridge-space where all possible bridges pre-exist as eternal forms, and the most fit bridges sort of… insinuate themselves into the physical instantiation.” You’d want to know: does that description change any prediction about which bridges get built, how long they take to find, where they appear in the search trajectory? If the answer is no, then “Platonic bridge-space” is doing exactly zero engineering work. It’s a skyhook dressed in a lab coat. Levin’s triangle example has the same problem. The geometric constraint is real; triangles have three angles that sum to 180 degrees, full stop. But the inference from “real geometric constraint” to “evolution accesses a Platonic morphospace with something like causal efficacy” is a skyhook inference, and Kauffman’s “order for free” already handles the crane version without the metaphysical overhead.

Here’s what I’d call a deepity lurking in the framing. “Mechanisms are subsets of more encompassing living processes” sounds profound, and in one reading it’s perfectly true: organisms do more than their parts in isolation. But in the reading that actually does philosophical work in your argument, it’s a claim about causal hierarchy, a prediction about where the explanatory action lives, and that reading is either testable or it’s just the true-but-trivial reading wearing a borrowed coat. You can’t have both readings simultaneously without paying a philosophical toll. Pick one and defend it. That’s not an attack; it’s the only way the claim becomes interesting.

The res potentiae framing is where I want to apply what I call the “what is it good for” test. Whitehead, Kauffman, and Epperson want to give ontological weight to the adjacent possible, which is a genuinely interesting project. But Montévil and Mossio already do the heavy lifting here without invoking a possibility-field: constraints regenerate conditions for their own persistence through measurable metabolic work, full stop. When you ask “what’s the thermodynamic cost of agency searching a field of possibilities,” you’re not being reductive. You’re asking whether the field is a crane or a skyhook. If the field exerts causal or quasi-causal influence, that influence has to show up somewhere in the energy ledger. If it doesn’t, we haven’t explained anything; we’ve redescribed a mechanism in vocabulary that feels more hospitable to our intuitions about mind and purpose.

Now here’s the part that should give everyone in this conversation some pause, because it applies to me as much as anyone else. Several moves in your framing aren’t purely interpretive. They’re making implicit empirical claims, and implicit empirical claims incur burden of proof whether or not they announce themselves as empirical. When the old ontology gets called “woefully inadequate” for evolutionary cosmology, that’s a verdict about explanatory reach that needs a failure condition to mean anything. When Whitehead’s agency-at-quantum-scales shows up to demonstrate that conscious scientists are “continuous with” the processes they study, that’s a prediction about where agency begins in the physical world. Predictions need falsification surfaces. What would the developmental record need to show for you to say the organicist framing is wrong about how reality is actually structured, not just less poetically satisfying than mechanism? That’s not a hostile question. It’s the same question you’d rightly demand of any mechanistic account claiming more than it earned.

The Discovery Institute angle is worth dwelling on, because it’s not about imputing bad faith to anyone. It’s about what I’d call the affordance problem. Once “non-physical influence” language enters the framing, a very predictable thing happens downstream of your intentions. Groups with preloaded conclusions walk through the door you’ve opened. The logical skeleton of “morphogenesis is so impressively organized that ordinary mechanism can’t fully account for it, therefore something beyond ordinary mechanism must be at work” is structurally identical to the eye argument, and the identity of the structure matters even when the people running it have completely different motivations. A framework that can’t specify what would count against it can’t specify who gets to borrow it, either. Durant et al.’s path-dependent planarian outcomes after RNAi knockdown are exactly the sort of data a Platonic convergence account should have anticipated and addressed: different starting conditions producing endpoints that don’t converge on a canonical form is precisely what you’d expect if thermodynamic constraint rather than formal pull is doing the work. The fact that this falsification surface wasn’t stated in advance is diagnostic of the problem, not proof that the view is wrong.

On panexperientialism: you’re careful to distinguish Whitehead’s version from constitutive panpsychism, and the distinction is real. But the combination problem that William James posed clearly enough a century ago doesn’t get dissolved by noting that “societies” and “dominant occasions” are richer than crude aggregation. If consciousness requires the kind of integration needed to produce the unified macro-experience we’re trying to explain, the account is starting to say “organization produces the relevant kind of experience,” which is the emergentist claim wearing process vocabulary. And the tuner model has a specific empirical embarrassment: a tuner loosened should reveal hidden signal, not produce systematic unconsciousness. The Massimini perturbational complexity data shows that targeted disruptions to effective connectivity reliably reduce level and content of experience across anesthesia, sleep, and disorders of consciousness. Filters don’t work that way.

You’re right to worry about engineering hubris under climate pressure. But notice the asymmetry in how the skepticism gets deployed: engineers get the humility lecture, and philosophers of organism get the benefit of the doubt that their underanchored frameworks are pointing toward something important. The oxygen crisis you invoke as a cautionary tale nearly wiped out anaerobic life. The universe does not organize itself around the preservation of what’s philosophically interesting to us, and Whitehead’s cosmology needs to survive that observation rather than quietly requiring the universe to be more hospitable to agency and experience than the fossil record suggests.

The Lakatosian question is simple and I mean it sincerely: is the process-relational research program generating novel confirmed predictions, or has it been accumulating auxiliary hypotheses to protect the hard core for about a century now? The gestural warrant toward a unification of physics and biology is real and I don’t want to dismiss it. But gestures have a renewal requirement. What would convince you the synthesis had failed? That’s the same discipline any honest mechanistic account owes us, and the answer matters more than anything else in this conversation.

A related article I published in January goes into considerably more depth on several of the threads you’ve pulled on here, particularly the falsifiability demands on process-relational frameworks, the specific failure modes of the brain-as-tuner model under perturbational data, the Discovery Institute affordance problem, and a fuller treatment of where I think Whitehead’s program is genuinely valuable versus where it quietly purchases exemption from constraint under the guise of interpretive humility. You can find it at:
When “No Scientific Evidence” Becomes an Unintentional Shield: The Philosophy of Matthew Segall, Whitehead, and the Testability of Consciousness and Falsifiability

I want to be transparent that I hold these views tentatively, if anything in that article misreads your position or commits the very immunizing moves it diagnoses, I genuinely want to know, because the goal here is not to win an argument against process philosophy but to evolve toward a framework that neither of us has to defend with special pleading. Your scholarship is exactly the kind of serious interlocutor this work needs, and I’d be grateful for whatever you think I got wrong.

Reply