Thursday, March 24, 2005

Re: A Critique of Social Science (Hank's Response)

Henry has been having some trouble with blogger, so I'm posting this for him. JB.

First I would like to thank you for taking the time to consider and criticize my writing. I allowed myself a little break after I read Joe's first response and feared that I might be a bit overwhelmed when I noticed upon coming back that there were three further additions. But to my relief these have merely refined the earlier arguments and done much of my work for me. So I will continue in my informal fashion and address the concerns that have arisen thus far.

The Issue of Inclusivity and the Barrier of Complexity

In terms of inclusivity I am largely in agreement with those criticisms that have so far arisen.

I am inclined, as Barry suggested, to declare that we ought to consider the physical or hard sciences those to which my criticisms are of little relevance. Everything else would then be soft science. There is of course the underlying philosophical issue, which Joe recognizes, that fundamentally, my arguments are universally applicable. That is, when I say, for instance:

"Many things will be nearly constant among the study populations… the effects of these factors will not be averaged out, nor will the experimenter have any sense as to which of them are significant."

There is no way that we can be certain that such a complaint could not be leveled against any experiment whatsoever. The constant that modifies the force of gravity may only appear constant. We have only been making physical observations for a very short period of time and there may certainly be significant variables that we have not accounted for because they change so slowly that the difference has been unnoticeable in the last hundred or thousand years (these are after all cosmologically miniscule time frames).

~> