| | | Outline |
The table of contents of The Computer Between My Ears:
Descriptions of the chapters:
The chapters in The Computer Between My Ears will examine the use of metaphors in specific scientific fields. In some of these, metaphors have been a problem for experts in that field. The "digital brain" dogma has in the past actually forestalled research in what now appear, in light of the failures of that dogma, to be promising avenues of cognitive science. In others, the problem lies in the popular understanding of that field. For example, sociobiology contained some tremendously appealing metaphors that have long outlasted the science that produced them.
The chapters that follow cover metaphor issues in each of several different scientific fields. The chapters are meant to build on each other. For example, concepts introduced in the Linguistics chapter will be relevant, though not crucial, to the Genetics chapter.
Despite many years of effort, the field of applied neuroscience is at least as notable for what it hasn't achieved as for what it has. Despite many theoretically appealing possibilities, there's no evidence, for example, that there is any kind of "coding" scheme, by which neurons could transmit "information" from one to another, in the way that computer scientists understand these terms. Nor is there much experimental reason to suspect that human thought has anything in common with the symbol manipulation computers do.
One thing that's interesting about this field is that several ideas that are currently being developed about cognition and artificial intelligence are in fact extensions of ideas that first surfaced many years ago, but were discarded or ignored at the time largely because they did not comport with the dominant metaphor of the digital brain. For example, Mark Tilden, a roboticist at Los Alamos National Laboratory, has been developing advanced electronic robot controllers that are not analyzable by any known digital method, and may not be readily simulated on a computer. These appear now to be closely related to ideas presented by William Ashby in a 1952 book, but which were largely ignored at the time. The 1950's also saw Gordon Pask construct a chemical automaton prototype that could learn in ways unforeseen by its creator. This too led nowhere, until its "rediscovery" in the 1990's. We still don't know the answer to the most important questions in this field, but there is ample reason to suspect that the dominant digital metaphor has interfered with the advance of understanding.
But the language we use to communicate with one another is shaped by our awareness of the world we live in. We move around in a real world, guided by our senses, and our experience. For example, the same word can be used to mean many different things--all covered by the same dictionary definition--according to our knowledge of how the world works. This is more than just the ambiguity between meanings, such as when a noun like "walk" can mean a path or an action. This is a case of multiple meanings within a single definition. The meaning of a "walk" depends on who it is that's walking: a baby, a parent, a dog, a recovering paraplegic.
Not just words, but the same sentence can be used in many ways: literally, metaphorically, ironically, or as a lie. Trying to formulate formal rules (interpretable by a computer, for example) for these uses is a vain application of the already flawed brain-is-a-computer comparison from Cognitive Science.
People are creative, and the language we use to communicate with one another is bent and shaped by that creativity. An entertaining set of examples can be seen by the variety of ways in which you can intimate mental deficiency in somebody else. The expressions "bats in his belfry" and having "a screw loose" are enshrined in idiomatic dictionaries. But you can read expressions like "he's a few bricks short of a load," "he's not wrapped too tight," and "his elevator doesn't go to the top floor," and still know what is meant.
Metaphors, the subject of The Computer Between My Ears provide a useful example. It may not be generally possible to create rules directing how a metaphorical sentence is to be understood. At least it has not been done yet, despite substantial effort from linguists and philosophers. There are many obstacles; but a significant one involves trying to establish the realm in which the comparison is valid. For example, when comparing neurons to wires, is it relevant that some neurons aren't long and skinny like wires? Does it matter that wire comes in many different gauges? The way in which we interpret these metaphors depends critically on shared knowledge of how our world works, not on the idealized structure of the language. The meanings of these metaphors will always be opaque to those who don't share our world.
In addition to difficulties decoding the "text," there are also substantial current questions about how exactly the reading of a gene is to be understood, even if the language is one we can read. The concept of a "gene" was invented long before anyone figured out that they were encoded in DNA. The idea that an individual gene--a conceptual invention of the scientists who first studied heredity--corresponds to a single location on a strand of DNA is an appealing one, but one for which there is no real evidence. It has the appeal of simplicity, but no force of data.
