Functional Is Not Optimal

If you look at what a random animal is doing at a random time, you will usually find a complex behavior, with many correlated parts.  Similarly if you look at a random physical part of an animal, you will find a complex structure, with many correlated parts.  The best way to make sense of such structures is usually to look for the functions they might perform, i.e., the way that they might contribute to survival and reproduction, now or in the past, of the animal or its parasites.  Sure, some structures are "spandrel" side effects of other functional structures, but (at the aggregate level) most are not. 

This, however, does not at all imply that most such structures are anything close to optimal, giving the best possible outcomes.  Consider that when we transplant plants or animals to a new continent, they often handily out-compete existing species.  This could not happen if the existing species were close to optimal.  Often their structures worked in the past, but have not adapted enough to changing context. 

It is similar for most human organizations.  A good way to understand what a random organization is doing at a random time, or to understand a random part of its physical or communication structure, is to ask how such structures once functioned to help the organization or its parasites to survive or grow.  But this hardly means that every organization is optimal; many are clearly dinosaurs losing to new rivals. 

A lot of human behavior does not immediately make sense in terms of survival and reproduction.  We sing, dance, joke, worship, collect, chatter, decorate, party, dribble, travel, argue, blog, and much more.  With a little thought, we can come up with functional explanations, such as signaling theories, for most of this behavior.  Many people, however, reject such explanations, saying functional explanations would imply human behavior is optimal, which it is clearly not.  Thus, they argue, odd human behavior must instead be due to random mistakes, spandrels, or our changing environment. 

This "head in sand" attitude is so contrary to what we know about animals and organizations that I have to conclude that not only is it seriously wrong, but that humans seem built to self-deceive about the functions of their behavior.  While our behavior may be far from optimal, there is surely a detailed correspondence between our behavior and the functions they perform.  But we prefer to be innocently unaware of the signals our behavior functions to send and receive.

From a lunch conversation with Mike Makowsky, Bryan Caplan, and Alex Tabarrok.

GD Star Rating
loading...
Tagged as:
Trackback URL:
  • http://www.stat.columbia.edu/~gelman/blog/ Andrew Gelman

    From an interesting article by Witold Rybczynski, I learned that the engineer/architect Ove Arup agreed with you:

    “…The idea that the correct functional, the correct structural and the best possible aesthetic solutions are one and the same thing must, I am afraid, be abandoned together with the older philosophers’ dream about the harmony and ultimate identity of truth, goodness, justice and beauty.”

    For example, Orup wrote:

    “A wall like the one at Highpoint would have been cheaper to build with bricks, but [Lubetkin] claimed it was functional and economic. It wasn’t functional at all: it had to be “Modern.” Functionalism really became a farce. What is wrong with a sloping roof? They can’t afford to pay what it costs to make a flat roof really waterproof. Lubetkin didn’t care. He just cared for the picture in the architectural magazines.”

  • Eric

    Stephen Jay Gould took the extreme that because some spandrels exist, because every species of the past was eventually replaced by another, it is reasonable to make the null hypothesis that everything is a quirky accident, there is no ‘direction’ to evolution. Evolutionary biologists/sociologist/psychologists, on the other hand, try to explain patterns and such as being functional. I agree with the evolutionary crowd. Clearly there are exceptions to the rule, but as Hayek said about traditions, it is prudent to assume they had some value, as opposed to being arbitrary.

  • http://mcohm.livejournal.com chesh

    One reason for this misconception could be the phrase, commonly taught along with the ideas of evolution and natural selection, “survival of the fittest”. While it’s true that Organism A in Environment A may be the fittest organism to evolve in it’s niche, Organism B (occupying a similar niche) in Environment B (a similar environment to A, but with colder winters, hotter summers, and meaner predators) will likely prove itself much fitter when moved to Environment A, and Organism A will quickly become extinct. Species are not likely to evolve to be significantly fitter than their environment requires for survival. A more accurate phrase to teach would be “survival of the fit enough”.

  • http://www.iidb.org RBH

    Which is to say that evolution is a satisficing process, not an optimizing process, where the satisficing is against a multi-term fitness non-linear equation rather than a nice linear single-term equation.

