Evolution and Culture
The notion of evolution is very general, and can be applied to many processes. Biological evolution is one form of evolution that has been examined in great detail. The title of this talk might even suggest that biological evolution is about to encounter a new development. This development would be the creation of a silicon artifact that would be susceptible to at least some of the central laws and constraints of biological life including evolutionary principles. Thus, one might begin to paint a picture in which artifacts are created that can reproduce and in reproduction exhibit some form of variation. One might then be tempted to speculate about the life of these artifacts and even examine their moral status. While all of this can produce good stories, and while it may create just those stories that allow us to focus on human characteristics more closely and clearly, we will not bother ourselves with these stories today. The story that we will examine is about cultural evolution.
Cultural evolution is similar to biological evolution. Biological and cultural evolutions are alike in that they involve changes, selection pressures, and populations containing diversity, and stable, recognizable populations even when there is variation in the individuals. Unlike biological evolution, cultural evolution contains intentional variations, selection pressures, and changes. Genetic variation is not an intentional response to the environment, but some cognitive, cultural variations are. Briefly, variation and selection are coupled in cultural evolution, but not in biological evolution. Further, selection pressures have an intentional component in cultural evolution. Humans create selection pressures to accomplish particular goals. In biological evolution this is not so. The evolving organism does not create environmental selection forces to promote the organism's or population's own goals.[1]
While there is a great deal that might be debated in this characterization of cultural evolution, I will assume that in general it is fairly accurate. The elements of culture vary, some of the variations are selected to continue, some of the variations are selected to be eliminated, to some degree humans attempt to intentionally create new variations, and to some degree they also create and modify selection conditions to achieve certain goals.[2]
In the remainder of this essay, I will apply these ideas to the introduction of silicon artifacts into the process of cultural evolution.
Silicon Artifacts
Silicon artifacts are computational devices. Currently, these computational devices are silicon based. However, there is no need to limit the discussion to only those devices that are silicon based.[3] I will use the phrase "silicon artifact" to signify any computational artifact.
What is of interest about silicon artifacts is that some of them seem to be behaving intelligently; they actually seem to perform tasks that otherwise could be performed only by intelligent humans. The extraction of information from a table, the sorting of information, and the performing of mathematical and logical operation on information would seem to be the kinds of things humans do and talk about doing. This latter notion is important. One might contend that organisms other than humans also have computational skills. However, what is distinctive in the human case is that we can talk about these computational skills and bring them under our intentional control. I can, if I so chose, create a bad argument. I can talk about that argument. I can use that argument in interaction with others. I can pass along the argument, the method of representation, and the criteria of appraisal to others. Of course, I can do so with good arguments as well. The cultural selection pressure on the advancement of novel arguments as good or bad arguments is as clear today as it was for Socrates.
The advancement and selection of arguments or, more generally ideas, is a prime factor in cultural evolution. The interactions of the human agents in this process are the primary focus of the story of cultural evolution. However, I now want to suggest that silicon artifacts (computational devices) have fundamentally altered the story. Silicon life is a form of life in which humans interact with computational devices at the intentional level. In this sense, silicon life is not about the simple fact that there are artifacts that compute, but that the computations, the communication of the computations, and the effects of the computations of these artifacts fundamentally alter the intentional life of humans who have been the primary focus of cultural evolution.
Arguments about Silicon Life
There are several arguments that can be advanced about the introduction of computational devices into our cultural evolution. I will focus on two of these. One concerns the notion that silicon life, in some sense, replaces a portion of human life; the other concerns silicon life as a servant.
The first argument sketch can be put in the following way.
1) Artifacts can be designed that can replace human effort.
2) Human effort should be replaced by the artifact's effort when there is a clear advantage for doing so.
3) Silicon artifacts have been created that can replace human cognitive effort.
4) There is an advantage to replacing human cognitive efforts.