In fact, there is ample evidence that this is far from the case. During development, strands of DNA are often combined in different ways to create more elaborate proteins than are directly coded in the chromosomes. Sometimes the same few strands are combined and then rearranged to form more than one protein. It was always an inadequate explanation to say that your DNA contains all the proteins that make up your body. After all, where are the instructions to put them together? But it now seems clear that even this simple statement is not correct, and that there do exist proteins not directly encoded in your DNA.
A specific version of the progress metaphor is the idea that the evolution of animal species is analogous to the growth of individuals. ("Ontogeny recapitulates phylogeny.") The idea had a long history before it was formalized as the "recapitulationist" theory of Ernst Haeckel in the 19th Century. Discredited by the turn of the century, the idea lingers on in several forms, not least in the form of the conviction that humans are more "advanced" than whales or baboons. But there are more subtle forms as well. You hear an echo of this theory whenever you hear another, supposedly inferior, race referred to as "childlike."
Natural selection is often compared to a force, acting on "units". Scientists argue over whether the units of evolution are individuals, populations, or individual genes. This is what gives scientists fits when they try to find explanations for the possibility that certain bacteria (and maybe more) can control the rate at which they mutate. But the comparison to a force is just that. Natural selection is not a force at all; it is a vast collection of life stories that can sometimes be viewed in terms of forces, but sometimes not.
Genetic advances have brought some clarity and some confusion to the study of evolution. For example, we can analyze genotypes of different species, and trace them back to common ancestors. But sometimes what these new techniques show is evidence of more complex histories than had been previously thought. For example, we now know that "horizontal gene transfer," the exchange of genetic material between two members of different species, has been much more common than supposed. According to findings from the Human Genome Project, hundreds of your genes have been inherited from bacterial invaders in your ancestors' pasts. We also know that sometimes genetic change can be quite dramatic. Two species of Himalayan deer are nearly identical, though their chromosomes are very distinct. They may have become species through a random genetic event, rather than through slow evolution. Each of these findings presents challenges to the simple, classic, picture of evolution as the accumulation of small mutations in the march of progress.
However, trying to create classifications that can handle all individuals of all plant and animal species (not to mention trying to account for sub-species and races) presents some confusing conundra. The Linnean classification system contains an implicit comparison between species and ideal Platonic "types." But as usual, the world is hardly so tidy. By some definitions, a single individual can belong to two separate species. Other definitions create species where two members may have less in common with each other than with members of other species. One classic case is of a few species of seabirds whose range is at some distance from either pole. Travelling at a constant latitude, neighboring populations of these birds can interbreed, but the ones across the pole from each other cannot. These appear to be different species, but there is no good place to draw the line between them. Similar situations are known to exist in many varieties of insects.
Advances in genetic techniques provide some insight into taxonomy. Actually, as is typically the case, the advances are making some details clearer at the expense of the clarity of the whole picture. For example, there is currently no way to predict from genes alone whether two animals are from different species. In many cases, the genetic variation within a species is greater than the average variation between one species and a near neighbor.
In a constantly evolving world, where continua are the norm, platonic ideals are seldom seen, and the definition of "species" even for animals like beavers--or people--is not nearly as clear as it once seemed.
But there are only shaky reasons to believe that many of the uses to which we put our brains have existed long enough for evolution to have been a factor. We speak, and other animals do not, so there are physical differences, which presumably were subject to evolutionary pressures. But the assumption that we can explain all of speech by appeal to biological evolution is simply an analysis led by a hard-line devotion to natural selection as the only mechanism of evolution. Many forces conspired to evolve us, and progressively refined adaptation to circumstances is only one.
Another misuse of evolutionary metaphor occurs when writers talk about an organism's "design." Daniel Dennett, for example, writes of the problem of evolutionary biology as equivalent to "reverse engineering" the design of some plant or animal. But this approach constitutes a grave mistake. I can take apart a device or a computer program to figure out how it worked only because I can be certain that most of the components I find inside are there for a specific reason. For an organism, this is frequently not the case, since much of history is contingent, dependent on chance, and limited by the developmental pathways available. If you know about the existence of genetic drift, horizontal gene transfer, and the instability of chromosomes, trying to come up with an evolutionary explanation for every feature of an animal is a fool's errand. Men have nipples for no other reason than it was important to have them so that women could have them, too. Many of us carry the genes for proteins that were apparently given to us by bacteria our single-celled ancestors ate billions of years ago.
