Life works through emergent collective intelligence: A conversation with Philip Ball

In How Life Works: A Users Guide to the New Biology (2023 University Chicago Press: 552 Pages) Philip Ball explains how recent research in cellular and molecular biology has begun to reveal how life emerges through dynamic, self-organizing processes. Ball explains how life emerges not by planning for every contingency but by enabling robust adaptive processes. Many of the ideas intersect with core themes in the field of collective intelligence. What follows is a lightly edited set of responses from Philip Ball to a set of question posed by Collective Intelligence co-editor Scott E Page.

Scott E Page (SP): In the book, you summarize much of science’s current understanding of how biological life self-organizes. I want to jump right in to some of the biggest, most provocative ideas. The book opens with a shot across the bow. You describe the Human Genome Project, while important, as an oversell. You describe how scientists have identified few straightforward causal relationships between genes and traits, behaviors, or diseases. DNA is not a blueprint. Creating life is much more than running ACTG double helixes through some biological Turing machine. I’m convinced. How then should we think of DNA?

Philip Ball (PB): DNA is a resource for making molecules that are useful to the cell and the organism—a resource that acts as an evolutionary memory of what has been found to work previously, but which can also generate phenotypic novelty. In this regard, I believe there’s a fuzziness to the question of whether any given molecule is functional or not. Some clearly are, some are pretty clearly not, but there are also genomically encoded molecules that seem to have the potential to find collective functions in certain circumstances—a pool of stuff that probably contributes to evolvability.

DNA—or more properly chromatin, where the DNA is bound to proteins that help structure it in chromosomes—is also a reconfigurable material that cells manipulate to control gene activity, for example, by partitioning in space groups of genes that are co-transcribed. So it really is wonderful stuff. But I don’t see how we can hope to find a “program” in there that specifies the organism. Rather, it encodes the ingredients with which cells can exert their agency and, in creatures like us, produce a wide range of phenotypic outcomes that themselves can change over time. It’s really hard to find a good analogy for talking about the genome, because we don’t know anything quite like it. But the “blueprint” is not a good metaphor any longer.

SP: The book covers much more than DNA. You widen the aperture chapter by chapter. You start at the smallest level and work up from genes, to RNA, to proteins, to networks, cells, and tissues and then bodies. At each level, the concept of signaling appears. In places, you state that we should interpret biological signaling as more of a conversation than a transmission. This seems important. What makes signal passing a conversation and why are conversations necessary?

Ball: Signaling tends to have a rather specific meaning in molecular and cell biology: the transmission of information to and through a cell via interactions between molecules, which “informs” the cell about itself and its environment and typically leads to some change in behavior (such as activation of a gene). For example, the arrival and binding of a hormone at a cell surface could trigger a cascade of interactions that ends with transcription of a gene. I guess here you’re using it in a somewhat broader sense to mean any kind of information flow in an organism (or even an ecosystem)—that might include, for example, the way electrical activity spreads in the brain. At any rate, one of the key messages of my book is that information flow in living systems is omnidirectional and multi-scale.

This is in contrast to the common perception that information starts with the genome. I’m not sure any biologist literally believes that—it ’s obvious, for example, that environmental cues can change gene expression patterns via epigenetics. But when we think about, say, the developmental process of an embryo, it is often implied that this unfolds according to the plan encoded in the genome. Francis Crick’s Central Dogma doesn’t exactly say that, but the two notions are easily elided. The Central Dogma says that sequence information passes from the gene to messenger RNA to the protein, but not back from the protein to DNA. People still argue about whether that is strictly true, but I’m not so interested in that debate: in the strict sense, the Central Dogma gives a fair account of what happens in transcription and translation. However, this does not mean the information in a gene is even the sole determinant of what a protein’s structure is—for example, alternative splicing of proteins happens in a context- dependent manner. It is certainly not the sole determinant of that a protein’s function is, because the role of a protein (especially one with a “disordered” structure) can be context- dependent and modified by other proteins. Even at this level, then, the actual information flow is complex.

When we think about cell signaling pathways generally, the picture is often not one of linear and prescriptive transmission of the signal. Again, the chains of molecular interaction can be complex (branched, for example) and context- dependent . There is often a fuzziness about which molecule interacts with which, rather than highly selective “ lock- and- key ” transmission of a signal. Some key processes in the cell’s information ecology, such as transcriptional and post- transcriptional gene regulation and mRNA splicing, involve groups of molecules such as proteins, noncoding RNAs, cofactors, and microproteins that work combinatorially and contextually via fuzzy rules of interaction—these really do look like molecular conversations.

