Tuesday, July 13, 2010

Quick Note on Complex Systems

In the comment section of this post, Mattheus and I come to exactly opposite conclusions on how to gain knowledge about complex systems (well, not exactly opposite since I see the utility of both a priori and a posteriori knowledge). I start by writing:

"See, I think you turn the issue on its head. If economics were a simple, clean science like math you could use pure a priori. Some mathematicians now use quasi-empirical methods for tougher problems where the issues at hand aren't as clean-cut, but obviously most of it is a priori knowledge. When you get into more complicated phenomena that the human mind simply can't capture, more empiricism is introduced - like in chemistry or physics. The science of truly complex systems: biology, and the study of human society - have to rely even more heavily on empiricism because the processes are just too complex to faithfully render with a priorism."
Mattheus responds:

"If the human mind can’t capture it, then we shouldn’t be bothering with elucidating the truth of its claims. After all, in order to process empiricism, you NEED a theory behind it otherwise you’re just looking at a lot of data with no way to interpret it. And of we can’t understand it at all, then both modalities are worthless and we’ll never come to terms with it.

The better way to understand complex systems is not by directly observing millions of variables interacting in unknown ways, but to understand each component logically. Mises creates his evenly rotating economy as a thought creation – to understand the difference between profit and interest – because he knows that the economy is far too complex to understand at face value."
For me, "complex system" means a system with relatively simple components or constituents that interact to form a very complicated whole. The concept of complexity is very closely related to that of chaos, which should shed some light on my hesitation to rely exclusively on a priorism to understand any complex system. The agents may be simple. Man may act, as Mises says, and we may be able to derive an awful lot from that. In fact, in Mattheus's initial post which I was responding to, he describes what Mises derived from "man acts" as subjective preferences, time preferences, and the division of labor. For all the back and forth over Mises here, I'm not aware of a single economist that challenges these three fundamental characteristics of economic agents. And that's not surprising. In a complex system, agents are relatively simple and probably could be understood just lying on a couch thinking about them. It's not the only way to do it, but a lot of people do economics that way. There are considerably fewer disagreements among microeconomists than macroeconomists. The disagreements that do exist are largely empirical disputes: the magnitude of certain elasticities, etc. This is to be expected, because agent behavior is (relatively) simple.

The problem comes in when those agents start interacting in the complex system that is human society (both the market, the state, and any variety of other institutions). This is where my understanding of what a "complex system" is is important. Relatively simple agents that on their own have predictable behavior can produce alternatively erratic, structured, or evolutionary group behavior. Or they can fall victim to a sudden cascade of collapse. The interaction of billions of simple agents can have complex and unpredictable results that can leave a priorists high and dry simply because we aren't smart enough to think through all the contingencies on our couch.

This isn't to say that theorizing about complex systems is unfruitful by any means. We need to think through the possible contingencies and the possible complex interactions. But you can't expect to be able to report back on how the economy works after a good couch-sit the way that Mises does. At best you've come up with some ideas on how it could work, and even probably how it does work under certain circumstances.

So what does empiricism have to offer? Well, at the very least it can test theories. The nature of complex systems gives us reason to doubt our a priorism, but presumably we can still observe the system to see if our predictions hold true. Mattheus references "millions of variables", asserting the problems with empirical work on complex systems. I'm not quite sure what millions he is referencing. You can look at just a couple variables to see if they behave the way you expected them to behave. The point is it's an iterative process. We have some ideas about how the system as a whole behaves, but the complexity of the system gives us serious doubt about how much faith to put in those ideas. So we look to see which ideas best match up with reality, we look for interesting trends that are left unexplained and then we try to go back to the drawing board and explain them. This is more the scientific method than it is Kant's "pure reason" at this point, but that's ultimately what the problem requires.

