Causality in complex systems/Workshop 2/Notes

[edit] Causality and explanation

Since the workshop (and the entire year's effort) is devoted to the study of causality in complex systems I had spent some time prior to the workshop thinking about causality. Although I hadn't formulated it as elegantly as Dowe, I had independently come to the conclusion that causality is essentially the transmission of a conserved quantity (such as energy or momentum). This notion of causality is limited to what one would intuitively think of as direct physical causation. It does not cover, for example, "The gardener caused the plant to die by not watering it." I was comfortable with that limitation and was happy to exclude the use of the term "cause" in those situations. One can explain that the plant died because the gardener failed to water it, but I would not use the term "cause" for cases of this sort.

Similarly I would be happy not to use the term "cause" in "The sounding of his telephone's ring tone caused John to press the phone's answer button." Again, one can look to the ringing to explain why John pressed the answer button, but I wouldn't say that the ringing caused him to press it. Nor would I use "cause" in "The red light caused Mary to stop the car" or in "Jane's early morning meeting caused her to set her alarm clock for 6:00am." or in "The fact that that's where the money is caused Willie Sutton to rob banks." Or what about this case: Joe tripped and fell after stepping on a banana peel thereby staying out of the way of a drunk driver whose car would otherwise have killed him. Did something cause Joe to remain alive? I wouldn't want to use the term "cause" here.

One of my fundamental examples was to deny that Meredith caused the light to go on by pressing the light switch. I argued that if anything it was the electric current flowing through the light bulb that caused it to go on; Meredeth's action simply enabled that flow. Similarly I wouldn't use the term "cause" to say that the input to a computer program caused it to produce the output or that a failed bit caused the output to be wrong. At best (in my view), the fact that the input was as it was is part of the explanation for why the output was as it was. The failed bit is another part of the explanation.

Although "explanation" seems to me to be the appropriate term for these examples, see Scientific Explanation for a discussion of how difficult it is to construct a philosophical theory of scientific explanation. My sense of explanation seems to come closest to that attributed to biologists, who are said often to "describe their explanatory goals as the discovery of mechanisms rather than the discovery of laws." This seems to me to be similar to how computer scientists—I am a computer scientist—think about explanation. To further compound the difficulties, though, see The Stanford Encyclopedia of Philosophy article on Mathematical explanation.

When I suggested Dowe's theory of causation as the best way to think about causation, I was told quite firmly by Sandra Mitchell, the one professional philosopher at the workshop, that transmission of conserved properties is considered a philosophical dead-end—to a great extent because it doesn't deal with the examples listed above. (Would one say that the fact that Dowe's theory fails to cover more metaphorical uses of the term "cause" caused philosophers to reject the theory?) After a bit of discussion it became clear to me that arguing for Dowe's approach was not worth the effort—especially since what I was really interested in was how to understand how things happen in complex systems—i.e., mechanisms and explanations—rather than to construct a philosophical theory of causation. (Since we didn't discuss the status of the philosophical theory of explanation I can't say what Mitchell's position is on that.) So I ceded the terms "cause," "causation," and "causality" to the philosophers and focused on what turns out to be my real interest, energy and complex systems.

For even more issues related to causation, see "Causality in Complex Systems - The Problem of Modularity", Mitchell's workshop presentation, and the Stanford Encyclopedia of Philosophy articles on Causal processes (edited by Dowe) and The Metaphysics of Causation.

[edit] Energy flows, cycles, agents, and Persistent Turing Machines

Leading up to the workshop I had become increasingly convinced that tracing the energy flows in a complex system would reveal valuable information about how the system is organized. After all, without energy, nothing happens. (See my Thursday thought piece.) Furthermore, the only forces that cause anything to happen are the fundamental forces of physics; everything else is built upon them. (This is not a reductionist position. See my "The reductionist blind spot" paper.)

I was also struck by comments about energy and cycles by Cliff and Anne-Marie. Cliff remarked that all complex systems have a few basic underlying cycles; that everything else that happens in those systems "rides on the back" of those cycles; and that those cycles are the inevitable result of the dissipation of energy by the system. See Morowitz, The Emergence of Everything.

An aspect of nonequilibrium thermal physics is that the flow of energy from an energy source to an energy sink through a nonequilibrium intermediate system leads to the organization of the intermediate system. This is quite general. It is sometimes referred to as the fourth law of thermodynamics.

This is also known as the max-entropy principle: systems increase entropy at the maximum rate available to them. See Morel and Fleck, "A Fourth Law of Thermodynamics" for a basic discussion, which does not mention max-entropy as a principle, and Kleidon (his Facebook page) "Non-Equilibrium Thermodynamics and Maximum Entropy Production in the Earth System", which does. See also the quirky Institute of Human Thermodynamics "Variations of the 4th Law of Thermodynamics" for 17 (and counting) versions of the fourth law. The page encourages readers to vote for the one they consider the best.

There was a good deal of discussion during the workshop about attractors and limit cycles. Anne-Marie noted the important distinction between attractors that result from energy wells and those that are maintained by the structure and operation of the complex system itself, what Cliff referred to at the system's global constraints. Although one tends to think of attractors as closed sequences of state transitions, I decided to understand cycles somewhat differently. In particular, I understood the fundamental cycles of a complex system to be something like the instruction execution cycle of a computer—or the rule application cycle of the Game of Life or the DNA-to-protein cycle of living cells. In other words, the basic cycles of a complex system are services that can be arrayed in such a way that more complex phenomena can be built on top of them.

Furthermore these cycles do work (in the sense of physics: energy transferred by a force acting through a distance) hence they require energy to operate. Thus it seemed to me that an important way to understand complex systems was to understand their basic cycles along with the energy pathways that power them.

I was also struck by the fact that all services (as in Service Oriented Architecture) are cycles. Otherwise we wouldn't be able to define finite processes that perform them. In addition I noted that all platforms are essentially services. All this came together to convince me that this direction might provide a methodological framework for analyzing any complex system.

Finally, after the workshop I recalled work that Wegner and Goldin have been doing on interactive computing. In particular they defined what they call a Persistent Turing Machine, which it seems to me is a good way to model agents in an agent-based system. With that in mind, I jotted down some notes ("Persistent Turing Machines, Agents, and Cycles") on how Persistent Turing Machines show that all complex systems that can be modeled as agent based systems are based on simple cycles.

So it is this area on which I have been focusing since the workshop.

[edit] Amazing niche occupation and Max-Entropy

After the workshop my wife and I spent a weekend in Sydney. We visited the Sydney aquarium and land animal parks. I was overwhelmed with the diversity of life. It's difficult to believe how many different forms of life exist on earth, each occupying its own energy niche. It is also amazing to realize that different as they are, all these life forms are build upon the fundamental DNA-to-protein cycle.

Recall that the Max Entropy Principle (MEP) is the hypothesis that left to itself, a system will dissipate any available energy gradient as fast as possible given the constraints under which it is operating. The multiplication of life forms and the way that so many energy niches seem to be occupied seems to offer some empirical—or at least arguable—evidence for MEP.

[edit] Table of Entity Types