The same is true at the level of how cells differentiate and acquire their eventual fates in the developing organism—that process isn’t deterministic in a sort of paint- by- numbers fashion, but comes about via a dialogue between cells, perhaps informed by larger- scale information such as deformation of a tissue layer. The neural networks of the brain wire up this way too: there is no prescribed program for that, and there simply can’t be, since there is no way the genome can encode all the information to specify trillions of synaptic junctions.

As for why such conversations are necessary: to be viable at all, all organisms (including bacteria) have to be error- tolerant—for example, robust against the inevitable stochasticity at the molecular level—and adaptive to contingencies of the environment. If there is simply a prescriptive program that has to be followed, it’s never going to work. Conversations allow the integration of contextual information to reach an appropriate decision. And the more complex the organism, the more important it is to have that flexibility and adaptability.

SEP: Going just a big deeper, in multiple chapters, you describe how emergent higher level structures and organizations influence micro level behavior. Biology builds the worlds within which it must evolve. You introduce two formal concepts downward causation and causal emergence. These may not be familiar to everyone. Can you explain the what these terms mean? And how do they operate, in say, tissues or proteins?

Ball: Downward causation seems to be a contested phenomenon in science, but it’s something of a mystery why that should be so. It says that sometimes what happens at a lower level can be caused by factors at a higher level. This happens every time you lift your coffee cup to take a sip. The atoms in the cup are moving not because microscopic events such as Brownian motion of the surrounding air molecules are pushing them, but because you, the agent, moved them. Some will say, ah, but you did that because of all those microscopic events in your brain, all those motions of sodium and potassium ions across the membranes of your neurons. But you will never be able to point to any actual cause in those microscopic events—at that level, causation has run out through your fingers.

This isn’t just a metaphorical way of speaking. There are various measures by which causation can be quantified, and it’s possible to show in some relatively simple model complex systems that measures of causal power are higher at high levels of the system, in which the fundamental components (like atoms) are course- grained into larger entities (like cells or brains), than at lower levels. Causation isn’t, in these cases, flowing from the bottom up.

So what we see in systems like that is casual emergence: causation from higher- level , emergent structures and systems. Again, this is nothing more than what we intuit already, but it can be formalized. No one says that the outcome of a football match was caused by the atoms of the players—that is as silly as it sounds. Yet for some reason there are people who think there is something mystical or spooky about causal emergence. They get confused by the distinction between a system’s behavior involving nothing more at the microscopic level than the known forces of nature, and that behavior being caused by those forces.

SEP: To clarify, the structures and capacities that evolve do so given chemical, informational, electrical, and, in particular, spatial and physical constraints. But this logic is deeper than the “form follows function subject to physical laws” logic described in Thompson’s seminal On Growth and Form. The claim is that the system evolves the structural constraints that enable robust functionality? Canalization would be an example, right?

Ball: Yes, this clearly has to be what happens, almost tautologously: only robust organisms will be evolutionarily viable. By robust, we might mean capable of surviving fluctuations, within certain bounds, in the environment but also stochasticity from the molecules up—a cell can’t, for example, depend on its molecules, moving by diffusion, all being present in a precise location at a precise concentration at a precise time. My interest is in what generic features of biological systems enable such robustness.

And this is what we are starting to understand at the molecular level. We see it, for example, in the way many biomolecules work combinatorially in groups—for example, the committees of transcription factors, cofactors, noncoding RNAS, and DNA regulatory regions that govern eukaryotic gene regulation, or the noncoding RNAs and RNA binding proteins that can make translational splicing context- dependent , or the microRNAs that influence post- transcriptional regulation. Combinatorial systems offer more with less—they can create many outcomes from a few components, much as olfactory receptors do—but they may also engender a tolerance to fluctuations, so that the right outcome will happen even if the precise group of components varies. We can also see how a reliance on mesoscale collective behaviors, for example, in the formation of biomolecular condensates through aggregation and phase separation to create partitioning of space and molecular segregation, can be more robust than some process that demands a precise chain of molecular signaling.