  • http://profile.typekey.com/tschoegl/ Adrian Tschoegl

    That a transplant outcompetes the incumbent, or simply does very well, does not say that the existing species were not close to optimal. They may well have been, given the ecological niche, including predators, they occupied. The transplants typically have left their predators behind, giving them an advantage. Think of the rabbits and cane toads in Australia. Alfred W. Crosby made a major argument out of this in his “Ecological Imperialism: The Biological Expansion of Europe, 900-1900”, in which he argued that this was a major factor in the success of Europeans in colonizing the New World, Australia, and Oceania. Jared Diamond built on the idea in his “Guns, Germs, and Steel”, treating Europeans as transplants who brought with them diseases new to the territories they conquered, diseases that facilitated their conquests by decimating the indigenes.

  • TGGP

    How big a role do “spandrels” play in mainstream evolutionary theory today? It is claimed here that PubMed has only two papers using the term other than the one in which it is coined. I am not all that familiar with PubMed and do not know whether it should be expected to have most of the significant papers on evolutionary biology.

  • Urstoff

    Optimality needs defined boundaries: optimal for the immediate environment, optimal for the immediate environment plus extant terrestrial species, optimal for any terrestrial environment, optimal for any terrestrial and extra-terrestrial environment, etc. Within certain boundaries (such as immediate environment with metabolic trade-offs considered), the structures of most organisms are optimal (or near enough). If one incorporates history as a boundary for optimality, then by definition of natural selection the structures of an organism are optimal.

  • http://profile.typekey.com/sentience/ Eliezer Yudkowsky

    RBH, evolution certainly does not satisfice. There’s no point where differential replication says “Okay, we’re fit enough.” Anything marginally fitter will continue to outcompete. Evolution doesn’t maximize in the sense of finding global maxima, either. It’s a gradual improving pressure.

    I suspect that the huge confusion between “functional” versus “optimal” (or as I would put it, between a product of cumulative optimization pressure versus the global maximum) is the fault of the insufficiently-damned Stephen J. Gould, whose deliberately dishonest crackpot pseudoscience is still thought by his many innocent victims to be the leading edge of evolutionary biology.

  • http://profile.typekey.com/brucekbritton/ Bruce Britton

    The logic here seems to be this:

    1. ‘With a little thought we can come up with functional explanations for people’s behavior.’

    2. ‘People reject such explanations.’

    3. Conclusion: People must be built to self-deceive about their behavior — they think it is not functionally explainable when it actually is.

    By the same logic, and equally true premises:

    1. With a little thought we can come up with non-functional explanations for people’s behavior (random mistakes, spandrels, changing environment).

    2. People disagree with such explanations.

    3. Conclusion: People must be built to self-deceive about their behavior — they think it is functionally explainable when it actually isn’t.

    It seems that the fact that we can ‘with a little thought’ come up with a class of explanations for some phenomena, and that ‘people reject such explanations’ should not necessarily make us “…have to conclude…” that those people are deceiving themselves.

    What else would be needed to make us ‘have to conclude’ that people are deceiving themselves?

    Ok, now suppose we have that. Wouldn’t we need to have some additional information to ‘…have to conclude…’ that people’s tendency to deceive themselves is ‘built in?’

    And what would that be?

  • http://profile.typekey.com/robinhanson/ Robin Hanson

    Bruce, the force of the argument was not based on “we can come up with” but on how common useful functional explanations seem to be for animals and organizations.

  • Anonymous

    Robin, I don’t mean to be snarky, but you seem to be saying that if a class of explanations is common and useful to those who believe them, then that is a forceful arguement that people who reject them are deceiving themselves.

    Say a class of explanations is common among scientists in a particular field. Is that a forceful arguement that other scientists in that field who reject those explanations are deceiving themselves?

    But what I’m really interested in is self-deception, which I believe does exist in some situations. At a minimum, for self-deception you need a single person with two separate parts, where one part believes a proposition, and the other part disbelieves it, and neither part knows what the other part believes. This is like the familiar dual self situation which appears in self restraint and other forms of self control, except that in self deception the two selves don’t know about each other, while in self control they do know about each other.

  • http://www2.blogger.com/profile/15437653761628928730 Kevembuangga

    Just to tease the singularitarians.
    If “Functional Is Not Optimal” (with which I agree) it is likely that intelligence as we know it is not optimal but does this implies that “super-intelligence” is in need of technological enhancements and not just of a paradigm shift?