The nature of complex systems also speaks to a common critique you hear of macroeconomics, that because they are in such stark disagreement it is somehow bogus science. Russ Roberts calls macroeconomics "ex post story telling". I think this critique is wrong-headed. We know complex systems act in unpredictable ways. They can form spontaneous order and then suddenly collapse. Or they can reorder into something else. Robust systems are best, but you don't always get robustness. Why would you expect one macroeconomic theory to be able to describe all of this. As I've said before, most macroeconomic theories and models are "necessary but not sufficient". The question usually isn't "is this one right?". Often the models are thoughtfully constructed and tested and usually do hold under certain circumstances. The right question is "is it right under this circumstance?". If you think of things in that sense, what you have isn't a self-contradictory science of the macroeconomy. What you have is a fairly insightful science of one of the most complex systems around, but a science where all the scientists are bizarrely combative. Macroeconomics sometimes feels like a cardiologist telling a pulmonologist that pulmonology is "wrong" because their patient died of a heart attack. If doctors explained illness the way economists explained recessions - trying to find a single cause rather than a mix of causes that are more or less important in a given circumstance - our health care system would fall apart. But the point is, the problem isn't with macroeconomics. The problem is with macroeconomists.

Although all those thoughts of mine are based in my understanding of the macroeconomy as a complex system, I'm getting a little far afield and I have to start getting ready for the day. So I want to end with some links.

First, the work of the Santa Fe Institute shaped a lot of how I think about complex systems. They do a lot of interesting work across a lot of fields. Notable economists working at the Institute are Rajiv Sethi (often cited by Krugman and Thoma) and Sam Bowles (spoiler alert because I'm sure it will come up in the comment section - Bowles is a Marxist).

There is a New England Complex Systems Institute as well. I've glanced through their work before but I'm not as familiar with it.

George Mason University has a Center for Social Complexity.

Two books that see the economy as a complex system that I recommend are Paul Krugman's The Self Organizing Economy, and Thomas Schelling's classic Micromotives and Macrobehavior. Of course, Hayek is great for these insights as well, and has done more than most at looking into the mechanisms that organize simple agents into complex systems, namely the uses of knowledge in society. I think it's probably fair to say that Hayek emphasizes the order of these systems, whereas Krugman and Schelling give a hearing to both the order and the disorder (or at least sub-optimality) of complex systems. All three men are Nobel laureates and all are very good.

And finally, here's a complex systems journal.

7 comments:

  1. The problem comes in when those agents start interacting in the complex system that is human society (both the market, the state, and any variety of other institutions).

    This is another mistake you make. You see "the market," and "the state" as abstractions that humans have to deal with that upsets our apriorism, but it doesn't. Humans deal with humans. Everything that can be said about human action holds true apodictically in a market setting. A marketplace is nothing more than a gathering of people to meet ends. There's nothing "alien" about it.

    But you can't expect to be able to report back on how the economy works after a good couch-sit the way that Mises does.

    You seem awfully confident to denounce praxeology without so much as a glance at it. For maybe the 100th time, I suggest you read Mises first hand. I recommend Human Action, personally.

    ReplyDelete
  2. This is another mistake you make. You see "the market," and "the state" as abstractions that humans have to deal with that upsets our apriorism, but it doesn't. Humans deal with humans.

    Could you clarify? When do I deny that it's all humans interacting with humans? When do I make the market or the state an abstraction with which humans interact? I certainly don't intend to make that statement, and I don't think I did either. Humans interact with other humans. We call the set of a specific type of those interactions a "market" or a "state", but humans are always interacting with other humans.

    And none of this upsets a priorism in theory. What it does is make a priorism considerably harder because the complexity of agent interaction increases exponentially. You can't think through what goes on when two agents make a trade and just extrapolate that to what goes on when millions of agents make trades that all influence each other. A priorism is valid theoretically, but frustrated practically. That is the point. Complexity doesn't invalidate a priorism, but it does make a priorism less appealing as a strategy and a posteriorism more appealing.

    Everything that can be said about human action holds true apodictically in a market setting.

    Yes. You can continue to say things about the action of humans. What you cannot necessarily describe with confidence is the behavior of the system. Essentially, you risk committing a fallacy of composition as systems become more complex. That is the major liability of Austrian economics: fallacies of composition.

    A marketplace is nothing more than a gathering of people to meet ends. There's nothing "alien" about it.

    Couldn't agree more.