Robustness can also come from molecular promiscuity—the way, say, intrinsically disordered proteins can interact with many partners, creating a degree of redundancy in protein interaction networks that is not simply about making tightly prescribed chains of communication. This seems to be increasingly important as systems get more complex: the amount of intrinsic disorder in the proteome is correlated with the organismal complexity.

But yes, canalization is another route to robustness. This refers to a term introduced by Conrad Waddington in his famous landscape picture, whereby cell lineages get channeled into particular types as they differentiate during the course of development. The key here is that the cell states don’t need to be exactly defined—as long as they are in the right ballpark (say, in terms of which genes are active and which are not), they’ll be in the right state. The modern equivalent of that image is supplied by dynamical systems theory, whereby cells have particular basins of attraction in the complex, high- dimensional landscapes of gene activity—in which case it generally turns out that the high- dimensional picture can be collapsed into a low- dimensional one in which only a few key genes matter for defining cell fate. We can see canalization too in the way the body tends to react in similar ways to different diseases—by inflammation, say, or coughing. That physiological canalization is surely a consequence of the fact that it is present too for healthy physiological functioning. The question then becomes that of what factors promote canalization: how do complex systems acquire relatively smooth dynamical landscapes rather than extremely rough ones where it becomes all too easy to jump into a different state?

SEP: Collective Intelligence readers like ideas and concepts that travel well across disciplines. These ideas, downward causation and causal emergence, seem relevant to people who study ants, humans, cities, social, political and economic systems, and even artificial intelligence. Put on your general science writer hat and walk us through how and why these ideas should be part of the broader scientific lexicon.

Ball: Well, this is of course the lexicon of complexity science generally! I’m going to nail my colors to the mast here by saying that I approach this from the perspective of someone who, long ago, studied statistical mechanics, phase transitions, and critical phenomena. From that perspective, many of the ideas that are sometimes paraded as novel insights from complexity theory seem to me to be old and well established ideas from statistical physics that go back to Maxwell, Boltzmann, and van der Waals. The notion that, in a complex system of many interacting components, one can see universal behaviors like phase transitions, critical phenomena, and even the same critical exponents, was of course pointed out by the likes of Pierre Curie and Ernst Ising. To my mind this is physics as fundamental as anything that quantum chromodynamics or string theory has to offer, in the sense that it speaks to very general laws of nature that arise in many-body systems. From that perspective, it is relatively unsurprising to me that we see things like collective behavior, phase diagrams, and criticality in animal groups, in social systems (like traffic) and in economics—and in biology and neuroscience. How could it be otherwise, unless we believe (and there has sometimes been this implication in biology) that such systems, especially living systems, somehow evade the laws of (statistical) physics? Evolution does not work with a tabula rasa—it works with material systems governed by the laws of collective dynamics and is going to need to work rather hard if it wants to suppress those. But why would it even do that, rather than taking advantage of such “order for free,” as indeed it does in things like biomolecular condensates?

My general take on this is that we have allowed too much exceptionalism in the disciplinary domains, especially in the way that biology has been presented as “governed by the genome”—which is to say, governed by fiat rather than subject to the same rules of collective behavior as the rest of the universe.

Page: One comes away overwhelmed by the complexity of life and awed by how it “works” (to borrow your term). I cannot help but be reminded of an observation of Thomas Schelling regarding cities:

“Tens of millions of people making billions of decisions every week about what to buy and what to sell and where to work and how much to save and how much to borrow and what orders to fill and what stocks to accumulate and where to move and what schools to go and what jobs to take and where to build the supermarkets and movie theaters and electric power stations, when to invest in buildings above ground and mine shafts underground and fleets of trucks and ships and aircraft – if you are in the mood to be amazed, it can amaze you that the system works at all.”

You describe many examples that might have appeared in Ripley’s Believe it or Not—rearranged tadpole eyes, the many tasks carried out by the sonic hedgehog protein, lopped off fish tails regrowing the same pattern. One need not even be in the mood to be amazed. Which did you find most amazing? What’s the shiniest marble in a book?