Another reason reverse engineering is a flawed analogy is that intuiting the purpose even of human artifacts is rarely so simple. The software world is filled with "legacy code", programs written by somebody who has moved on or forgotten how they work. Many of these are rife with subroutines never called and conditions never fulfilled. In his memoirs of life as an industrial chemist, Primo Levi wrote about a paint recipe he invented that lived on decades after an ingredient was rendered superfluous. Trying to infer the purpose of these useless remnants can be very trying.
The classic evolutionary psychology analysis typically conflates "heritable" with "inevitable." Organisms are the result of complicated growth of an individual in its environment, each working on and changing the other. A behavior can have an important genetic component, but still not be inevitable. Breastfeeding infants provide a good example of how instinct and biological imperatives are often overrated. "Instinct," we are told, will help a baby latch onto their mother's breast. But many young parents have learned that this is poppycock, at best. If my daughter had been left to rely on instinct alone, she would have starved to death. She resisted feeding (from breasts and bottles), and had to be physically pushed onto the breast when she made the mistake of opening her mouth to yell. After that--and seeming surprised every time--she'd feed happily, but the process was excruciating for everyone concerned. It seems that a lot of the breastfeeding "instinct" is in fact cultural knowledge passed down from mothers to daughters. The La Leche League, a volunteer support organization, was founded to deal with the collapse of cultural knowledge about breastfeeding precipitated by "modern medicine" and infant formula purveyors. Without their help, instinct would often be useless.
The sociobiology project lives on in many ways, for example, in the ways authors seek biological justification for modern social mores, and in the related discipline of evolutionary psychology. A host of comparisons are regularly made between human society and groups of baboons, chimps, gorillas, or prairie dogs, depending on the sociologist in question. We are told that studies of social groupings of other animals will provide insight into our own. Like the rest of the metaphors covered in this book, there is some truth to be found in these comparisons, but seldom as much as is extracted. Sarah Hrdy, the author of a recent popular sociobiology book makes this claim:
The comparison is made here between the ways stone-age women strove for status in their tribes and the ways modern women would achieve status. Perhaps it is a telling analogy, but is there any justification for saying that we are biologically equipped to make the comparison unconsciously? That is, getting a good job must feel different than finding a trove of yams, though presumably both will make the finder happy and grant increased "status." But why assume that biology--the action of our genes--is more important than the reasoning faculties a woman could bring to bear on the situation? Is this evolution at work, or common sense? Committed sociobiologists would say there is no difference, but with what evidence? Our meager understanding of how brains really work make this conjecture little more than a guess based on the application of a congenial worldview.
Quite a lot has been written about the "balance of nature," where the image is of nature as an unspoiled Arcadia, with all forces equally opposed, so that a state of gentle equilibrium is maintained. This image is usually opposed to the image of Arcadia, invaded by people.
But nature, however fascinating, is almost never in equilibrium. Population biologists long ago discovered that predator-prey populations can exhibit behavior as chaotic as any weather system, with constant booms and busts as predators swing between feast and famine. Even the standard story of forest progression, with scrub yielding to pine forests which in turn yield to hardwoods, turns out to mask a tremendous struggle for dominance between species, without an equilibrium in sight.
Where this becomes important is that we are now in control of much of what used to be wild. In many of the important ecosystems in the world, we have fenced in the wild land, and we manage it for preservation's sake. But our image of what untrammeled nature would be like is often quite at odds with what was the case, and this leads to tragic mismanagement, ranging from unnecessary fires to the inadvertent extinction of protected species, and terrible conflicts between the "contained" wild, and people who live near it, or who advocate for it.
