Joe's Jottings
Jottings Number 68, Reply D, by Benjamin Lloyd:
Date: Wed, 22 Jan 97 17:11:09 -0800
Your jotting brought to mind the work of several scientists who deal with simplicity, complexity, and their applicability to life. I apologize for the length and obtuse nature of this discussion: I just got on a roll, and couldn't stop (I need an editor). The first is Murray Gell-Mann, who won the Nobel prize in physics 25 or 30 years ago, and helped to start the Santa Fe Institute in the early 80's. He won the prize and created his reputation for theorizing the existence of quarks as the fundamental building blocks of matter. As such, his efforts were reductionist to an extreme. In the early-to-mid 70's, however, he expressed the opinion that the study of complexity represents the future of physics: we will (he believes) eventually develop the small set of mathematical principles that describe the "laws" of physical behavior in our universe (though they may not apply in any other universe). However, those laws will not be sufficient to describe all events in the universe (kind of a Godelian result). In fact, they will be unable to approximate any of the most interesting events. Gell-Mann feels, therefore, that there is need to study both the simple (exemplified by the physical laws, such as general relativity, quantum gravity, etc.) and the complex. He calls this study "plectics," a term with a rather forced etymology, but one, given the success of his other invented term, the "quark," we might find ourselves using in the future. A basic idea of plectics (as far as I understand it) is that it is not sufficient to try to explain all significant physical effects (including biological, psychological, technological) starting only from a simple set of laws/principles (from the ground up, if you will). One reason for this is that as soon as some randomness is added to the system (which exists even at the most basic quantum level), the result is unpredictable (complex). At this point, it becomes necessary to apply chaos or complexity theory. As an aside, the point of balance between stability predicted by basic laws and complete chaos is an example of homeostasis in complex systems. Several prominent thinkers in a number of fields, not limited to the sciences, believe that optimal "progress" is achieved at that homeostatic point. In fact, the theoretical biologist Stuart Kauffman postulates that biological (and human) evolution (in a neo-Darwinian sense) works best when we define "fittest" (as in survival of the) as those organisms that thrive at that point of homeostasis. In the business world, David Whyte, in his book "The Heart Aroused," suggests that success in business is achieved by organizations and individuals walking the knife edge between stability and chaos. With these views in mind, it's easy to see how we might arrive at the unspoken expectation we have of an engineer: walk the path between stability and chaos, and create views of parts of the chaos that convert them to stability. In other words, humanity progresses (evolves) by subduing the wild (another image popular over the last thousand years, and particularly close to home here in the American West). In so doing, however, we are (literally, in some cases) playing with fire. Our attempt to control fire, to box it to make it do our bidding, is one of the oldest examples both of humanity's conceit and of our desire and willingness to live on the edge. The risk of explosion, conflagration, etc. is apparently one that we are willing to take in order to reap the benefits of controlling fire. In your note, the Amish, however, recognize that the risks of adopting or using new technology are not limited to the physical, but must include the societal. To apply (overly-broadly, no doubt) the balancing point to the Amish situation, they too work on that boundary between stability and complexity, but understand that it is almost impossible to transform the complex into the simple ("Simple Gifts," though not Amish, is literally a hymn or anthem to this belief; the Copeland setting, with very simple harmonic structure, exemplifies the beauty of that simplicity, as opposed, perhaps, to more aleatoric music of the last 30 years). The Amish, then, exert control over the effects of complexity by attempting to limit its introduction into their society. In business, we attempt to "manage risk" in the ways Phillip Capper describes, developing defensive postures (in the same way the Amish do) that minimize the impact of risk on progress, but we can not eliminate it, nor, as one might infer from the previous discussion, should we try to: too much progress is achieved BECAUSE of failures, not in spite of them. It has been popular over tha last thirty years to use the computer to model events and systems. Neural networks and other such systems comprised of a large number of very simple components are used to simulate "life." What we find when we perform these experiments is that we can not, regardless of how many rules or laws we add to the enivironment, emulate life, but we can produce some results that provide us with knowledge about how things may work. At the same time, the idea that a butterfly flapping its wings in Guatemala affects the weather in Chicago can cause us to withdraw completely from the process, throwing up our hands at the immensity of the task of understanding such a system. Gell-Mann says we must approach it from both directions, and in fact, to take horizontal slices (at certain levels of complexity/simplicity) to shed light on the problem. He uses the example of human behavior and thought: the reductionist (simple) view is that we can describe the action of quarks, which determine the action of atomic components, which determine the actions of molecular components, etc. on up through body parts, etc. which finally determine behavior. Attempting the complete journey, he would argue, is impossible, and thus, non-productive. Rather, we have different specialists who focus at different levels: partical physicists, molecular biologists, bio-chemists, neurologists, psychologists, behavioral psychologists... Each of these specialties tries to advance in both directions: towards the simplicity of suspected cause, and towards the complexity of presumed effect. Now, in relation to information systems, however they are implemented: through traditional development, implementing packages, or packaging components, object or otherwise, they all fail eventually. They fail for the same reason that species become extinct, societies fall, or individuals die: they encounter a set of events, usually unforeseen, that they were not programmed or designed to handle. It is important to note that those events almost always come from outside the "system," and are introduced as a side effect from the system's existence in the context of a larger environment. Incorporating or embracing the immediate environment as part of the system (a la Gaia), factoring external effects on and by the environment helps, if merely by reducing the risk of unforeseen events. But it can not solve the problem for two reasons: 1. This expansions increases the complexity of the system, therefore increasing the opportunity for unforeseen internal failure. 