Ball: That’s a lovely way to put it! And yes, sociological metaphors are extremely useful in thinking about how life works—more so, very often, than machine metaphors. For me the most marvelous aspect of working on the book was that, to be honest, I started out with this bag of what seemed like unconnected stuff about transcription factors and embryology and developmental signaling pathways and vague ideas about agency, but little by little I began to see patterns and continuity and common themes—a sort of logic to it all. I don’t by any means want to imply that it all now makes sense and adds up to a clear and coherent story. But I think I can sense the way the story is headed. And since finishing the book, I have been seeing again and again—both in small ways, like seemingly obscure bits of molecular biology that actually make perfect sense within this new framework, and in big ways, like biologists I respect saying that perhaps it’s time for a new narrative—that this viewpoint seems to fit with what’s happening within the discipline. I feel we’re on the cusp of a significant change in the discourse of biology.

That’s perhaps too nebulous an answer to your question though, so let me give a specific example of something I’m very excited about biomolecular condensates. These are fairly loose little blobs or droplets of biomolecules, generally proteins but also often containing noncoding RNAs and other components, that can form spontaneously in cells and conduct a wide range of tasks, from gene regulation to stress protection, DNA repair, and molecule storage. They are compartments, rather like the familiar organelles—mitochondria , lysosomes and so on—but with no surrounding membrane. Some are more or less permanent features, like the nucleolus where the ribosomes are assembled. Many are transient, forming when needed and then dissolving again. They were initially thought to form by liquid- liquid phase separation, like vinegar droplets in the oil of salad dressing, and that might indeed be a valid way to think about them, but they generally exist as non- equilibrium structures with a range of materials properties.

Since they were first identified in 2009, they have been cropping up everywhere, involved in all manner of cell functions. But the crazy thing is that we had no idea of their existence until then. Or rather, biologists had been seeing them—in fact, ever since the early days of cell theory in the 1803s—but didn’t know what to do with them or how to think about them. So, they were given vague names: speckles, granules, bodies, but there was no unified picture of what they were or what they did. They’re a whole level of cell organization we overlooked, largely (I think) because they didn’t fit the conventional picture of how molecular biology worked. It was supposed to have this digital precision, not the messiness of liquid blobs. Condensates are a great example of how we’re being forced—by experiment and observation!—to think about how life works in new ways.

SEP: My previous question was a bit of set up. Schelling follows that observation with “Amazement, needn’t be admiration…”. In the systems you describe, many of the constraints were created by the process itself. Are their particular cases where your amazement and admiration aligned, and some where they didn’t? In other words, were there times when you thought “evolution did what?”

Ball: OK, well evolution made the proteins of more complex organisms less structured, on average: as I said above, there is more intrinsic disorder in our proteome than in that of bacteria, and the correlation between such disorder and measures of organismal complexity is quite good. What the heck is going on there?! Why would you make the components of a more complex system “messier” or “fuzzier” than those of a less complex system?

I don’t see any way this can be just some kind of epiphenomenon—it looks like a deliberate strategy (to speak anthropomorphically). This realization was one of the penny- dropping moments in writing my book. As I hope the book shows, once you realize how this stuff seems to work, it makes perfect sense. That’s the only way such complex systems can work. No company can run by giving each employee a list of tasks they must conduct, come what may, without fail, every day. Imagine doing that for a sports team. “Machine thinking” makes no sense in biology. Disordered proteins are, by the way, very good at making biomolecular condensates. All these new things fit together.

SEP: I now want to turn to what many readers will find the most unexpected topic in the book: agency. The old school “DNA as blueprint metaphor” implies a machine-like view of organisms that have a “goal” (let me put that in quotes) of surviving now and in the future, that is, reproduction. Genes provide the blueprints. The best blueprints win out. Different blueprints for different niches. A robin here. An elephant seal there. Some sea plankton over there. You claim that what’s missing is meaning. Meaning leads directly to the concepts of purpose and agency. Quite a pivot. We jump from the deep end of one pool (biology) into another (philosophy of life). Central to your argument is the notion of affordances. You write “living things are not non equilibrium structures that happen to resist the disordering pull of entropic decay; they are structures that have evolved capabilities specifically to do so.” Elaborate on the biology facts that support that claim.

Ball: The facts are simply what we observe to be the case in the living world; what matters is the framing. That’s to say: I don’t think it is controversial that living organisms have goals. My goal right now is to see if I can find a way of putting this point concisely and persuasively. A bird might have, at some point, the immediate goal of finding food for its young. If we want to deny those things, we’re then not talking about biology.

