The Psychotic State Number 4 November 12, 1998 BILL GATES AND THE STRUCTURE OF SCIENTIFIC REVOLUTIONS I have recently been in a more psychotic state than usual, and so you are going to have to bear with me a bit this month: this is a more eclectic essay than usual. What has been making me nuts is, as you know, the endless flamewars we've been having about the relative merits of Windows NT vs. Unix. With the time I save not participating in those discussions I have been pursuing an occasional avocation of mine reading books about the history and philosophy of science. Recently I read a rather famous and important book in that field, "The Structure of Scientific Revolutions", written in 1961 by Thomas S. Kuhn. If you'll read along for a while, I'll describe some of what it is he says. I promise, at the end of the essay, it will have something to do with what we all do for a living. Kuhn's great contribution to the history of science was to point out that scientific progress is not "teleological" (a greek word meaning "directed toward an end"). The traditional view of the history of science - if not of all of history - is that the present is the "goal" of the past. Yesterday is just a step on the road leading inevitably to today. In the context of history of science, this means that the science of past eras is regarded as essentially the same as the science of the present. As science progresses, great scientists uncover new physical laws, which make the understanding of the world more and more complete and accurate. Today, we still have some unanswered questions, which (one hopes) will be elucidated with further experiments and theories. Kuhn rejects this whole picture. He points out that at almost any moment in time, scientists have remarkably complete metaphysics, experimental techniques, and mental images that are used to describe the world. This scientific framework he calls a "paradigm". What has happened over the history of science is not simply the filling in details in the modern understanding of the world. The entire framework has been replaced periodically. A paradigm is not precisely a scientific theory - it is more like the set of assumptions and practices on which the theory is based, that lie "beneath" the theory. The paradigm is the thing that is not explicitly discussed nor challenged under usual circumstances. It is what holds the scientific community together. A paradigm is not supposed to be a complete understanding of the world. Having a paradigm does not solve all the problems, it simply is the "playing field" on which the solutions are attempted. There are plenty of problems that a paradigm leaves for the scientist to solve, and this is what science is doing most of the time. The vast, vast majority of scientific achievements, important though they may be, do not result in anyone changing an underlying assumption. So, for example, when a chemist produces a new reaction, or a physicist explains an event in a collider, it only changes the state of knowledge incrementally. This sort of thing Kuhn terms "normal science". Every now and again some phenomenon is observed that cannot be fitted into the current paradigm. Most of the time the scientists ignore the apparent problem, and usually eventually someone figures out a way to resolve it without any drastic reshaping of the way science is done. Not always. Sometimes the problem sits around for quite a while with no one paying too much attention to it; sometimes it gradually precipitates a "crisis", in which some people begin to question more and more of what they took for granted. The crisis is sometimes resolved through the creation of a new paradigm. This "paradigm shift" is not a wholely rational process, and scientific "evidence" may or may not succeed in persuading people to make the leap. Not all members of the scientific community participate in it. Sometimes there are people who never leave the old paradigm and in essence, according to Kuhn, they leave the community and cease to be scientists. Let me give you an example of a paradigm shift. In the physics of Aristotle, all bodies had a natural position in the world. Bodies always seek their natural location. When a body fell, it was going downwards, towards that point. As Kuhn says: "Contemplating a falling stone, Aristotle saw a change of state rather than a process... [B]ecause the stone was impelled by its nature to reach its final resting point, Aristotle saw the relevant distance parameter during the motion as the distance _to_ the final end point, rather than as that _from_ the origin of motion." (p 124) Now consider a pendulum. Certainly, for as long as there have been ropes, people have tied weights on them and seen them swing around. When Aristotle saw such an arrangement, he would fit it into his conceptual framework. To him it was some sort of very complex constrained fall. The weight would try to reach its natural point, below, and be constrained by the rope. It would swing around and eventually come to rest at the lowest point allowed by the rope. Galileo found through his experiments that there was certain regularities to the behavior of falling bodies, ones which could not be contemplated in Aristotle's terms. His analysis changed the view of a falling body from a change of state to a _dynamical process_, not just a simple transition from higher to lower. Galileo focussed on what happened _as_ the body fell. Galileo considered the relevant parameters as the position and velocity of the body, not simply the starting and ending points. Note that what Galileo did was to _change the conceptual terms in which the problem is understood_, not simply to figure out a way to discuss this motion in the then-current understanding. Without going into it too much, let me say that Galileo's paradigm was not the same as Newton's. Newton saw the