Saturday 25 June 2016

Splitting and simplification

Splitting and simplification are common operations which scientists routinely apply. Splitting can be applied to almost anything, but it is most commonly applied to the subject matter of scientific theories. Simplification is an operation that is applied to scientific theories themselves. In science, splitting subject matter up is sometimes given a technical name: reductionism.

Reductionism has become a bizarrely unpopular part of science. The most common complaint is that splitting things up leads to a focus on the parts and ignoring their interactions. Sometimes it is emergent properties that reductionism supposedly neglects.

That gets us on to the topic of simplification. Simplification is another basic tool in the scientific toolkit. The most common associations for many scientists when they think of simplicity are Occam's razor and the parsimony principle. However these are the tip of the iceberg. Simple theories act as foundations for more complex theories. It is easy to make complex theories that completely explain data - the problem is that such theories are poor at making predictions. To get a good predictive theory you need to find smaller theories.

I would say more about the virtues of simplicity. However, Boyd and Richerson have a fine essay titled: Simple Models of Complex Phenomena: The case of cultural evolution The essay is targeted at anthropologists - who are often complexity worshippers. Simple theories are incorrect theories, they seem to think. Perhaps because so many of their colleagues are confused about the topic, Boyd and Richerson go to considerable lengths to argue the virtues of simplicity.

Many social scientists don't like simplification. They think accuracy is more important. Jonathan Haidt expresses this perspective here:

I'm actually a anti parsiomonist: I'm opposed to the pursuit of parsimony. I take Occam very literally. Let's pursue truth. If two theories are equally good at describing reality take the simplest - but don't pursue simplicity. Whoever made us didn't give a damn about simplicity, so stop pursuing simplicity in psychology.

This attitude is a confused one which fails to understand why scientists prioritize simplicity so highly.

Many of the complaints about reductionism are about simplification. The fear is that after splitting things up, interactions between the components will be ignored, or the diversity of the components will be neglected. Perhaps it should be stressed more that splitting and simplification are technically orthogonal operations.

As far as terminology goes, the term "reductionism" has some weak points - it is an "-ism" - like Mormonism and Catholoicism. The term "reductionism" sounds as though it denigrates that which it reduces. Reduce" doesn't mean "split" - so the term doesn't say what it means: it is confusing. Lastly, reductionism is jargon. I think I am going to start using the terms "splitting" and "simplification" instead.

Saturday 18 June 2016

The Balkanization of Darwinism

In modern times the study of Darwinian evolution has become split across multiple academic departments. There are evolutionary biology departments, ecology departments and departments of human evolution. Other departments study evolutionary economics and evolutionary epistemology. As far as I can tell, practically nowhere studies evolutionary theory itself.

The success of the field of evolutionary biology seems to be part of the problem to me. There are a number of academics who self-identify as "evolutionary biologists". There are evolutionary biology departments at universities all over the world. It seems to me that students of evolutionary theory who confine their attention to biology are missing out on Darwinian physics, and applications of evolutionary theory outside biology. Is this just a case of specialization? or do these folk not understand that Darwinism is more broadly applicable? Experience suggests that the latter hypothesis is usually the more accurate one.

For example, Mark Ridley's "Evolution" textbook says (3rd edition, page 4):

Evolution means change in living things by descent with modification

"Living things"? Since when is evolutionary theory limited to "living things". What about Darwinian physics? To me that's a classic example of the confusion associated with evolutionary biology. Those folk think they have a monopoly on evolutionary theory. What they actually have is bad terminology which is confusing the next generation of students.

Larry Moran once wrote: "Call me an evolutionary biologist". Well, OK - but it seems like a term of abuse to me. When I refer to people as "evolutionary biologists" I am usually referring to people who don't understand the true scope of Darwin's legacy.

Cancer evolution

For a long time, immune system evolution and brain evolution were some of the best documented-examples of evolution taking place within organisms as part of development (assuming that we don't count symbiotes such as parasites and gut bacteria). Such evolution is contrary to the textbook orthodoxy that changes to individuals over their own lifespan do not count as being a type of 'evolution'.