But some will say: sure, we act for reasons and because of goals. But this is just a manner of speaking. Our genes have programmed us to behave in a way that ensures their replication. The first problem with that is obvious or should be: we have simply transferred all the agency, all the goal- directedness , to genes, which actually don’t possess those attributes at all, as they are (in the usual way of thinking) simply inert pieces of DNA. The response might be: of course they are, we’re just speaking metaphorically! But that’s no good, because it’s then the metaphor itself that has to do the actual work: the metaphor is what is meant to be acting in the real world. So, I’m saying that if we want to have a scientific explanation for an event, an action, we surely need to focus on the thing that acts.

What’s more, saying that all goals are somehow caused by genes “wanting” to replicate becomes a reductio ad absurdum. We can no doubt construct a story about how my wanting to put this point concisely and persuasively in my responses here is ultimately about my wanting to survive—which is me acting as a survival machine for my genes. But honestly, what an unscientific fairy tale that would be! The truth is that complex organisms like us have evolved brains precisely so that we don’t have to rely on genes “telling us what to do” in every circumstance, because of course genes can’t possibly do that, especially for situations we and our ancestors have never before encountered. The human brain, in order to have the versatility needed to make smart, adaptive decisions in all kinds of unfamiliar circumstances, has also to have become capable of all manner of things for which it is impossible to identify some unique, well supported and testable explanation, such as concocting the theory of general relativity. And while humans are unusual in some of our abstract cognitive capacities, there’s plenty of continuity too with the rest of the natural world.

Here, then, is what I mean by agency: the ability to change oneself and one’s environment in order to achieve (or try to achieve) some self- determined goal. I believe that all living organisms, even single- celled ones to some degree, have agency—that it is intrinsic to life itself. It does not, as some people want to insist, demand any capacity for deliberative and reflective decision- making (although in higher organisms it can include this). And making meaning is a necessity for all agents. What I mean by that is that an agent needs not only to be responsive to its environment but also to filter the information it gets: to be able to distinguish which signals it should attend to from those it can ignore. This is a requirement for any system that, among other things, needs to operate relatively efficiently, as any organism does. Meaning, at its most basic level, means just that: assigning value to information, so that some is “listened to” in decision- making and some is not. And I argue in the book that this is one way to think about the evolutionary process: it is very good at producing entities with agency, and which create meaning in their environment. There is absolutely no reason I can see to suppose that the evolutionary process itself has agency or meaning (even though it can be very hard to talk about it without using those terms!).

And just to be clear: there is nothing at all in what I have said that conflicts with the Neodarwinian view of evolution. This is not at all some “alternative theory” of biology or evolution! Some people argue that recognizing the agency of organisms does demand some revision, perhaps substantial revision, of traditional evolutionary theory. I follow those debates with interest but am largely agnostic about them.

SEP: You then slyly (and parenthetically) add that these capabilities are best regarded as “cognitive capabilities.” People should read the entire book for the full argument. I want to focus on two of these capabilities here: memory and prediction. Both associated with cognition. Take these one at a time. You claim biology should not be seen as an instantaneous process. It maintains a history. How? If possible, link this back to the concept of downward causation?

As for prediction: You cite work that shows how energy-efficiency implies the ability to predict. Unpack that for us?

Ball: Well, it might be simplest not to maintain a clear separation between these two notions of memory and prediction! Because I think what this comes down to is that organisms need to create and maintain an internal representation of their environment. That is itself a kind of memory, which is being constantly updated with new information, and it is what enables the organism to make predictions.

Here I am drawing on work by Susanne Still, Gavin Crooks, and others, who have shown that for any system to act in an energy-efficient way in a random or unpredictable environment, it must make predictions in order to correlate with the environment. In simple terms, it has to be able to avoid continually just bumping into things, literally or metaphorically. And they show that to achieve this, the entity has to gather information selectively: to distinguish what information is useful (leads to better predictions) from what isn’t. So, it needs a memory, but also needs to minimize the storage of useless information about the past.

An example of a simple use of memory in an organism with no brain is the way a bacterium like E. coli exhibits chemotaxis, swimming up a gradient in nutrient concentration. It can only measure that gradient by taking two readings of nutrient concentration in different regions—at different times—and comparing them to see if the concentration has risen (in which case it should keep going in the same direction) or fallen (in which case it might take a new direction at random). So it has at least to be able to remember and store one reading until it’s taken the next.