5) Therefore, these silicon artifacts ought to replace human cognitive effort.[4]
There are two points to make about this argument sketch. The first is that the advantage might be the removal of a danger or the lessening of a burden, or it might be an increase in the level of activity that the human can sustain or the improvement of the human effort. In the case of silicon artifacts the emphasis is generally placed on the latter notions. The effect of replacement in the case of silicon life is to either increase productivity or increase computational and inferential ability. The second point to note is that the argument is intended to have a normative thrust that is consistent with the ideas of cultural evolution presented above. If one accepts the argument, then new silicon artifacts should be introduced that support the replacement of human cognitive effort.
The second argument sketch is a bit different.
1) Artifacts can be designed that assist humans in their various efforts.
2) Silicon artifacts can be designed that assist humans in their cognitive efforts.
3) Artifacts ought to be designed so as facilitate human endeavor.
4) Therefore, silicon artifacts ought to be designed that facilitate human endeavor.
This second argument sketch emphasizes silicon life as a servant of human endeavor. This is compatible with having to learn new ways of thinking and behaving, but places an obligation on the designers of silicon artifacts to create artifacts that facilitate human endeavor rather than replace human effort. The cultural selection pressures, therefore, are placed on the production of new artifacts that in some sense are servants of human intentions.
I will attempt to show that the first argument is misleading since human cognitive effort is not so much replaced as redirected and intensified, and that the second argument provides a reasonable and desirable path for placing selection pressures on the evolution of silicon life.
Silicon Life as Replacement
Consider U.S. industry's experience with computer-integrated manufacturing (CIM). CIM research has resulted in the development of a number of sophisticated automatically controlled manufacturing plants, up to and including "lights out" robot-populated factories.
These systems have been a disappointment. Despite the promises to the contrary, most so-called CIM facilities have been characterized by "islands of automation" that focus on discrete areas of production that cannot work in concert with other plant operations. Many of the facilities are inefficient and costly to run. Others have alarming safety records that can be traced almost entirely to poor systems design, although most of the accidents are attributed to "human error." These limits have been well publicized, and commercial manufactures have been reluctant to make the multimillion-dollar commitments these systems require.[5]
Silicon life begins to evolve in the workplace. Whether one is looking at "office automation" or manufacturing there is a significant interaction of humans and computational artifacts. Beyond the workplace, all of the other parts of our lives are beginning to share in silicon life. The production of newspapers, books, movies, records, entertainment devices, transportation vehicles, home appliances, and weaponry all bear the mark of silicon life. Kukla and colleagues present a rather unpleasant fact about silicon life, at least silicon life in the factory. It seems in some sense most unsatisfactory, and I believe its unsatisfactoriness can be located in the belief that silicon life will require less human effort and that the human effort expended will require less intelligence.[6]
In the mythology that surrounds the first argument about silicon life, not only physical labor, but also intelligence is transferred from the human to the machine. Not only are our artifacts stronger, but they are also quicker and smarter. Within the myth, people need to exert less physical and mental effort. Artifacts handle the problems. The myth incorporates the notion of a technological fix in a very strong form. Not only can there be a technological fix for human physical problems and limitations, but also for human cognitive problems and limitations. The ultimate work place has no lights. There is no need. The machines do the work. They think. They plan. They act.
The infusion of silicon life marks a profound change in the process of cultural evolution. Human intentionality in generating variations and selection pressures are to some degree transferred to computational artifacts. One might think of an industry in which there are genuinely no lights on. Computational artifacts monitor human desires and propensities to spend, and understand enough of human psychology to plan the development of new products. Other computational artifacts reason about the design of other artifacts. Yet other computational artifacts reason about the manufacturing of the new artifacts. Computational agents that understand business reason about the profitability of the production of the new artifact. Collectively these computational artifacts deliver instructions to yet another computational artifact that sets the ordering of material, the generation of parts, and the assembly of the parts in motion. The advertising computational artifacts map out the advertising campaign. The new artifact is advertised, produced, distributed, and, eventually, consumed.
