1.2 Information is Fundamental

Physical scientists have become increasingly committed to physicalism over the past four centuries. Physicalism is intentionally a closed-minded philosophy: it says that only physical things exist, where physical includes matter and energy in spacetime. It seems, at first glance, to be obviously true given our modern perspective: there are no ghosts, and if there were, we should reasonably expect to see some physical evidence of them. Therefore, all that is left is physical. But this attitude is woefully blind; it completely misses the better part our existence, the world of ideas. Of course, physicalism has an answer for that — thought is physical. But are we really supposed to believe that concepts like three, red, golf, pride, and concept are physical? They aren’t. But the physicalists are not deterred. They simply say that while we may find it convenient to talk about things in a free-floating, hypothetical sense, that doesn’t constitute existence in any real sense and so will ultimately prove to be irrelevant. From their perspective, all that is “really” happening is that neurons are firing in the brain, analogously to a CPU running in a computer and our first-person perspective of the mind with thoughts and feelings is just the product of that purely physical process.

Now, it is certainly true that the physicalist perspective has been amazingly successful for studying many physical things, including everything unrelated to life. However, once life enters the picture, philosophical quandaries arise around three problems:

(a) the origin of life,
(b) the mind-body problem and
(c) the explanatory gap.

In 1859, Charles Darwin proposed an apparent solution to (a) the origin of life in On the Origin of Species by Means of Natural Selection, or the Preservation of Favoured Races in the Struggle for Life. His answer was that life evolved naturally through small incremental changes made possible by competitive natural selection between individuals. The basic idea of evolution through naturally occurring “selections” is now nearly universally endorsed by the scientific community because a vast and ever-growing body of evidence supports it while no convincing evidence refutes it. But just how these small incremental changes were individually selected was not understood in Darwin’s time, and even today’s models are somewhat superficial because so many intermediate steps are unknown. Two big unresolved problems in Darwin’s time were the inadequate upper limit of 100 million years for the age of the Earth and the great similarity of animals from different continents. It was nearly a century before the earth was found to be 4.5 billion years old (with life originating at least 4 billion years ago) and plate tectonics explained the separation of the continents. By the mid-20th century, evolutionary theory had developed into a paradigm known as the Modern Synthesis that standardized notions of how variants of discrete traits are inherited. This now classical view holds that each organism has a fixed number of inherited traits called genes, that random mutations lead to gene variants called alleles, and that each parent contributes one gene at random for each trait from the two it inherited from its parents to create offspring with a random mixture of traits. Offspring compete by natural selection, which allows more adaptive traits to increase in numbers over time. While the tenets of the Modern Synthesis are still considered to be broadly true, what we have learned in the past seventy or so years has greatly expanded the repertoire of evolutionary mechanisms, substantially undermining the Modern Synthesis in the process. I will discuss some of that new knowledge later on, but for now, it is sufficient to recognize that life and the mind evolved over billions of years from incremental changes.

Science still draws a blank trying to solve (b) the mind-body problem. In 1637, René Descartes, after thinking about his own thoughts, concluded “that I, who was thinking them, had to be something; and observing this truth, I am thinking therefore I exist”1, which is popularly shortened to Cogito ergo sum or I think, therefore I am. Now, we still know that “three” is something that exists that is persistent and can be shared among us regardless of the myriad ways we might use our brain’s neurochemistry to hold it as an idea, so intuitively we know that Descartes was right. But officially, under the inflexible auspices of physicalism, three doesn’t exist at all. Descartes saw that ideas were a wholly different kind of thing than physical objects and that somehow the two “interacted” in the brain. The idea that two kinds of things exist at a fundamental level and that they can interact is called interactionist dualism. And I will demonstrate that interactionist dualism is the correct ontology of the natural world (an ontology is a philosophy itemizing what kinds of things exist), but not, as it turns out, the brand that Descartes devised. Descartes famously, but incorrectly, proposed that a special mental substance existed that interacted with the physical substance of the brain in the pineal gland. He presumed his mental substance occupied a realm of existence independent from our physical world which had some kind of extent in time and possibly its own kind of space, which made it similar to physical substance. We call his dualism substance dualism. We know now substance dualism is incorrect because the substance of our brains alone is sufficient to create thought.