But history also matters for organisms because they maintain an internal state which, if you like, informs the decisions they take. Kevin Mitchell and Henry Potter put this nicely in their description of what an agent is: Any attempt to understand or explain the causes of an organism’s behavior is doomed to fail if it takes a purely instantaneous view of the physical system. It is not enough to account for how an organism behaves upon detecting some external stimulus or physiological state of affairs—the “triggering cause.” We must also understand why the system is configured such that it behaves in that way—the “structuring causes.”

This is obviously true of our brains, right?—what we do and how we respond to a stimulus depends partly on how we’re feeling! It’s remarkable how often “what the brain does” seems to be a good analogy for how organisms behave more generally, and it’s in this sense that I think a cognitive model of agency is useful.

SEP: Once a biological entity can predict, you claim that this enables them to experience the world as meaningful, and that they can choose how to best achieve their goals. Choose may be too strong a word here, but they can follow gradients to minimize the difference between what they are currently experiencing and what they would most prefer. (See Friston’s Free Energy Principle). If, as you write, organisms exert agency within a “web of meaning” this would imply that organisms in a common environment, say bacteria in my gut, could exhibit collective intelligence. Is that a useful frame?

Ball: I think so. Cells exhibit a kind of collective intelligence when they build a body during embryogenesis, for example. They are constantly signaling to one another—chemically, mechanically, electrically—and reconfiguring themselves accordingly. There’s nothing spooky about this. For example, cells in the inner cell mass—basically a blob of embryonic stem cells inside the cavity of a blastocyst in early development—differentiate into two types, epiblast cells (which form the fetus) and primitive endoderm (which form the yolk sac). But they do so somewhat at random, forming a conglomerate in which both cell types are scattered. Then they have to segregate so that the primitive endoderm forms a layer over the epiblast. They seem to do that by mechanical segregation, though the details are still not well understood—it’s analogous to a sort of phase separation, or a wetting process, but we can also see it as a kind of collective intelligence, a negotiation between cells. Bacteria certainly exhibit comparable, collective goal-directed behavior, for example, in the formation of biofilms.

And incidentally, to reinforce an earlier point: it’s a bit absurd to suppose that a structure like the layered epiblast embryo is somehow encoded in the genome. It’s like saying that the optical brilliance of a diamond is encoded in the carbon atom. Structures like this arise because biological cells have collective materials properties governed by physical laws. Again, the genome provides the molecular resources.

SEP: Last question. Organismal goals are not the goals of the evolutionary process. However, if organisms evolve agency to achieve goals, could this mean that the evolutionary process itself has a goal? As a social scientist, I’m predisposed to think that a system of purposive actors will itself exhibit purposeful behavior. How would a biologist react to this line of thinking? As you note, this seems out of alignment with the Modern Synthesis.

Ball: The idea that evolution itself might have a goal is generally regarded as heretical. And for good reason: it is not at all obvious how the Darwinian process of random mutation and natural selection could give rise to such a thing.

However, the question is now being asked, or at least something like it. Does evolution have directionality, at least? We have mercifully thrown off the old anthropocentric idea that evolution was all about making us as some kind of Darwinian pinnacle. Nonetheless, it does seem clear that biological evolution has created organisms with ever greater complexity over time—there are plenty of bacteria still around, of course, but there weren’t complex multicellular animals until around 560 million years ago, and animals with central nervous systems and complex brains and cognition are still more recent. Some researchers are asking if there might be a kind of natural law that impels the genesis of greater complexity, not just in biological evolution but in any evolving system that has some degree of selection—for example, as defined by thermodynamic or dynamic stability. If so, this might constitute an arrow of time independent of (but not inconsistent with) that imposed by the second law of thermodynamics. Certainly, the second law—the idea that entropy increases in all processes of change—does not obviously seem by itself to explain the increase in complexity over time in evolving systems.

Whether this bears on the issue of purpose is another matter. My own view is that Darwinian evolution is evidently a great way of making entities with purpose and agency. The question of whether that agency feeds back into the evolutionary process itself is a subtle and controversial one. But that does at least seem to be what happens for humans: by virtue of our agency, we are no longer at the mercy of natural selection. We have medicines and technology, and we can even edit our own genomes and, in principle, eliminate some deleterious gene variants that natural selection has not removed. So it is not just acceptable to ask how organismal agency might affect evolution; we have clear evidence that it can. I’m very interested in such questions but tend to take a conservative view about them simply as a default position that acknowledges my own ignorance about evolutionary biology.

SP: Thank you!