Ridley's "Evolution" textbook says (3rd edition, p.4):

Developmental change within the life of an organism does not count as evolution in the strict sense and the definition referred to evolution as "a change between generations" to exclude developmental changes.

Of course, generations of a multi-cellular organism is one thing, and generations of its cells are another. Developmental changes can indeed be a form of evolution from the perspective of somatic cell generations. Now it looks as though the evolution of cancers is turning into another well-documented example of Darwinian evolution during development which involves somatic cell lines. Some quote illustrate the cancer-related discoveries:

Over the last 2 years, there have been an unprecedented number of publications focused on cancer evolutionary processes in solid and haematological cancers, a trend that is set to continue over the next decade. [...] It is increasingly clear that many advanced tumors follow a branched, Darwinian evolutionary trajectory. This has been demonstrated in childhood ALL [1], pancreatic cancer [2, 3], colorectal cancer [4], clear cell renal carcinoma [5, 6], breast cancer [7, 8] and prostate cancer [9] among others. [source]

...

Cancer development within an individual is also an evolutionary process, which in many respects mirrors species evolution. Species evolve by mutation and selection acting on individuals in a population; tumors evolve by mutation and selection acting on cells in a tissue. The processes of mutation and selection are integral to the evolution of cancer at every step of multistage carcinogenesis, from tumor genesis to metastasis. [source]

...

Iconic examples of evolution (birds evolving from dinosaurs, hominids evolving an upright posture, or a lineage of lobe-finned fish evolving four legs and moving onto land) might seem unrelated to the growth of a cancerous tumor, but the process underlying them both — natural selection — is identical. We typically think of natural selection acting among individuals, favoring those carrying advantageous traits and making those traits more common in the next generation. However, the key elements of this process — variation, inheritance, and selective advantage — characterize not just populations of organisms in a particular environment, but also populations of cells within our own bodies. The cells lining your intestines, for example, are not genetically uniform; there is variation among them. [source]

A wikipedia page offers a summary of the topic.

The topic is receiving interest partly since evolutionary theory is involved in the treatment of cancer.

Somatic evolution becoming more orthodox will help with acceptance of cultural evolution. For one thing, once the textbooks get rewritten, people will no longer be able to point at them, saying that changes to individuals over their own lifespan do not count as being 'evolution' - by the definition of the term.

Refactoring Darwinism: where to start

The term "refactoring" commonly refers to a set of useful computer programming techniques which involve making structural changes to a program without altering its function. Refactoring is often done before making functional changes. If you launch into making changes without some preliminary refactoring, you are often more likely to make mistakes. Refactoring frequently allows you to make changes without breaking tests - so you can verify that your changes don't break anything. If you need to make a bunch of changes, it sometimes helps to make ones that you can verify do no damage first.

I argued in refactoring science that science can benefit from refactoring too. I agree with the many folk who think that Darwinism is in need of some changes. This raises a number of issues about what changes need making. One such issue is where to start from. Darwinism has had two main identifiable releases: Darwin's Darwinism and the Modern Synthesis. Many seem to assume that we should build on the most recent release of the theory, the Modern Synthesis. For example, here's Massimo Pigliucci:

It makes just as little sense to talk of "Darwinism" in modern science as it does to talk about Mendelism or Newtonianism. The current theory of biological evolution is the Modern Synthesis, and if one wants to make the point that cultural evolution works in the same way, one needs to take on board the most refined version available of the theory, not its earliest draft. Incidentally, I think that's why we should avoid talk of Lamarckism as well as Darwinism altogether: they refer to murky (in the first instance) or outdated (in the second) ways of thinking about biology, and it doesn't help to resurrect them as if the last two centuries of science hadn't happened.

This rather assumes that the Modern Synthesis was an improvement over Darwinism. However, some have argued that the Modern Synthesis consists of a bunch of overstatements. For example, Mesoudi (2011) has argued that cultural evolution is Darwinian, but not Neo-Darwinian. The modern synthesis is often billed as uniting Darwin's idea of natural selection and Mendelian genetics. The problem is that Mendelian genetics only applies to inheritance via DNA. As a result, other forms of inheritance were sidelined by the synthesis. From such a perspective, the Modern Synthesis looks like a bit of a broken version of Darwinism. Failing to encompass cultural evolution is a pretty serious flaw. In this case, maybe we should start from an earlier revision of Darwinism - one without so many problems.