But is it at all reasonable to think that silicon life can accomplish all of this? Clearly the tone of the comments by Kukla and colleagues suggests a healthy dose of skepticism. Silicon artifacts might not achieve the level of autonomy that is required by the myth.
Consider the many computational devices with which we interact. I am sitting at my computer, fingers at the ready. Pressing the keys that transmit my intentions to the computer, which in turn represents my key presses with familiar characters on the screen and very unfamiliar representations in its memory units. The intentionality is stripped from my ideas. The computer does not understand them. As a result I must retype and edit. Something that I did not formerly know I now must know. I must know how to edit. That is a new imposition on me. Further, I want my essay to read well and look nice. More impositions! I must learn more. But, perhaps I am deceived. Perhaps this is only a passing image of silicon life. The real image will be produced shortly. New computational mechanisms will be able to understand what I am writing. A new computational proofreader will appear. It can be set to a variety of styles and correct my awkward manipulation of my native language. But why stop there? If the computational artifact can understand my writing, then it can represent it and reason about it. Perhaps yet another computational agent will be generated that will critique and improve my essay. It will check the logic of my arguments as well as the quality of my references. When it is finished I will have a well-written, well-organized, well-argued essay that includes all of the relevant references indicating whether they are consistent or inconsistent with my thesis.
The difficulties in building the collection of computational artifacts that could actually do the tasks connected with my writing this essay are numerous and complex. For the artifact to actually write this essay without my intervention would seem to be an even more difficult task. However, this is not correct. The silicon artifacts that critique and rewrite my essay would need to be able to pass the Turing test of an intelligent computational artifact, and if that test could be passed, then the artifact could simply write the essay. Briefly, the Turing test is a guessing game that pits humans against both a computational artifact and a human. The computational artifact would pass the test and display its intelligence, if the humans performing the test could not with regularity determine whether the artifact or the human were responding to the testers' questions. The test is designed to show that I would have just as much reason to say that the artifact was intelligent as I would to say that the human was. Of course, when the artifact that passed the test generated its responses it would have engaged the processes that my very intelligent word processor would have needed to engage.
I want to suggest that the evolution of silicon life can be regarded as the evolution of silicon artifacts that replace human activity. The replacement need not be total and need not be perfect. The replacement of the human by the artifact may only occur in certain regions of human life, "islands of automation." The replacement artifact may not perform exactly like the human agent. In this sense the artifact does not need to pass the Turing test to replace the human. The cultural selection pressures can be such that if the computational artifact is "good enough," then that computational artifact can replace the human agent.
For the computational artifact to be "good enough" certain conditions must be satisfied. These conditions are as various as the parts of human life that are the targets of the replacement. In general, however, the replacement artifact must perform in a way that is close to the way in which the human agent would perform. For example, one might have a criterion that allows replacement if the computational artifact performs "just like" the human agent in eighty percent of the cases that it is designed to handle. The notion of being "just like" must also be specified. The criterion in this case may be that the artifact and the human produce the same result or behavior. Further criteria may specify the advantages of the replacement in terms of human goals. The replacement might increase profit, might distribute knowledge that would be otherwise difficult to distribute, might remove dangers from human life, or might reduce the amount of time needed to do a particular chore. In brief, the replacement argument can be filled with good intentions. Thus, the criteria of replacement (cultural selection pressures) would specify the conditions under which the artifact is "good enough" and the conditions under which some cultural goal is being advanced.