Physicalism is an ontological monism that says only one kind of thing, physical things, exist. But what is existence? Something that exists can be discriminated on some basis or another as being distinct from other things that exist and is able to interact with them in various ways. Physical things certainly qualify, but I am claiming that concepts also qualify. They can certainly be discriminated and have their own logic of interactions. This doesn’t quite get us down to their fundamental nature, but bear with me I and I will get there soon. Physicalism sees the mind is an activity of the brain, and activities are physical events in spacetime, so it just another way of talking about the same thing. At a low level, the mind/brain consists of neurons connected in some kind of web. Physicalism endorses the idea that one can model higher levels as convenient, aggregated ways of describing lower levels with fewer words. In principle, though, higher levels can always be “reduced” to lower levels incrementally by breaking them down in enough detail. So we may see cells and organs and thoughts as conveniences of higher-level perspectives which arise from purely physical forms. I am going to demonstrate that this is false and that cells, organs, and thoughts do not fully reduce to physical existence. The physicalists are partly right. The mind is a computational process of the brain like digestion is a biological process of the gastrointestinal system. Just as computers bundle data into variables, thinking bundles data into thoughts and concepts which may be stored as memories in neurons. Computers are clearly physical machines, so physicalists conclude that brains are also just physical machines with a “mind” process that is set up to “experience” things. This view misses the forest for the trees because neither computers nor brains are just physical machines… something more is going on that physical laws alone don’t explain.

This brings us to the third problem, (c) the explanatory gap. The explanatory gap is “the difficulty that physicalist theories have in explaining how physical properties give rise to the way things feel when they are experienced.” In the prototypical example, Joseph Levine said, “Pain is the firing of C fibers”, which provides the neurological basis but doesn’t explain the feeling of pain. Of course, we know, independently of how it feels, that the function of pain is to inform the brain that something is happening that threatens the body’s physical integrity. That the brain should have feedback loops that can assist with the maintenance of health sounds analogous to physical feedback mechanisms, so a physical explanation seems sufficient to explain pain. But why things feel the way they do, or why we should have any subjective experience at all, does not seem to follow from physical laws. Bridging this gap is called the hard problem of consciousness because no physical solution seems possible. However, once we recognize that certain nonphysical things exist as well, this problem will go away.

We can resolve these three philosophical quandaries by correcting the underlying mistake of physicalism. That mistake is in assuming that only physical things can arise from physical causes. In one sense, it’s true: life and minds are entirely physical systems following physical laws. But in almost every way that matters to us, it is false: some physical systems (living things) can use feedback to perpetuate themselves and also to become better at doing so, which in turn creates in them the disposition to do so. This inclination, backed up by the capacity to pursue it by capturing and storing information, is not itself a physical thing, even though it exists in a physical system. For simplicity, I will usually just refer to this kind of existence as information, which is a term that usually refers to physically-encoded functionality, but it is understood to have meaning beyond the encoding. Often I will call it function, since information only exists to serve a function and disposition is managed through functional capacities. Information, functions, capacities, and dispositions are not physical and do exist. Ideas are information, but information is much more than just ideas. The ontology of science needs to be reframed to define and encompass information. Before I can do that, I am going to take a very hard look at what information is, and what it is not. I am going approach the subject from several directions to build my case, but I’m going to start with how life, and later minds, expanded the playing field of conventional physics.

During the billions of years before life came along, particle behaved in ways that could be considered very direct consequences of the Standard Model of particle physics and general relativity. This is not to say these theories are in their final form, but one could apply the four fundamental forces to any bunch of matter or energy and be able to predict pretty well what would happen next. But when life came along, complex structure started to develop with intricate biochemistries that seemed to go far beyond what the basic laws of physics would have predicted would happen. This is because living things are a information processing systems, or information processors (IPs) for short, and information can make things happen that would be extremely unlikely to happen otherwise. Organisms today manage heritable information using DNA (or RNA for some viruses) as their information repository. While the information resides in the DNA, its meaning is only revealed when it is translated into biological functions via biochemical processes. Most famously, the genetic code of DNA uses four nucleotide letters to spell 64 3-letter words that map to twenty amino acids (plus start and stop). Technically, DNA is always transcribed first into RNA and from RNA into protein. A string of amino acids forms a protein, and proteins do most of the heavy lifting of cell maintenance. But only two percent of the DNA in humans codes for proteins. Much of the rest regulates when proteins get translated, which most critically controls cell differentiation and specialization in multicellular organisms. Later on I will discuss some other hypothesized additional functions of non-coding DNA that substantially impact how new adaptations arise. But, as regards the storage of information, we know that DNA or RNA store the information and it is translated one-way only to make proteins. The stored information in no way “summarizes” the function the DNA, RNA or proteins can perform; only careful study of their effects can reveal what they do. Consequently, knowing the sequence of the human genome tells us nothing about what it does; we have to figure out what pieces of DNA and RNA are active and (where relevant) what proteins they create, and then connect their activities back to the source.