Starting with Darwin's theory should allow cherry picking any useful bits from the Modern Synthesis. Starting from the Modern Synthesis would probably be followed by bunch of 'revert' operations - to get rid of the dogmatic and mistaken bits. The former operation looks easier to me than the latter one.

I have some other concerns about starting from the Modern Synthesis. It is vague. Of course Darwin changed his mind about some things too - and the various editions of "The Origin" describe rather different theories. However, it seems to me that Darwinism is more clearly specified than the Modern Synthesis was. The Modern Synthesis can be difficult for critics to criticize - since there's no canonical version. In science, that's not a good thing.

In some respects, the Modern Synthesis illustrates how not to produce an evolutionary synthesis. It subsequently became a bit of a straitjacket for researchers. Inevitably any new theory of evolution will also be incomplete - but there's no need for it to be dogmatic and mistaken.

Sunday 12 June 2016

Susan Blackmore: A new form of evolution

Here's Sue in 2016:

Susan appears to have renamed "temes" as "tremes".

More from Sue on that topic:

Susan says:

I first called these replicators the ‘temes’ – for ‘technological memes’ – but people were so confused by the spelling that I have changed the name to ‘tremes’. I am very sorry if this causes any more confusion!

Saturday 11 June 2016

Universal Darwinism in a nutshell

Introduction

Darwinism's domain has expanded dramatically over the last century, to cover culture, development and other fields. This has led many thinkers to wonder how far the Darwinism goes. Universal Darwinism proposes that it goes far. This page offers a brief introduction to the basic ideas involved.

Basics

First, some basics: evolutionary theory is based on population thinking. It considers the world as composed of a population, an environment, and operations that act on the population members. It assumes that part of the world can be broken down into a set of similar entities - a population. A finite set of simple operations can then be applied to these entities. Entities can come into existence, go out of existence, make copies of themselves and merge with other entities. They can self-modify, interact with one another and with their environment. They can be filtered and sorted. They can move around and can be moved around. Meta information associated with each operation explains where, when and under what circumstances it is to be applied. The operations are usually applied repeatedly over an extended period of time.

Coevolution

In practice sometimes it is important to consider more than one population. Where multiple populations of entities interact, models of coevolution and symbiosis can be applied. These consist of descriptions of the additional populations, together with more operations describing how the entities from different populations interact.

Simplified models

Many models are simplified and focus on a small set of these operations, often ignoring other ones completely. For example, they may assume that interactions take place randomly, that the population is of infinite in size, or that the environment is fixed. Sometimes the simplifications assist mathematical analysis, other times they make computer simulation easier. Darwinism is not a particular theory, rather it is a toolbox which can be used to construct theories. One thing that is out of bounds is "arbitrary" mutations. Unconstrained mutation predicts everything and is useless. Mutations must be constrained to be naturalistically plausible. Other operations are subject to similar constraints.

Expanded domain

Traditionally, the populations involved consist of living organisms with genes made out of nucleic acid. The organisms involved are usually of the same species. Universal Darwinism expands the domain of evolutionary theory to include a wide range of other types of populations. Often these are conventionally considered to be part of the environment. For example, in culture, populations of cars, books, buildings, coins and words can be considered. In neuroscience, populations can consist of nerve impulses, or axon branch tips. In geology, populations can consist of rocks, streams, mountains or islands. In physics, populations can consist of atoms, molecules, photons or observers.

Often the same kinds of operations that are conventionally associated with Darwinism can be applied in these new domains. Operations involving damage and disintegration clearly apply to a very wide range of entities. Also, many entities persist and can be copied. Copying is sometimes though of as being unique to living organisms, but it isn't so: copying is ubiquitous in nature. The essence of copying is that information in one place winds up in more than one place - and that happens whenever rocks crumble, whenever streams part and whenever starlight hits dust. High fidelity copying is also common. For instance, position is often copied with high precision when entities split.