If the foregoing is accurate, then it might seem that replacement is a good thing. However, as noted at the beginning of this section, replacement does not seem to work. Perhaps the answer is now a bit clearer. The cases that are not covered in the "good enough" condition require even more skill and intelligence to handle than the one's that are covered. Further, since the "just like" condition allows for representations and reasoning that are unlike human reasoning, when problems do occur the human must now be able to understand both the human representations and reasoning and the artifact's representation and reasoning. In this way what is required by the replacement is not a work force of less skilled and less intelligent humans, but a work force that is more skilled and more intelligent about some things. Thus, the advantage of replacing human cognitive effort with the efforts of silicon artifacts does not appear quite so obvious. Extending only slight, the introduction of silicon artifacts would require not only a greater amount of cognitive effort, but a more complex effort as well. Replacement can increase the cognitive burden on individuals. Still worse, the increase of the cognitive burden may not be directed at the domain. It might reasonably be assumed that the majority of the cognitive burden will be placed on understanding what the silicon artifact does and how one should interact with it. If the desired outcome of replacement is that human cognitive effort be reduced, then replacement would seem to be an inappropriate means to that end.
Silicon Life as Servant
The difficulties with the evolution of silicon life under the replacement banner might be viewed as merely temporary, practical problems. These practical problems are meaningful within a certain framework of intentions and values. The core ideas of silicon life as replacement generate problems only in the sense that the goals of the replacement are not satisfied. The replacement artifacts are not quite good enough, and they don't behave just like us. But if they did meet these conditions more fully, then profit would be increased, knowledge would be distributed, dangers would be removed from human life, and the time needed to do a particular chore would be reduced. Researchers and designers, therefore, should continue to strive to meet the conditions of replacement more fully. In much the same way that my word-processor creates a temporary inconvenience and imposes on me new cognitive demands, so to is there a temporary inconvenience in the whole of the evolution of silicon life. If we work hard enough and long enough, we will be able to satisfy the conditions of replacement to ever-higher degrees and will be able to reap the rewards in ever-greater amounts.
Perhaps there is another problem. Perhaps it is not the silicon artifacts that need to be changed. Perhaps it is our goals and values that need to be changed. A technological fix to a human problem is a well-entrenched part of our cultural evolution. We have promoted innovations that do not require a rethinking of human activity, but remove the problem by a piece of technology. We have taken as given human goals and desires and attempted to place selection pressures on ourselves that lead to the satisfaction of those goals. Perhaps there is something of value in this notion and perhaps there is something that is misguided.
What is of value in this notion is that technology is a servant. We have evolved our technologies to serve our goals, needs, and desires. While it is certainly the case that some technological innovations have displaced the efforts and activities of humans, this has not been the sole aim. In many, if not all, cases technological innovations have created artifacts that do things that humans cannot do. Humans cannot apply great pressures to metals in an accurate and repeatable way. They cannot sew uniformly for hours on end. The industrial aspect of our cultural evolution has many such cases. The introduction of silicon artifacts presents a new and novel element, however. In all of the previous innovations humans were clearly the masters. They determined what was to be done and how it was to be done. Those with power determined where human effort would be applied and were technological effort may be applied. While there is a great deal of moral and political critique that can be applied to these human decisions, they were still clearly human decisions. Humans make decisions and the technological innovations are the products of those decisions. Humans have a certain power that the inanimate machines did not have. We can think and reason; they cannot.[7] The challenge put forward by the silicon artifact is that this very thinking and reasoning is something that is turned over to an artifact. Silicon artifacts might think and reason, and in many cases might do so better than we can. When the tasks are sufficiently specific, advance human purposes, and human cognitive effort is not directed toward understanding the artifact, then it would be reasonable to consider the silicon artifact to be a cognitive servant. Adopting this stance leads to a vastly different understanding of silicon life.[8]
If Servitude Is Better than Replacement, Then How Did Silicon Life Get Here?
I will assume for the moment that silicon life has largely been geared to replacement.[9] In one way or another the evolution of silicon life has been driven by the effort to devise silicon artifacts that can do what humans do, although more precisely one might say that they have been devised to do what humans ought to do. The replacement has taken many forms. One of the most interesting rests on the ideas that human weaknesses are to be overcome, and human weaknesses exists where humans do not do what they ought to do. Thus, replacement has attempted to achieve "correct" human performance.