2. We can never define a system that is large enough such that there is nothing outside of it (again, a la Godel). An example of a failure is the dinosaur. As a complex-adaptive system, the dinosaur survived for millions of years (much longer than humans have been around), but an event from outside the system to which it had adapted caused its extinction. As humans, we are trying really hard not to drive ourselves to extinction, but we have little control outside of this world. Information systems can be designed to be adaptive to a certain extent, responding to external events by changing themselves, but that only works at a reductionist level: for those events we can anticipate. As soon as we introduce the level of complexity required to create a complex-adaptive system, we no longer have control over the system, and the results it provides us will incorporate a certain degree of randomness. If we can say anything about our current use or expectations of computers and information systems, it is that we can not tolerate randomness. Consider, on the other hand, a corporation such as HP: a good example of a complex-adaptive system. There is no one person who understands everything about how the company works, even if we leave out how each person works. The corporation is successful even though this is true. Even though we can not exactly predict either internally how a particular project will work, or how an individual will perform his/her job, or externally, how the market will respond to our product introductions, everything averages out to produce a profit. To survive longer, our information systems and our expectations for them need to change: maybe we need to be satisfied if the information system produces a good result some high percent of the time. As long as the results are "competitive" or produce a positive "return" overall, why should we care what the individual results are? This partly derives from the "simple" laws that enable computing: in the end, everything comes down to a binary yes or no. In that world, 1 + 1 always equals 2, and 9/3 = 3. In the real world, the non-digital, non-integral world, nothing is black or white. When we humans have to make a black or white decision from this complexity, we must guess, over-simplify, and apply our own (complex) logic. Quantum computing works this way: since we can not predict exactly the behavior of the quantum components (Heisenberg), but we have "trained" the components to act a certain way on average, we send a computation to thousands or millions of components in parallel, survey them, and take the preponderance of results (0, 1, or maybe) as the "correct" answer. It's sort of like democratic computing. I can see that this uncertainty would raise havoc in most of our current systems, e.g. Financial Reporting ("our net profit for the quarter was approximately $500M"), but insistence on exactness dooms our information systems to obselescence (and keeps IS professionals employed). Extrapolating from this (a big leap), those information systems that solve problems which, even in the real world, can be deduced directly from a simple set of laws, have been successful: accounting, mathematical modeling, etc. automate a clearly-understood and law-driven process. The calculator is an extreme example of this: think how successful the calculator has been (as an application and as a technology). On the other end of the spectrum, virtually any other use of a computer has achieved limited success at best: the lack of payback from Office Automation, the lack of acceptance of expert systems. But few businesses now are willing to take the Amish approach, and live only with a calculator to perform and compete in their business: we are willing to accept higher risk and a smaller payback in exchange for the efficiencies and the expanded possibilities that these more complex solutions give us. Before airplanes, no one died in plane crashes, but... Now, having reduced the information system problem to a very overly simplistic cause, and suggested the impossibility of getting out of the complexity mess, let me step back and say that there may be light at the end of the tunnel. Kauffman believes that the nature of our universe in regards to complex systems creates "order for free." The idea is that even in the most complex of systems, there are forces that tend to produce a type of order, to organize components of the system. He argues that without this "order for free," the complex primordial soup of chemicals could not possibly have produced life. This means that there is some tendency to define a state of homeostasis in all complex systems. Further, given some basic laws/rules that may include some motivation or goal, complex components may join together to produce a result that none could have done alone. This reflects back to the discussion of a corporation's success. The point is, though, that rather than producing stability, this process produces more, larger, and more complex components that now interact in this new complex system. Information systems, therefore, can never simplify: they either increase the complexity of a system, or change the configuration of that complexity. This, in combination with the recognition (thanks to the Amish example) that the introduction of complexity always affects the environment into which it is introduced, suggests that a major role of information systems implementation is to define simple rules/laws that will trigger the "order for free" within the now-more-complex environment. As humans, we strive, generally, for order, understanding, and simplicity: this is our comfort zone (see Socio- technical systems theories). If, however, we are participating in a competitive situation (such as HP), and need to advance in order to survive, we must change our internal motivation to find happiness either in the process of change, or in the moment itself. In short, working in the high-tech industry, or in any industry for that matter, we must constantly lean towards the complex and unsettled at the expense of the simple and stable. This pressure will always stress established systems, whether they are bridges, airplanes, information systems, or (sadly) societal and ecological systems (forget getting into discussions of our responsibilities in relation to other people or the environment). This could go on forever... A final note: Because it is impossible for an individual to plan, design, and construct a 747 or a highway bridge, the engineers responsible for those projects have had to develop a very formal method for communicating requirements and expectations to those who will perform the various steps and construct the various pieces of those projects. As soon as more than one person is involved in such a project, the complexity of the interaction of project participants as well as of the project itself render it impossible to identify all the potential failure points. The primary task then becomes the management of risk of risk, as opposed to the management of risk itself. In Information Systems, we are only beginning to develop formal methods to communicate and manage expectations. In addition to continuing to develop these, maybe we need to devote some energy to managing the risk of risk (mr^2 ?) in our projects. Ben Lloyd WCSO/WSS Technical Architecture
....................................................................... TO: PODOLSKY_JOE/HP-PaloAlto_om4@opnmail2