As a result of these shared operations, many of the parts of evolutionary theory can be applied to gain insights into these other realms. In particular, the concepts of fitness, adaptation and optimality which are traditionally associated with evolutionary biology turn out to apply widely to many types of system.This is Universal Darwinism in a nutshell.

Is the approach useful?

How useful is evolutionary theory in these other domains? It varies. In the case of cultural evolution, the answer is obviously, very useful indeed. In other cases, it often depends partly on how much overlap there is between the operations involved and those found in classical Darwinism. One common issue is limited heredity. With both DNA and culture, high fidelity inheritance systems allow open ended evolution and the accumulation of adaptations. However, with many simple systems, heredity is limited. A rock might inherit its position, momentum, chemical composition and temperature from its parent rock, but there's a limit on the number of times a rock can split into smaller rocks. Additionally, there are ceilings on the information carrying capacity of some entities. The tip of a lightning strike might carry information about charge and position - and encoding position might require a considerable number of bits - but there's still a limit on the amount of information so encoded. Such "information ceilings" represent additional limitations. These limitations don't mean that many standard Darwinian methods are not useful - but they should lead to constrained expectations regarding the scale of any resulting adaptations.

Relevance of population thinking

Another issue is the relevance of population thinking in the first place. With DNA-based organisms, a population of discrete similar entities is often the product of evolution - since the entities involved share a common ancestor. Often they flock together and interact with each other more than with other parts of their environment. In some of the other cases covered by Universal Darwinism, how to divide the world into entities is not always so clear and obvious. Even if the entities involved don't form a natural kind, it can still be useful to divide the world up into pieces and consider how they interact - that is simply standard scientific reductionism. However, the more the world divides neatly into pieces, the easier it will be to apply evolutionary theory to it. Note that even if you can't divide some aspect of the world up into pieces, you can still sometimes gain something by treat it as a single persisting entity - and then applying coevolution models.

Competing theories

A theory is only as good as the best of its competition. How have the phenomena covered by universal Darwinism been treated by scientists historically? We will consider two alternative theoretical frameworks:

  • One competing theory is from physics, or more specifically thermodynamics and statistical mechanics. This is the maximum entropy production principle. Instead of considering fitness maximization, maximum entropy production deals instead with entropy maximization. Maximum entropy production covers similar domain to universal Darwinism and successfully reproduces many of its predictions. However there are differences in emphasis and approach. For example, evolutionary theory has done a lot of useful work on path dependence and genetic drift. It has also made more study of when waste arising from conflict results in deviations from optimality.

  • Another competitor for universal Darwinism is Niche Construction Theory. Like universal Darwinism this deals with heredity transmitted via the environment and selection acting on environmental components. Niche Construction Theory is usually presented as a bunch of add-on components for Darwinism, though. Universal Darwinism produces very similar results by using standard Darwinism with no more axioms or complexity. Instead it treats the environment as a bunch of other Darwinian populations and reuses models of how different populations interact. As such, it would seem to be favored by Occam's razor. If Universal Darwinism is adopted first on these grounds, it is not obvious that there is any further explanatory work for Niche Construction Theory to do.

Deepening as well as broadening

Many biologists have long hoped that we will discover alien life - to illuminate which aspects of biology are fundamental and which are historical accidents. Universal Darwinism provides a very similar kind of illumination. In addition to the broadening of the domain of Darwinism, the new examples of evolution stimulate deepenings, revisions and generalizations of evolutionary theory. For example, cultural evolution stimulates the incorporation of intelligent design and directed mutations into evolutionary theory. I haven't covered these sorts of deepenings in this post - mainly through shortage of space - but several such deepenings are conceptually associated with Universal Darwinism.

Darwinism: what's in a name?

Some debate whether the resulting evolutionary theories should be called "Darwinian". Sometimes certain operations - such as entity copying - are required for the term to be applicable. Other times certain operations - such as inheritance of acquired traits - are forbidden. The term is used here mostly as a way of paying homage to Darwin's pioneering work on organic and cultural evolution. The term Darwinism is commonly used, though it has some associated baggage. In the end this is a terminology issue - the science involved is more significant.