At first glance this does not appear to be a bad thing. After all one should applaud efforts that attempt to improve human performance by bringing actual performance in line with the ideal. It is exactly at this point, however, that the whole issue of the evolution of silicon life rests. What is to count as "correct" performance and what is to count as the ideal are at the heart of attempts to decide which innovative silicon artifacts should be pursued and which social selection pressures should be brought to bear on them.
A central concern in replacement and servitude in silicon evolution is the role of reason in values. While this is surely a broad topic, I will narrow it to a specific area. Let us assume that doing what we ought to do means, in part, doing what is rational, and let us further suppose that rationality is impersonal. The first assumption makes rationality transcend individual persons and even the collection of all persons, while the second assumption places decision processes in rationality, and especially logicist rationality. The history of philosophy can be read as supporting the replacement thesis by supporting theses two assumptions.[10]
The first assumption rests primarily on epistemological considerations. One part of the story would begin with Aristotle's distinction between matter and form, and the idea that logic is the instrument for knowing. The notions of matter and form can be understood as the precursors of the notions of "domain specific" and "generic knowledge," and logic can be though of as the "blue-print" for an inference engine. This idea of logic as an engine for performing inferences can then be traced through the modern works of Frege, Russell, and Tarski. Another part of the story can be traced to the birth of modern philosophy, the Renaissance, and the scientific revolution. Galileo realized that there was a need to assert that our ideas about the world are fundamentally distinct from the world of appearance. Although this movement to separate the world as it is from the world of appearance can be traced to Plato, it is the scientific revolution that pushed it forward. This separation paved the way for the rejection of the Cartesian distinction between mind and body. This distinction may only be one of appearance, and in fact there may be a fundamental similarity between mind and body. Hobbes, Hume, and others began to identify this similarity in mathematical description and symbol manipulation. In some manner or another the anti-Cartesians rallied round the slogan "Cognition is computation." A third part of the story is the extension of formalist accounts of reasoning. Leibniz postulated the possibility of automating calculations. Euler pioneered the representation of concrete problems in abstract ways that were amenable to computation. Babbage, Boole, and Russell continued this line of thought. As these three lines of thought began to converge, Turing introduced his test of intelligence that relied on the possibility that knowledge and reasoning were logical and representable and that a device that appeared to perform in an intelligent way and incorporated the general principles of reasoning and representation was indeed intelligent.
The second assumption rests primarily on a set of movements within moral and political theory. The story to be told here is one of the evolution of decision-making and the growth of a computational worldview. Within this view, all value decisions are computations. They may be questions of logic and consistency as they were with the Kantians. They may be questions of maximizing happiness as they were with Mill and the utilitarians. They may be questions satisficing and equilibrium as they are with Rawls and other liberal theorists. Which question one focuses on is a rough indicator of the kind of moral principle one thinks is important. In addressing these questions one might add the political-economic theories of the free market, the rational consumer, and the rational decision maker, as well as the notions of preference and choice. What is of interest in all of these movements is the effort to make morality and normative decision making generally a computational activity. That normative activity might be one in which traditional logic, probabilistic mathematics, or game theory play the dominate role, and each of these can be turned into an abstract, impersonal, and logical computational operation. In an extreme form, it leads to the picture of a gigantic computable function into which all of the values and preferences of society can be put, and the "correct," "right," or "good" choice can be produced. In concert with the epistemological aspect, the slogan for morality and value might become "Valuation is computation."