Animals take information a step further by processing and storing real-time information using neurochemistry in brains2. While other multicellular organisms, like plants, fungi, and algae, react to their environments, they do so very slowly from our perspective. Sessile animals like sponges, corals, and anemones also seem plantlike and seem to lack coordinated behavior. Mobile animals encounter a wide variety of situations for which they need a coordinated response, so they evolved brains to assess and then prioritize and select behaviors appropriate to their current circumstances. Many and perhaps all animals with brains go further still by using agent-centric processes called minds within their brains that represent the external world to them through sensory information that is felt or experienced in a subjective, first-person way. Then first-person thinking contributes to top-level decisions.

While nobody disputes that organisms and brains use information, it is not at all obvious why this makes them fundamentally different from, say, simple machines that don’t use information. To see why they are fundamentally different, we have to think harder about what information really is and not just how it is used by life and brains. Colloquially, information is facts (as opposed to opinions) that provide reliable details about things. More formally, information is “something that exists that provides the answer to a question of some kind or resolves uncertainty.” But provides answers to whom? The answer must be to an information processor. Unless the information informs “someone” about something, it isn’t information. But this doesn’t mean information must be used to be information; it only has to provide answers that could be used. Information is a potential or capacity that can remain latent, but must potentially be usable by some information processor to do something. So what is fundamentally different about organisms and brains from the rest of the comparably inert universe is that they are IPs, and only IPs can create or use information.

But wait, you are thinking, isn’t the universe full of physical information? Isn’t that what science has been recording with instruments about every observable aspect of the world around us in ways that are quite objectively independent of our minds’ IPs? If we have one gram of pure water at 40 degrees Fahrenheit at sea level at 41°20’N 70°0’W (which is in Nantucket Harbor), then this information tells us everything knowable by our science about that gram of matter, and so could be used to answer any question or resolve any uncertainty we might have about it. Of course, the universe doesn’t represent that gram of water using the above sentence, it uses molecules, of which there are sextillions in that gram. One might think this would produce astronomically complex behavior, but the prevailing paradigm of physics claims a uniformity of nature in which all water molecules behave the same. Chemistry and materials science then provide many macroscopic properties that work with great uniformity as well. Materials science reduces to chemistry, and chemistry to physics, so higher-level properties are conveniences of description that can be reduced to lower-level properties and so are not fundamentally different. So, in principle, then, physical laws can be used to predict the behavior of anything. Once you know the structure, quantity, temperature, pressure, and location of anything, then the laws of the universe presumably take care of the rest. Our knowledge of physical laws is still a bit incomplete, but it is good enough that we can make quite accurate predictions about all the things we are familiar with.

Physical information clearly qualifies as information once we have taken it into our minds as knowledge, which is information within our minds’ awareness. But if we are thinking objectively about physical information outside the context of what our minds are doing, that means we are thinking of this information as being present in the structure of matter itself. But is that information really in the matter itself? Matter can clearly have different structures. First, it can differ in the subatomic particles that comprise it, and there are quite a variety of such particles. Next, how these particles combine into larger particles and then atoms and then molecules can vary tremendously. And finally, the configurations into which molecules can be assembled into crystalline and aggregate solids is nearly endless. Information can describe all these structural details, and also the local conditions the substance is under, which chiefly include quantity, temperature, pressure, and location (though gravity and the other fundamental forces work at a distance, which make each spot in the universe somewhat unique). But while we can use information to describe these things, is it meaningful to say the information is there even if we don’t measure and describe it? Wouldn’t it be fair to say that information is latent in physical things as a potential or capacity which can be extracted by us as needed? After all, I did say that information is a potential that doesn’t have to be used to exist.