How revolutionary?

Another issue is how revolutionary Universal Darwinism is. On one hand, most evolutionary textbooks need to be completely rewritten to cover the new domains involved. On the other hand Universal Darwinism expands the domain of Darwinism at very little cost in terms of additional axioms or complexity. It is often no more complex and traditional Darwinism. From this perspective it is no revolution at all. Having said that, expansions of the domain of Darwinism often represent revolutions - in the fields which are being invaded. For example, Darwinism represents significant revolutions in anthropology, economics and physics. In biology, the field was significantly transformed after adopting Darwinism. It seems reasonable to expect a similar large impact in other fields.

Sunday 5 June 2016

The evolution of natural selection

Modern definitions of "natural selection" include both differences in survival and differences in reproductive success. E.g. see here.

For Darwin, natural selection was all about the survival and preservation of individuals. Some quotes from Darwin should illustrate the point.

In The Origin of Species, Chapter 4, Darwin wrote:

If such do occur, can we doubt (remembering that many more individuals are born than can possibly survive) that individuals having any advantage, however slight, over others, would have the best chance of surviving and of procreating their kind? On the other hand, we may feel sure that any variation in the least degree injurious would be rigidly destroyed. This preservation of favourable variations and the rejection of injurious variations, I call Natural Selection.

In The Origin of Species, Chapter 3, Darwin wrote:

Owing to this struggle for life, any variation, however slight and from whatever cause proceeding, if it be in any degree profitable to an individual of any species, in its infinitely complex relations to other organic beings and to external nature, will tend to the preservation of that individual, and will generally be inherited by its offspring. The offspring, also, will thus have a better chance of surviving, for, of the many individuals of any species which are periodically born, but a small number can survive. I have called this principle, by which each slight variation, if useful, is preserved, by the term of Natural Selection, in order to mark its relation to man's power of selection.

In The Origin of Species, Chapter 4, Darwin wrote:

Natural selection acts solely through the preservation of variations in some way advantageous, which consequently endure.

Darwin repeatedly uses the term 'preservation' and repeatedly says he is talking about individuals (rather than traits). In The Descent of Man, Darwin wrote:

The survival or preservation of certain favoured words in the struggle for existence is natural selection.

In a letter to Charles Lyell dated 1860, Darwin regretted the use of the term "Natural Selection," preferring the term "Natural Preservation". He wrote:

Talking of “Natural Selection”, if I had to commence de novo, I would have used "natural preservation".

Of course, Darwin did also refer to differences in reproductive success. He called these "sexual selection". Darwin wrote:

this leads me to say a few words on what I call Sexual Selection. This depends, not on a struggle for existence, but on a struggle between the males for possession of the females; the result is not death to the unsuccessful competitor, but few or no offspring.

For Darwin, natural selection and sexual selection were separate. Most modern authors would say that natural selection includes sexual selection.

Darwin's distinction between "natural selection" and "sexual selection" corresponds broadly to my terms "natural elimination" and "natural production". Darwin's term "sexual selection" is totally unusable in this context - since both sexual and asexual reproduction can result in different numbers of offspring.

In the 1930s the term "natural selection" was hijacked by other evolutionary biologists and redefined away from what Darwin originally meant by it. Here's Satoshi Kanazawa (2003):

In the 1930s, however, biologists redefined natural selection to subsume sexual selection and began to contend that differential reproductive success was the currency of natural selection.
Here is Geoffrey Miller in The Mating Mind, 2000, p.8:

In the 1930s, biologists redefined natural selection to include sexual selection, because they did not think sexual selection was very important. Following their precedent, modern biology textbooks define natural selection to include every process that leads some genes to out-compete other genes by virtue of their survival or reproductive benefits. When one biologist says "evolution through natural selection," other biologists hear "evolution for survival or reproductive advantage." But non-biologists, including many other scientists, still hear "survival of the fittest." Many evolutionary psychologists, who should know better, even ask what possible "survival value" could explain some trait under discussion. This causes enormous confusion, and ensures that sexual selection continues to be neglected in discussions of human evolution.