Although this sketch is far too quick, I believe that it is in broad measure accurate. What I mean by this is that it is a story that could be convincing built from the writings of Western philosophers and intellectuals. This is enough to indicate how we arrived at a point where the replacement position "makes sense." Given this story, it would seem to follow that if a being could be rational, then it would not make any epistemological mistakes and if it were going to act correctly, it would only make decisions that were rational. Thus, the fact that we have epistemological and moral failings is due to our nature. Our rational nature is somehow compromised. Perhaps it is our lack of objectivity. Perhaps it is our lack of will in being consistent in our reasoning. Perhaps it is our inability to follow out long chains of reasoning. Perhaps it is our inability to do complex calculations. Whatever the case, the defect is a cognitive defect in us. Now, it was desirable to eliminate the defects of our physical nature. Artifacts were created to perform exactly, predictably, and without tiring. In the case of our silicon artifacts, they reason objectively, consistently, and continuously. Just as one set of artifacts replaced our physical limitations, so to ought silicon artifacts replace our cognitive limitations. The "lights out" factory is the ultimate elimination of our limitations both physical and cognitive.
Controlling the Evolution of Silicon Life
One important thing has been left out of the story of the evolution of silicon life is the fact that in large measure the evolution of this form of life is something under our control. There is nothing necessary in the evolutionary story. In particular, there is no step that is the necessary next step. Granted that there are some "steps" that will seem necessary in the accomplishment of certain goals, it will not follow that these steps will be taken or should be taken. Even more there is nothing that makes it necessary for our goals to remain fixed. A central feature of cultural evolution is that change is the norm. The introduction of artifacts is one part of the change; the introduction or alteration of selection pressures is the other. I want to suggest that it is the introduction or alteration of the selection pressures on the innovations of silicon life that provides the means for controlling the evolution of silicon life.
A central aspect of the introduction and alteration of selection pressures is the way in which the process rests on an understanding of human nature or human life. However, the way in which we understand human nature is not independent of the way in which we understand the development of artifacts.
Man's nature is indeed so malleable that it might be on the point of changing again. If the computer paradigm becomes so strong that people begin to think of themselves as digital devices on the model of work in artificial intelligence, then, since for the reasons we have been rehearsing, machines cannot be like human beings, human beings may become progressively like machines. During the past two thousand years the importance of objectivity; the belief that actions are governed by fixed values; the notion that skills can be formalized; and in general that one can have a theory of practical activity, have gradually exerted their influence in psychology and in social science. People have begun to think of themselves as objects able to fit into the inflexible calculations of disembodied machines; Machines for which the human form-of-life must be analyzed into meaningless facts, rather than a field of concern organized by sensory motor skills. Our risk is not the advent of superintelligent computers, but subintelligent human beings.[11]
This passage from Dreyfus raises many interesting issues. From the point of view of this presentation, the most important thing is the concern that the silicon life may not only become a model of human life, but might be mistaken as the way human life ought to be, or even worse it might be mistaken for the way human life is.
Returning to the illustration of my very intelligent word-processor and the Turing test of intelligence, it should be noticed that there was an important sort of ambiguity in the way that this material was presented. Collins provides a useful approach to the ambiguity. Noting that Turing's real claim was that by the turn of the century opinion would have changed so much about our computation artifacts that one would be able to say that machines think without expecting a chorus of objections, Collins notes that
There are at least four ways in which we might move toward such a state of affairs:
(i) Machines get better at mimicking us;
(ii) We become more charitable to machines;
(iii) We start to behave more like machines;
(iv) Our image of ourselves becomes more like our image of machines.[12]
These four ways reflect the sorts of concerns that have been raised in the context of the evolution of silicon life. In terms of the very intelligent word processor, I might (i) find that it mimics what I do very well; (ii) lower my standard of what counts as intelligent; (iii) begin to behave in a way that is compatible with it, (iv) begin to think that either it is doing things correctly and I am not, or that the view I have of what I am doing is mistaken and what I am really doing is what it is doing. What should be clear about these options is that they are free choices. We are able to construct variations of concepts and alter selection pressures in such a way that after the alteration any one of these notions will have come to pass. The consequences of these choices may be extremely disturbing, and perhaps dangerous. But, it still is the case that it will be our free cultural choice that will shape the conclusion. If this is so, then how ought our choice to be guided?