One common comparison is about economic growth, comparing growth of an economy to the growth of animals or people: Russia, or the nations of central Africa, don't have a "mature" economy, for example. The metaphor implies an intrinsic fault, correctable with time and perhaps the care of a loving parent. But though there are useful points of comparison, the growth of nations is only tenuously comparable to the growth of people. For example, my growth did not interfere with my sister's growth; when I was little, my family had enough food for both of us to grow up on. But the growth of nations is not independent. England became the first industrial nation in the 19th century. It took France much longer to industrialize, partly because of the rural nature of that country, but partly because England had already done it, and that made it more challenging for France to compete in areas already dominated by the English. The same dynamic prevails today. If Ghana, say, wants to industrialize, what can they make that isn't made better or cheaper somewhere else already? Cars? Clothes? Electronics?
Another invidious comparison involves the analogy between the federal debt and the debt of a corporation or an individual. Staying out of debt is considered good for individuals, therefore it must be good for the country as well. But the bank that holds my mortgage only grudgingly believes in my promise to repay, whereas our nation's debt--and the universal belief in its ability to repay it--is a pillar of the international economy, supporting any number of smaller governments, institutions, and individuals. Repaying my debt will have repercussions to no one but myself, but rapidly reducing the supply of US bonds in the world's economy will have serious and unpredictable consequences.
Of further interest in this chapter is the phenomenon that, as is probably inevitable in any endeavor where money is at stake, the choice of dominant metaphor seems often oddly coincident with the interests of those in power. Using the example above, if industrialization is seen as a purely intrinsic process, with the important variables being domestic ones (such as the "growth of civil society", or the "maturation of financial markets"), the interference of foreign aid is understood to be not necessarily benign. To put it crudely, this metaphor implies that if Ghana wants to industrialize, it's largely their problem, and the best thing the United States can do is to stay out of the way.
The final chapters will focus not on individual scientific disciplines, but on more general issues.
The choice of domain is a risk for any metaphor, but when you're comparing one real-world process to another, the risk is often masked. This is not so when considering numeric models, which have quite strict boundaries. For example, consider a computer model of a brain, and imagine that computer sitting next to some experimental subject (the person) it's modelling. Now imagine a little molecule of some psychedelic drug travelling up the person's aorta. How will the computer model keep track of that? One can argue that the modeled domain is not appropriate, and that the model should also include the blood chemistry and digestive system. Now our human takes a bite of the ham sandwich that was on his plate, causing who knows what havoc in his digestive and neural systems. So perhaps the model should include the food about to be eaten as well. But what about the food in the refrigerator? The brain of the person preparing the food? And what about the grizzly bear about to crash through the wall and chew off our subject's leg? What will the adrenaline rush from that do to his brain?
To make a model work, you have to choose boundaries between what's modeled and what's not, but these are inevitably arbitrary. Weather models have this exact problem. You can try to account for all the sources of energy in the atmosphere, but it's impossible to get them all. Just when you think you have, you realize that there is a cyclonic effect of driving on the right side of the road, or notice that the heat from millions of exhaust pipes and chimneys may not be negligible.
Another great risk of computer modelling is the discrete and digital nature of computers, compared to the continua of the natural world. Despite their underlying quantum nature, most natural phenomena are observable only as parts of a continuum. For some processes this is less of a problem than for others. Planetary motion presents problems readily solved by computer models. The errors in these kind of calculations can apparently be made arbitrarily small, by increasing the precision of the modelling. But in non-linear, possibly chaotic, systems small errors in precision can lead to huge errors in the solution. In systems like these, it is not possible for a discrete model even to approach perfect accuracy. This is a well-known result in meteorological models, but is sometimes ignored in other domains, where the system may be, but is not is not known to be, non-linear.
Briefly, the theory here borrows from George Lakoff the observation that comparisons are a fundamental way that we understand our world. We make comparisons to categorize new objects, but also to characterize new situations and events. This is automatic, and an integral part of what it means to be human. The construction of metaphors in human language follows naturally from this facility. The viewpoint in The Computer Between My Ears also takes from the work of Eva Kittay the conviction that there is an irreducible meaning in metaphor, that is not available to simple analysis, but is only understood by people who share enough knowledge of their world to make the comparisons they create mutually comprehensible.
| | | Outline |