The answer is no, physical things contain no information. Physical information is created by our minds when we describe physical things, but the physical things themselves don’t have it. Their complex structure is simply physical and that is it. The laws of the universe then operate uniformly at the subatomic level as particles or waves or whatever they really are. The universe doesn’t need to take measurements or collect information just as a clock doesn’t; it just ticks. It is a finite state machine that moves ahead one step at a time using local rules at each spot in the universe. This explanation doesn’t say how it does that or what time is, but I am not here to solve that problem. It is sufficient to know that outside of information processors, the universe has no dispositions, functions, capacities or information. Now, how close particles get to each other affects what atoms, molecules and aggregate substances form, and can create stars and black holes at high densities. But all this happens based on physical laws without any information. While there are patterns in nature that arise from natural processes, e.g. in stars, planets, crystals, and rivers, these patterns just represent the rather direct consequences of the laws of physics and are not information in and of themselves. They only become information at the point where an IP creates information about them. So let’s look at what life does to create information where none existed before.

Living things are complicated because they have microstructure down to the molecular level. Cells are pretty small but still big enough to contain trillions of molecules, all potentially doing different things, which is a lot of complexity. We aren’t currently able to collect all that information and project what each molecule will do using either physics or chemistry alone. But we have found many important biochemical reactions that illuminate considerably how living things collect and use energy and matter. And physicalism maintains that given a complete enough picture of such reactions we can completely understand how life works. But this isn’t true. Perfect knowledge of the biochemistry involved would still leave us unable to predict much of anything about how a living thing will behave. Physical laws alone provide essentially no insight. Our understanding of biological systems depends mostly on theories of macroscopic properties that don’t reduce to physical laws. We are just used to thinking in terms of biological functions so we don’t realize how irreducible they are. Even at a low level, we take for granted that living things maintain their bodies by taking in energy and materials for growth and eliminating waste. But rocks and lakes don’t do that, and nothing in the laws of physics suggests complex matter should organize itself to preserve such fragile, complex, energy-consuming structures. Darwin was the first to suggest a plausible physical mechanism: incremental change steered by natural selection. This continues to be the only idea on the table, and it is still thought to be correct. But what is still not well appreciated is how this process creates information.

At the heart of the theory of evolution is the idea of conducting a long series of trials in which two mechanisms compete and the fitter one vanquishes the less fit and gets to survive longer as its reward. In practice, the competition is not head-to-head this way and fitness is defined not by the features of competing traits but by the probability that an organism will survive and replicate. So all I mean by “more fit” is able to survive longer. Birds can and apparently have evolved into less physically-fit forms to produce more spectacular plumage, possibly sometimes even evolving themselves into extinction3. Also, under conventional evolutionary theory, simpler forms can emerge with equal likelihood of more complex forms, leading to things like blind cave fish4. Genetic recombination provided by sexual reproduction allows the fitness of each trait to evolved independently of the fitness of individual organisms. No one trait may make a life-or-death difference, but over time, the traits that support survival better will outcompete and displace less capable traits. Finally, evolution includes the idea that mutation can change traits or create new ones. If you look over this short summary of evolution, you can see the places where I implicitly departed from classical physics and invoked something new by using words like, “traits”, “fitness”, “probability”, and “compete”. These words are generalizations whose meaning relative to evolution is lost as soon as we think about them as physical specifics. Biological information is created at the moment that feedback from one or more situations is taken as evidence that can inform a future situation, which is to say that it can give us better than random odds of being able to predict something about that future situation. This concept of information is entirely nonphysical; it is only about similarities of features, where features themselves are informational constructs that depend on being able to be recognized with better than random odds. Two distinct physical things can be exactly alike except for their position in time and space, but we can never prove it. All that we can know is that two physical things have observable features which can be categorized as the same or different based on some criteria. These criteria of categorization, and the concept of generalized categories, are the essence of information. For now, let’s focus only on biological information captured by living organisms in DNA and not on mental information managed by brains. Natural selection implies that biological information is created by inductive logic, which consists of generalizations about specifics whose logical truth is a matter of probabilities rather than logical certainty. Logic produces generalities, which are not physical things one can point to. And the inductive trial-and-error of evolution creates and preserves traits that carry information, but it doesn’t describe what any of those traits are. Furthermore, any attempt to describe them will itself necessarily be an approximate generalization because the real definition of the information is tied to its measure of fitness, not to any specific effects it creates.