  • http://scienceblogs.com/tfk Josh Rosenau

    “RBH, evolution certainly does not satisfice. There’s no point where differential replication says ‘Okay, we’re fit enough.’ Anything marginally fitter will continue to outcompete. Evolution doesn’t maximize in the sense of finding global maxima, either. It’s a gradual improving pressure.”

    Put slightly differently, evolution moves a population to a point which is fitter than anything around it, but not necessarily to the fittest point. Which is to say, it satisfices.

  • TGGP

    Josh, what do you mean by “fitter than anything around it”? If what is “around it” has fitness, it must also be acted on by natural selection. Then neither can ever be “fitter than anything around it” provided everything “around it” does not go extinct with no new entrants. See the Red Queen hypothesis.

    In addition I would note that any mutation that appears and increases fitness will be expected to become more common, independently of any sort of “minimum standard”. Imagine a population of aesexual beings in an unlimited resource environment. Each reproduces itself above the replacement rate (it has two children before it dies). A mutation appears that causes 5 children to be had. This mutation will become more common even though the previous norm could be said to be “satisficing”.

  • http://scienceblogs.com/tfk Josh Rosenau

    I don’t think I meant “around it” in a geographic sense, though I confess that I’m no longer sure, and I think my claim holds either way, if you’ll grant reasonable assumptions about spatial autocorrelation of fitness in neighboring populations.

    If we think about fitness landscapes (rather than geographic ones), evolution will push a population towards a local maximum, as Eliezer Yudkowsky said. My point is that that evolution is slow. I was not claiming that the minimum standard is static, and I don’t think RBH was either. Because of the process you describe, the minimum standard will rise slowly (slowly because evolution is slow).

    There was a time when ecology (my field) was focused on finding equilibrium rules for communities based on how natural selection would drive populations in a community in an unchanging environment. This generated a lot of fascinating math, and a lot of conundrums. Those challenges have largely been resolved by pointing out that environments do change. The organisms that survive and reproduce over the long run need not be the very best ones at each moment in time. The genetic lineages that persist are the ones that fail to perish under the full range of ecological conditions that the community experiences.

    At any moment in time, the population is at a point which satisfices, a point at which the individuals that are not fit enough to survive and reproduce have little or no presence. The survivors were not (all) optimal, they are what was sufficient in that era. Predators, prey and other biological resources will adapt to that new population (Red Queen), but not instantly. The population will have to adapt to that new status quo, but it will not respond instantly, nor will its response be rapid. No doubt it has to keep running just to stay in one place, but that place will almost certainly not be optimal at any given point it time, and the population is not trying to stay in one place. It has to shift as the fitness landscape shifts.

    A mutation that boosts fecundity (as in your example) means more infants competing with one another and with parents. In good years, that isn’t a problem. In harsh years, it might mean that none of the offspring grow up and produce more than two offspring (depending on the details of the organism’s life history). There is some optimal number of offspring in any given population and any given environment. When scientists measure fecundity in British great tits, they do not find it matching their modeled optimum, and environmental variability is the answer I recall them giving. I don’t have the reference handy, but now I have to look it up and blog it.

    With regard to your question about “spandrels” above: I don’t see the word itself in the literature much, but the concept definitely comes up in conversation. I wasn’t around before the spandrels paper, but I’m told that people are less apt to insist that every feature of an organism must be adaptive than they had been. I couldn’t say how much of that results from the spandrels paper and how much reflects the greater appreciation for environmental stochasticity I described above, as well as the neutral (and nearly neutral) theory in population genetics. Those theories led to research which showed that a major fraction of genetic change is not due to selection, so the major advances in molecular biology all build on an understanding that adaptation is not the only way to explain biological data.

  • _Felix

    “humans seem built to self-deceive about the
    functions of their behavior.”

    That reminds me of what I added to Uncyclopedia’s section about evolutionary psychology:

    There is mounting evidence that evolutionary psychology may be an innate urge built-in to the brains of all human beings. It would have met the need, in our cave-dwelling ancestors, for an spurious position of authority to retreat to in the attempt to justify their most comfortable habits, thus imparting a genetic advantage of some kind yet to be made up.