pendulum problem as one of forces acting on the body - a force of gravity plus a force from the rope - which causes the body to accelerate in various ways. The force of gravity is caused by an innate property of every material thing, which is that matter attracts other matter. This attractive property is the same as the one that holds the earth in orbit around the sun. (This analysis is the one you would learn in an introductory physics course in high school or college.) Suffice it to say that Einstein regards the problem differently yet - for him the problem is motion in a curved space-time system. In each case the world is the same -- a weight on a rope -- but the language and the framework in which the data are analyzed are significantly different. The interesting thing is that everyone always has to operate inside a paradigm. Without some conceptual framework, the world is just too difficult to understand. We need something that allows us to distinguish what is significant from what is not, in order to sort out the forest of data we are presented with, and in order to allow us to look for the correct data to increase our understanding. Having said all this - and I don't have the space here to elaborate on it - we must note that the power of the paradigm can also be blinding. As Kuhn would have it, the paradigm you use actually changes the world in which you live. So, for example, the Aristotelian would see no regularity to the pendulum, because he wouldn't look for it. Galileo, on the other hand, seems to have _measured_ that the period of the pendulum - the time it takes to go back and forth - is independent of the amplitude with which it swings. This, it turns out, is not correct; it's only roughly true if the swings are small enough. Thus, the Aristotelian paradigm prevents the beholder from seeing a regularity of the world, but it forced Galileo to see too much. Here's an even better example: The ancient Greeks believed that the cosmos was a sphere of stars surrounding the earth, with the sun and the planets being perfect, unchanging spheres. It was Copernicus who proposed that the earth and the planets went around the sun. "Can it conceivably be an accident, for example, that Western astronomers first saw change in the previously immutable heavens during the half-century after Copernicus' new paradigm was first proposed? The Chinese, whose cosmological beliefs did not preclude celestial change, had recorded the appearance of many new stars in the heavens at a much earlier date. Also, even without the aid of a telescope, the Chinese had systematically recorded the appearance of sunspots centuries before these were seen by Galileo and his contemporaries. Nor were sunspots and a new star the only examples of celestial change to emerge in the heavens of Western astronomy immediately after Copernicus... The very ease and rapidity with which astronomers saw new things when looking at the old objects with old instruments may make us wish to say that, after Copernicus, astronomers lived in a different world." (p116-7) What Kuhn points out is that during a period when science is in crisis and a paradigm is breaking down, the exponents of the old and new theories tend to talk "through" each other. Frequently, that's because they are often using the same words to talk about the same problems, but the words have utterly different meanings. "Consider ... the men who called Copernicus mad because he proclaimed that the earth moved. They were not either just wrong or quite wrong. Part of what they meant by 'earth' was fixed position. Their earth, at least, could not be moved." (p149) For an individual scientist, making the paradigm shift means learning to see the old problem in a new way, to use the old words in a different conceptual framework. I guess it's time for my punchline. Something occurred to me because I was reading Kuhn and the NT/Unix flamewars at the same time. If anyone is talking "through" each other, it is the people participating in this debate. And it seems to me that there is a good reason for that. I think that in some ways, the OS is a paradigm in Kuhn's sense. It's not just a bunch of code that keeps us from having to write disk drivers every time we want to read a file. For one thing, our use and knowledge of the OS includes us into or out of a whole community of users and developers. More than that, it actually defines the way we look at the machine, and our expectations for it. It permits or discourages certain abstractions about what the computer can do - even what it is. So, for example, a Unix expert confronted with a machine that cannot talk to the network might explain it a wrongly configured routing table, whereas the Mac user simply sees a down box. When something you mostly understand fails, the problem is easily put into a framework, and that framework somehow lessens the emotional impact of the failure. Or, to quote an example we discussed recently in a tech meeting, the NT user might perceive the machine's monitor and the GUI interface thereon to be inherently part of the operating system of the CPU, whereas the user of X windows sees the potentialities of the arrangement completely differently. In such a circumstance, the usual sorts of discussions/flamewars are inevitably fruitless. The aesthetic and conceptual framework in which the statements are embedded are not explicit, and mostly what everyone is saying is "I'm a member of the Unix (or NT) community" over and over. I don't mean to overstate this, and I really don't mean that the power of the OS as a paradigm has as powerful a hold - in either the positive or negative sense - on our thoughts as the paradigms Kuhn was talking about. But I do think that it goes some way towards explaining why the flamewars never persuade anyone, and why they are ultimately so boring. Read Kuhn instead.