Enchantment and Fun
- Despite their remarkable technological power, formal methods are not sufficient to guarantee that software will be simple and will satisfy users. User-centered design will produce more satisfaction but is unlikely to reduce the complexity of software. Designs based in linguistic analyses of work are most likely to reduce complexity while satisfying users.
Peter Denning, "Beyond formalism" - The design work should be playful.
Pelle Ehn, "Scandinavian design: on participation and skill" - I see the computer as the enchanted technology. Better, it is the technology of enchantment.
Allen Newell, "Fairy tales"
While it is difficult to identify the kinds of features that free choices will have, there are some features that emerge from the considerations of work and play. If the replacement thesis is rejected and one holds that silicon evolution should produce servants, then some of the choices in the previous section are a bit easier. The last two options should be rejected. Thus, the first two options are the ones that should concern us. Should we attempt to create things that are better cognitive mimes or should we loosen our criteria?
In attempting to come to a decision about these we should keep several things in mind. First, we ought not to assume that the highest values that we seek to promote are captured in rigid formalisms. There is a lot more to life and the creation of silicon servants than that. As Denning notes, there are clear limitations to the formalists orientation, and we might be better to concentrate on the communicative activities that structure and give meaning to our world. Second, we ought to keep in mind that life should be enjoyable. As Ehn notes, design work should be fun! In silicon evolution this notion of fun can be lost. In particular, it can be lost when we are inclined to accept the notion of replacement.
But where then does that leave us? Which of the remaining options should we pursue? To my mind pursuing either is pursuing both. As Newell notes, there is enchantment here. The enchantment is that intelligent behaviors (conditional actions) can be "stuffed" into all manner of things. The intelligence that is stuffed in need not be exactly like ours. It need only be similar enough that these things obey our commands and carry out our requests. In this way the whole of the artificial world comes to be enchanted. But it is not just the enchantment that comes with the command. There is also the enchantment that comes with constructing the spells for the enchanted world and discovering new worlds and new adventures.
Silicon life can continue to evolve. We have a free choice about how it will evolve. That choice is an important one. If I had to vote today on the direction that silicon evolution should take, I would vote for the particular direction that produced silicon servants the allowed our social world to become more fun and more enchanted.
Notes
1 This view borrows heavily from S. Toulmin's analysis of evolution. See S. Toulmin, Human Understanding (Princeton, NJ: Princeton University Press, 1972).
2 Cultural evolution as described above allows for a mapping of some traditional normative concerns in moral and political philosophy onto the process of cultural evolution. The notion of intentional variation can be mapped to discussions of virtue, freedom, and rights. The notion of intentional selection pressures can be mapped to discussions of state power, punishment, and distributive justice. The notion of a goal can be mapped to the discussions of the good life or the ideal society. Thus, the notion of cultural evolution can be closely tied to other more traditional normative concerns.
3 New devices based on more exotic materials may be created. In fact, even those devices that today people popularly speak of as silicon devices are much more complex and rely on many other materials in their construction and operation.
4 I am calling these argument sketches rather than arguments since much more detail would need to be supplied to mark the argument clear. In a manner consistent with the spirit of this essay, they are consistent with the notion that arguments are like ideas that are subject to selection pressures.
5 "Designing Effective Systems: A Tool Approach," C. D. Kukla, E. A. Clemens, R. S. Morse, and D. Cash in Usability; Turning Technologies into Tools (P. S. Adler and T. A. Winograd, NY: Oxford University Press, 1992), p . 42. It should be noted that authors of this paper are represent a variety of areas. The authors include a technical writer, human factors consultants, and manufacturing engineer.
6 Several researchers have examined this point. P. Adler "New technologies, new skills." California Management Review, Fall 1986. L. Hirschhorn, Beyond Mechanization. Cambridge: MIT Press, 1984. R. Jaikumar, "Postindustrial manufacturing," Harvard Business Review, Nov.-Dec. 1986, pp. 69-76. A. Majchrzak, The Human Side of Factory Automation, San Francisco: Jossey-Bass, 1988. R. Walton and G. Susman, "People policies for new machines." Harvard Business Review, March-April 1987, pp. 98-106. S. Zuboff, In the Age the Smart Machine, NY: Basic Books, 1988.