We know that evolution works as we are here as evidence, but why did processes that collected biological information form and progress so as to create all the diverse life on earth? The reason is what I call the functional ratchet, and also previously called an arms race. A ratchet is a mechanical device that allows motion in only one direction, as with a cogged wheel with backward angled teeth. Let’s call the fitness advantage a given trait provides its function. More generally capable functions will continuously displace less capable ones over time because of competition. This happens in two stages. First, useful functionality provided by entirely new traits will tend to persist because it provides capabilities other organisms lack. Second, variants of the same trait compete head to head to improve each trait continuously. It is often said that evolution is directionless and human beings are not the “most” evolved creatures at the inevitable endpoint, but this is an incorrect characterization of what is happening. Evolution is always pulled in the direction of greater functionality by the functional ratchet. What functionality means is local to the value each trait contributes to each organism at each moment, so because circumstances change and there is a wide variety of ecological niches, evolution has no specific target and no given function will necessarily ever become advantageous. But the relentless pull toward greater functionality has great potential to produce ever more complex and capable organisms, and this is why we see such a large variety. Simpler forms, like blind cave fish, have actually become increasingly functional relative to the needs of living in dark caves. The functionality that is lost because it is no longer sufficiently useful doesn’t violate the one-way movement of the ratchet; it has just become irrelevant because the ratchet changed. So it is not at all a coincidence that life is more diverse now than in the past or that human intelligence evolved. I will discuss later on how the cognitive ratchet created human brains in the evolutionary blink of an eye5.

Note that while the word function suggests we can list the effects the trait can cause in advance, I am using it in a more abstract sense to include any general effects it can cause whether they are knowable or not. In practice, because any effects caused by the trait in specific situations are more likely to be preserved over time if they create net benefits to survival, a collection of effects that are probably more helpful than not overall are likely to evolve for the trait, given that it can change over time. The trait has arguably been causing effects continuously for millions to billions of years, all of which have contributed probabilistically to the trait’s current functionality. However, for entirely physical reasons, traits are likely to be highly specialized, usually having just one fairly obvious functional effect. Any given protein coded by DNA can only have a small, finite number of effects, and it will likely only be used for effects for which it does a better job than any other trait. My point is that the exact benefits and costs can be very subtle and any understanding we acquire is likely to overlook such subtleties. Beyond subtleties, cases of protein moonlighting, in which the same protein performs quite unrelated functions, are now well-documented. In the best-known case, some crystallins can act both as enzymes that catalyze reactions or as transparent structural material of eye lenses.6 But even proteins that can only perform one enzymatic function can use that function in many contexts, effectively creating many functions.

Induction, the idea that function is a byproduct of a long series of trial and error experiments whose feedback has been aggregated, is sufficient to explain evolution, but the mind also uses deduction. I noted before that where induction works from the bottom up (from specifics to generalities), deduction works from the top down (generalities to specifics). From the deductive perspective, we see functions in their simplified, discrete forms which cause specific effects. The body is an agent, limbs are for locomotion, eyes are for seeing, hearts are for pumping blood, gullets are for taking in food, etc. Viewed this way, these functions describe clear contributions to overall survival and fitness, and detailed study always reveals many more subtle subsidiary functions. Of course, we know that evolution didn’t “design” anything because it is used trial and error rather than discrete deductive causes and effects, but we know from experience that deduction can provide very helpful and hence functional support, even though it is not the way the world works. Why and how it does this I will get into later, but for now, let’s review how deduction sees design problems. Deduction begins with a disposition, which is a tendency toward certain actions, that becomes an intent, which is an identified inclination to achieve possible effects or goals. Effects and goals are inherently abstractions in that they don’t refer to anything physical but instead to a general state of affairs, for which the original and driving state of affairs concerning life is to continue to survive. The manipulation of abstractions as logical chess pieces is called deductive reasoning. Techniques to reach goals or purposes are called strategies, designs, or planning. I call the actions of such techniques maneuvers. All these terms except disposition, function, cause, and effect are strictly deductive terms because they require abstractions to be identified. I will expand more in the next chapter how disposition, functionality, and causality (cause and effect) can be meaningful in inductive contexts alone without deduction. My point, for now, is that while evolution has produced a large body of function entirely by inductive means, deductive means can help us a lot to understand what it has done. Provided we develop an understanding of the limitations of deductive explanations, we can be well-justified in using them. I am not going to credit the philosophy of biology with fully exploring those limitations, but we can safely say they are approximately understood, and so on this basis it is reasonable for biologists both to use deductive models to explain life and to characterize evolutionary processes as having intent and designs. There is, however, an unspoken understanding among biologists that the forces of evolution, using only inductive processes, have created something functional that can be fairly called functional. This something has to sit beneath the surface of their deductive explanations because all explanations must be formed with words, which are themselves abstract tools of deductive logic. In other words, information and function are very much present in all living structures and has largely been recorded in DNA, and this information and function are not physical at all. Physicalists go a step too far, then, by discounting the byproducts of inductive information processes as the incidental effects of physical processes.