I share the concern that this redefinition can be a source of confusion. Darwin is widely credited with the idea of natural selection. I wonder how many students realize that the natural selection they are taught is quite different from the natural selection that Darwin came up with.

Most of the sources I can see that mention this redefinition argue that its consequences were negative, and we should go back to Darwin's conceptual split between destruction and reproduction. That's essentially the split which I argued for in my essay natural production and natural elimination.

Satoshi Kanazawa (2003) says:

I concur with Miller (2000, pp. 8–12), Campbell (2002, pp. 34–35), and others in the current generation of evolutionary psychologists and believe that we should return to Darwin’s original definitions and treat natural and sexual selection as two distinct processes. I am fully aware that this view is still controversial and in the minority, but I firmly believe that the conceptual separation of natural and sexual selection will bring theoretical clarity to evolutionary biology and psychology.

Geoffrey Miller in The Mating Mind, 2000, p.8-9 says:

In this book I shall use the terms "natural selection" and "sexual selection" as Darwin did: natural selection arising through competition for survival, and sexual selection arising through competition for reproduction. I am perfectly aware that this is not the way professional biologists currently use these terms. But I think it is more important, especially for nonbiologist readers, to appreciate that selection for survival and selection for attracting sexual partners are distinct processes that tend to produce quite different kinds of biological traits. Terms should be the servants of theories, not the masters. By reviving Darwin's distinction between natural selection for survival and sexual selection for reproduction, we can talk more easily about their differences.

Friday 3 June 2016

Superstimuli

Superstimuli are stimuli that are larger than life. They are typically exaggerations of a particular natural stimulus. The stimulus involved could be attractive, aversive, or it could evoke other reactions - such as hunger, fear or jealousy.

Superstimulii are typically ways in which organisms manipulate other organisms. Extra-strong stimuli evoke powerful reactions - sometimes against their owner's will and best interests. This makes them useful for those who seek to manipulate others.

A picture is worth a thousand words, so without further ado, here are some superstimulii:

-
Superstimuli can arise in organic or cultural evolution. In the organic realm, flowers present superstimuli to insects - including genitalia mimicry, powerful scents and super-sweet nectar. Cultural evolution has led to super-sweet chocolate gateau, super satisfying donuts, super-smelly perfumes, and all manner of exaggeration of female ornaments. Pornography makes extensive use of sexual superstimuli.

Superstimuli are often used in marketing and advertising. The Daisy advertisement used a fear superstimulus relating to the end of the world for political ends. Environmental activists regularly invite people to help to save the planet. Highly attractive female models are used to sell many products.

Wednesday 1 June 2016

Beyond memetic algorithms

I used to be a genetic algorithms enthusiast - but more recently I have come around to the idea that they are based on an old and out-of-date type of evolution - one that doesn't learn from its mistakes very efficiently. In the real world, genes invented brains that then led to culture and cultural evolution - which is a faster and better kind of evolution, one which is better able to approximate our ideals of inductive inference.

In this case, it seems sensible to follow nature's approach - and use virtual genes to control the parameters of neural networks (or a similar learning component) which are able to transmit their learned findings to each other via culture, bypassing their original genome. This is what is done in memetic algorithms.

Will nature go beyond this approach, and if so, can we help it along? One of the next steps seems fairly obvious: use a neural network (or a similar learning component) to help along the evolution of the virtual genes that control its own development by using intelligent design. In other words, use memetic and genetic engineering. This step finally closes the loop between brains and genes: first genes made brains and then brains learned how to do engineering and then became able to apply engineering techniques to their own genes.

There should be several benefits associated with using engineering techniques and intelligent design. Agents using self-directed evolution can select mutations deliberately, based on whether tests on them are likely to increase the knowledge base or otherwise lead to success. This eliminates some of the cost associated with testing random variants. Evaluation can be performed partly under simulation - which should be faster and cheaper than testing everything in real life. Interpretation of the results also changes as a result of being able to store a history of past successes and failures which can be processed using interpolation and extrapolation techniques in the development of new variants.