7 A similar account might be advanced for animals in terms of their capture, training, breeding, reproduction, and consumption.
8 This approach is most apparent in "Scandinavian design." This approach is strongly influenced by the later works of Wittgenstein. The following quotations from Pelle Ehn give a feel for this movement. "Fundamentally, democracy at work or industrial democracy concerns freedom, another value-laden concept. It concerns freedom from the constraints imposed by the market economy and the power of capital. And it also concerns freedom to practically formulate and carry out particular projects that further democratize work." (p.97) Noting that the participation in design requires that the participation makes a difference for the participants, that implementation of results is likely, and that the process is fun, Ehn points out "The first two points concern the political side of participation in design. Users must have a guarantee that their design efforts will be taken seriously. The last point concerns the design process. No matter how much influence participation may give, it has to transcend the boredom of traditional design meetings to really make the design meaningful and full of involved action. The design work should be playful." (p. 129) Pelle Ehn, "Scandinavian design; on participation and skill," In Usability; Turning Technologies into Tools (P. S. Adler and T. A. Winograd, NY: Oxford University Press, 1992) pp. 96-132. See also P. Ehn, Work-Oriented Design of Computer Artifacts (Hillsdale, N.J.: Lawrence Earlbaum, 1989); R. L. Ackoff, Redesigning the Future (New York: John Wiley, 1974); C. Floyd, "Outline of a paradigm change in software engineering," in Computers and Democracy - A Scandinavian Challenge (G. Bjerknes et al, eds.) (Aldershot, UK: Avebury, 1987).
9 Consider the following example. It can be read as expressing the idea of a technological fix and the assumption that replacement will take place. "We know how computers work, but they are not intelligent. Humans are intelligent, but we do not know how to design better ones. While AI researchers are trying to make computers intelligent, cognitive scientists and molecular biologists are trying to figure out how human intelligence works. An intelligent computer could use computerized semiconductor manufacturing facilities to design more intelligent versions of itself. If we understood the human genetic code, we might be able to design smarter humans. It is any body's guess which will come first - knowing how to humans smarter or knowing how to make computers smart. Either event promises to make life very interesting. Computers have already changed our lives beyond recognition, and by-products of AI research will accelerate this process. Change will accelerate if genuinely intelligent computers are developed. Now is a very good time for a technology junkie to be alive." W. A. Taylor, What Every Engineer Should Know About AI (Cambridge, MA: The MIT Press, 1988).
10 It is, of interest, that the view that philosophy supports the two assumptions is prevalent in the scientific and technological literature. For example see: G. Luger and W. Stubblefield, Artificial Intelligence: Structures and Strategies for Complex Problem Solving (Redwood City, CA: Benjamin/Cummings, 1993). The first chapter is both interesting in its presentation of an overview and sensitive in its understanding of the issues. "The challenging field of AI reflects some of the oldest concerns of Western civilization in the light of the modern computational model. The notions of rationality, representation, and reason are now under scrutiny as perhaps never before, because we computer scientists demand to understand them algorithmically! At the same time, the political, economic, and ethical situation of our species forces us to confront our responsibility for the effects of our artifices. The interplay between applications and the more humanistic aspirations for much of AI continues to inspire hosts of rich, challenging questions. We hope you will glean from the following chapters both a familiarity with contemporary concepts and techniques of AI and an appreciation for the timelessness of the problems they address." (p. 24)
11 H. Dreyfus, What Computers Can't Do (New York, NY; Harper and Row), p. 280.
12 H. M. Collins, Artificial Experts: Social Knowledge and Intelligent Machines (Cambridge, Mass.: MIT Press, 1990), p. 222.