Although it is possible to create, collect, and use information in a natural universe, it is decidedly nontrivial, as the complexity of living things demonstrates. Beyond the already complex task of creating it with new traits, recombination, and natural selection, living things need to have a physical way of recording information and transcribing information so that it can be deployed as needed going forward. I have said how DNA and RNA do this for life on earth. Because of this, we can see the information of life captured in discrete packages called genes. DNA and RNA are physical structures, and the processes that replicate and translate them are physical, but as units of function, genes are not physical. Their physical components should be viewed as a means to an end, where the end is the function. It is not a designed end, an inductively-shaped one. The physical shapes of living structures are cajoled into forms that would have been entirely unpredictable based on forward-looking design goals, but which patient trial and error demonstrated are better than alternatives.

Beyond biological information, animals have brains that collect and use mental information in real time that is stored neurologically. And beyond that, humans can encode mental information as linguistic information or representational information. Linguistic information can either be in a natural language or a formal language. Natural languages assume a human mind as the IP, while formal languages declare the permissible terms and rules, which is most useful for logic, mathematics, and computers. Representational information simulates visual, audio or other sensory experience in any medium, but most notably nowadays in digital formats. And finally, humans create artificial information, which is information created by computer algorithms, most notably using machine learning. All of these forms of information, like biological information, answer questions or resolve uncertainties to inform a future situation. They do this by generalizing and applying nonphysical categorical criteria capable of distinguishing differences and similarities. Some of this information is inductive like biological information, but, as we will see, some of it is deductive, which expands the logical power of information.

We have become accustomed to focus mostly on encoded information because it can be readily shared, but all encodings presume the existence of an IP capable of using them. For organisms, the whole body processes biological information. Brains (or technically, the whole nervous and endocrine systems) are the IP of mental information in animals. Computers can act as the IPs for formal languages, formalized representations, and artificial information, but can’t process natural languages or natural representational information. However, artificial information processing can simulate natural information processing adequately for many applications, such as voice recognition and self-driving cars. My point here is that encoded information is only an incremental portion of any function, which requires an IP to be realized as function. We can take the underlying IPs for granted for any purpose except understanding how the IP itself works, which is the point of this book. While we have a perfect knowledge of how electronic IPs work, we have only a vague idea of how biological or mental information processors work.

Consider the following incremental piece of biological information. Bees can see ultraviolet light and we can’t. This fact builds on prevailing biological paradigms, e.g. that bees and people see light with eyes. This presumes bees and people are IPs for which living and seeing are axiomatic underlying functions. The new incremental fact tells us that certain animals, namely bees, see ultraviolet as well. This fact extends what we knew, which seems simple enough. A child who knows only that animals can see and bees are small flying animals that like flowers can now understand how bees see things in flowers that we can’t. A biologist working on bee vision needs no more complex paradigm than the child; living and seeing can be taken for granted axiomatically. She can focus on the ultraviolet part without worrying about why bees are alive or why they see. But if our goal is to explain bees or minds in general, we have to think about these things.

Our biological paradigm needs to define what animals and sight are, but the three philosophical quandaries of life cited above stand in the way of a detailed answer. Physicalists would say that lifeforms are just like clocks but more intricate. That is true; they are intricate machines, but, like clocks, an explanation of all their pieces, interconnections, and enabling physical forces says nothing about why they have the form they do. Living things, unlike glaciers, are shaped by feedback processes that gradually make them a better fit for what they are doing. Everything that happened to them back to their earliest ancestors about four billion years ago has contributed. A long series of feedback events created biological information leveraging inductive logic captured as information rather than using laws of physics alone. Yes, biological IPs leverage physical laws, but they add something important which for which the physical mechanisms are just the means to the end. The result is complex creations that have essentially a zero probability of arising by physical mechanisms alone.