    You seem awfully confident to denounce praxeology without so much as a glance at it. For maybe the 100th time, I suggest you read Mises first hand. I recommend Human Action, personally.

    1. I'd be careful not to confuse one book with an entire epistemology. I can speak to an epistemology without reading the book.

    2. But of course I would like to read the book. Time is scarce, though, and so far I've personally found looking in more detail into Austrican business cycle theory to be more fertile than looking into Austrian epistemology.

    3. If I'm inaccurately describing Mises, I'd be happy to take it back - but I thoguht this was precisely his project. My language may seem flippant, but "couch sitting" is just a reference to Galen Strawson's own description of a priorism.

    4. Again, it's not precisely that a priorism doesn't work for complex systems, it's just that it's considerably more risky and less likely to produce fruitful results. This should be obvious from looking at other sciences - the more complex the system that is under investigation (using the definition I've used here), the bigger a role empiricism plays in the discipline.

    ReplyDelete
  3. RE: "Everything that can be said about human action holds true apodictically in a market setting.

    Yes. You can continue to say things about the action of humans. What you cannot necessarily describe with confidence is the behavior of the system. Essentially, you risk committing a fallacy of composition as systems become more complex. That is the major liability of Austrian economics: fallacies of composition."


    And I should also add, this is clarified by the point that generally speaking markets clear in The General Theory. Keynes pays lip service to the prospect of non-clearing labor markets (which as a real world prospect is something worth paying lip service to, at the very least), but markets clear for Keynes. Roger Garrison emphasizes and reemphasizes this in his book.

    So the difference between Mises and Keynes is not primarily about human to human interactions at all. Mises's a priorism and Keynes's theorizing (if not strictly a priorism) are coincident on this point.

    What differs between them is their depiction of the behavior of the system - something I'm claiming an a priorist approach to individual agent behavior and interaction has trouble with.

    ReplyDelete
  4. I don't know how it's possible, but I agree with your premise and disagree on the conclusion. I think it is because of the complexity of the system that makes empiricism more difficult, not vice cersa.

    That is the major liability of Austrian economics: fallacies of composition.

    Excuse me? This is the major liability of Keynesianism.

    Just because some employees are reluctant to accept cuts in nominal wage rates, labor as a whole should be reluctant to accept cuts. Just because some investors make poor decisions, private investment as a whole is untrustworthy.

    These are the messages I got from my understanding of GT.

    The statements that praxeology makes about action is valid for all man, without exception. That's the point of appealing to synthetic a priori. It doesn't make much sense for you to claim I'm making a fallacy of composition because I posit that humans "prefer sooner rather than later." The only statements praxeology makes about human action are those that are universally applicable to all man.

    My advice would be for you to study Misesian epistemology. The theories are important, but they are products of a rigorous underbelly. I suggest you acquaint yourself with it.

    ReplyDelete
  5. I'm suspicious of people who think they can refute someone "passage by passage" as it is always said of Hazlitt

    Extraordinary claims require extraordinary evidence. I would, however, acquaint yourself with Hazlitt's refutation if for no other reason than to see the Austrian objections to Keynesianism.

    ReplyDelete
  6. Excuse me? This is the major liability of Keynesianism.

    No, not at all. You have it backwards.

    Austrianism's blindspot is the fallacy of composition.

    Keynesianism's blindspot is the ecological fallacy.

    I discuss the two risks here:
    http://factsandotherstubbornthings.blogspot.com/2010/04/fallacious-fallacies-and-great.html

    ReplyDelete
  7. it does not follow that your ability to explain "all man" means you can explain the behavior of the system.

    We've agreed that "the system" is nothing more than multiples of people acting in concert. Praxeology is an instrument to understand the logic of action; it stands to reason that action on a large scale and be inferred from action on a small scale because action is action.

    I wasn't impressed at all with how Hazlitt addressed liquidity preference

    So try another chapter. Hazlitt analyzes the multiplier, the MPC, Keynesian theories on interest, etc. The entire book is given attention. Just because you found one chapter unconvincing doesn't mean there aren't substantial contentions in others. That's a fallacy of composition. :)

    ReplyDelete

All anonymous comments will be deleted. Consistent pseudonyms are fine.