How, exactly, do these feedback processes that created life create this new kind of entity called information and what is information made out of? The answer to both questions is actually the same definition given for information above: the reduction of uncertainty, which can also be phrased as an ability to predict the future with better odds than random chance. Information is made out of what it can do, so we are what we can do. We can do things with a pretty fair expectation that the outcome will align with our expectations of the outcome. It isn’t really predicting in a physical sense because we see nothing about the actual future and any number of things could always go wrong with our predictions. We could only know the future in advance with certainty if we had perfect knowledge of the present and a perfectly deterministic universe. But we can never get perfect knowledge because we can’t measure everything and because quantum uncertainty limits how much we can know about how things will behave. But biological information isn’t based on perfect predictions, only approximate ones. A prediction that is right more than it is wrong can arise in a physical system if it can use feedback from a set of situations to make generalized guesses about future situations that can be deemed similar. That similarity, measured any way you like, carries predictive information by exploiting the uniformity of nature, which usually causes situations that are sufficiently similar to behave similarly. It’s not magic, but it seems like magic relative to conventional laws of physics, which have no framework for measuring similarity or saying anything about the future. A physical system with this capacity is exceptionally nontrivial — living systems took billions of years to evolve into impressive IPs that now centrally manage their heritable information using DNA. Animals then spent hundreds of millions of years evolving minds that manage real-time information using neurochemistry. Finally, humans have built IPs that can manage information using either standardized practices (e.g. by institutions) or computers. But in each case the functional ratchet has acted to strongly conserve more effective functionality, pulling evolution in the direction of greater functionality. It has often been said that evolution is “directionless”, because it seems to pull to simplicity as much as toward complexity. As Christie Wilcox put it in Scientific American, “Evolution only leads to increases in complexity when complexity is beneficial to survival and reproduction. … the more simple you are, the faster you can reproduce, and thus the more offspring you can have. … it may instead be the lack of complexity, not the rise of it, that is most intriguing.”7 It is true, evolution is not about increasing complexity, it is about increasing functionality. Inductive trial and error always chooses more functionality over less, provided you define “more” as what induction did. In other words, it is a statistical amalgamation of successful performances where the criteria for each success was situation-specific. But what is “success”? Natural selections not made by brains depend only on the impact to survival and reproduction, but minds make real-time selections based on high-level models (desires) which must have evolved for reasons that aligned with survival or reproduction, but which may no longer accurately reflect a benefit to survival.

A functional entity has the capacity to do something useful, where useful means able to act so as to cause outcomes substantially similar to outcomes seen previously. To be able to do this, one must also be able to do many things that one does not actually do, which is to say one must be prepared for a range of circumstances for which appropriate responses are possible. Physical matter and energy are comprised of a vast number of small pieces whose behavior is relatively well-understood using physical laws. Functional entities are comprised of capacities and generalized responses based on those capacities. Both are natural phenomena. Until information processing came along through life, function (being generalized capacity and response) did not exist on earth (or perhaps anywhere). But now life has introduced an uncountable number of functions in the form of biological traits. As Eugene Koonin of the National Center for Biotechnology Information puts it, “The biologically relevant concept of information has to do with ‘meaning’, i.e. encoding various biological functions with various degrees of evolutionary conservation.”8 The mechanism behind each trait is itself purely physical, but the fact that the trait works across a certain range of circumstances is because “works” and “range” generalize abstract capacities, which one could call the reasons for the trait. The traits don’t know why they work, because knowledge is a function of minds, but their utility across a generalized range of situations is what causes them to form. That is why information is not a physical property of the DNA, it is a functional property.

Function starts to arise independent of physical existence at the moment a mechanism arises that can abstract from a token to a type, and, going the other way, from a type to a token. A token is a specific situation and a type is a generalization of that token to an abstract set of tokens that could be deemed similar based on one or more criteria. Each criterion permits a range of values that could be called a dimension, and so divides the full range of values into categories. Abstracting a token to a type is a form of indirection and is used all the time in computers, for example to let variables hold quantities not known in advance. An indirect reference to a token can either be a particular, in which case it will only ever refer to that one token, or a generality, in which case it is a type referring to the token. By referring to tokens through different kinds of references we can apply different kinds of functionality to them. Just as we can build physical computers that can use indirection, biological mechanisms can implement indirection as well. I am not suggesting that all types are representational; that is too strong a position. Information is necessarily “about” something else, but only in the sense that its collection and application must move between generalities and specifics. Inductive trial-and-error information doesn’t know it employs types because only minds can know things, but it does divide the world up this way. When we explain inductive information deductively with knowledge, we are simplifying what is happening by making analogies to cause-and-effect models even though they really use trial-and-error models. Cells have general approaches for moving materials across cell membranes which we can classify as taking resources in and expelling wastes, but the cells themselves don’t realize they have membranes and the simplification that materials are resources or waste neglects cases where they are both or neither. Sunlight is important to plants, so sunlight is a category plants process, which is to say they are organized so as to gather sunlight well, e.g. by turning their leaves to the sun, but they don’t pass around messages representing sunlight as a type and instructing cells to collect it.

To clarify further, we can now see that function is all about applying generalities to specifics using indirect references, while physical things are just about specifics. [Now is a good time to point out that “generalize” and “generalization” mean the same thing as “general” and “generality” except that a generalization is created by inference from specific cases, while a generality is unconcerned with whether it was created with inductive or deductive logic. Because I will argue that deductive logic can only be applied by aligning it to inductive findings, I will use the terms interchangeably but according to the more fitting connotation.] We can break generalities down into increasingly specific subcategories, arriving eventually at particulars.

Natural selection allows small functional changes to spread in a population, and these changes are accompanied by small DNA changes that caused them. The physical change to the DNA caused the functional change, but it is really that functional change that brought about the DNA change. Usually, if not always, a deductive cause-and-effect model can be found that accounts for most of the value of an inductive trial-and-error functional feature. For example, hearts pump blood because bodies need circulation. The form and function line up very closely in an obvious way. We can pretty confidently expect that all animals with hearts will continue to have them in future designs to fulfill their need for circulation. While I don’t know what genes build the circulatory system, it is likely that most of them have contributed in straightforward ways for millions of years.

Sex, on the other hand, is not as stable a trait. Sometimes populations benefit from parity between the sexes and sometimes with disproportionally more females. Having more females is beneficial during times of great stability, and having more males during times of change. I will discuss why this is later, but the fact that this pressure can change makes it advantageous sometimes for a new mechanism of sex determination to spring up. For example, all placental mammals used to use the Y chromosome to determine the sex of the offspring. Only males have it, but males also have an X chromosome. With completely random recombination, this means that offspring have a 50% chance of inheriting their father’s Y chromosome and being male. However, two species of mole voles, small rodents of Asia, have no Y chromosome, so males have XX chromosomes like females. We don’t know what trigger creates male mole voles, but a mechanism that could produce more than 50% females would be quite helpful to the propagation of polygamous mole vole populations, as some are, because they would be more reproducing (i.e. female) offspring.91011 The exact reason a change in sex determination was more adaptive is not relevant, all that matters is that it was and the physical mechanism was simply abandoned. A physical mechanism is necessary, and so only possible physical mechanisms can be employed, but the selection between physical mechanisms is not based on their physical merits but only on their functional contribution. As we move into the area of mental functions, the link between physical mechanisms and mental functions becomes increasingly abstract, effectively making the prediction of animal behavior based on physical knowledge alone impossible. To understand functional systems we have to focus on what capacities the functions bring to the table, not on the physical means they employ.

I have introduced the idea that information and the function it brings are the keys to resolving the three philosophical quandaries created by life. In the next chapter, I will develop it into a comprehensive ontology that is up to the task of supporting the scientific study of all manner of things.

  1. René Descartes, Discourse on the Method, 1637, Oxford University Press, 2006, part four, p 28
  2. Though plants, and probably all lifeforms, also have limited real-time information processing and memory. Learning was demonstrated in plants by Monica Gagliano, see Gagliano, M. et al., Learning by Association in Plants, Sci. Rep. 6, 38427; doi: 10.1038/srep38427 (2016).
  3. Fisherian runaway, Wikipedia
  4. Mexican tetra, Wikipedia
  5. Alexander Riegler, The Cognitive Ratchet: The Ratchet Effect as a Fundamental Principle in Evolution and Cognition, CLEA, Vrije Universiteit Brussels, 2001
  6. Graeme Wistow and Joram Piatigorsky, Recruitment of Enzymes as Lens Structural Proteins, Science
    New Series, Vol. 236, No. 4808 (Jun. 19, 1987), pp. 1554-1556
  7. , Evolution: The Rise of Complexity
    , Scientific American, January 16, 2012
  8. Eugene V. Koonin, The meaning of biological information, Philos Trans A Math Phys Eng Sci. 2016 Mar 13; 374(2063): 20150065.
  9. Jennifer A Marshall Graves and Catherine L Peichel, Are homologies in vertebrate sex determination due to shared ancestry or to limited options?
  10. Transcaucasian mole vole, Wikipedia
  11. Emily Singer, The Incredible Shrinking Sex Chromosome, Quanta Magazine

Leave a Reply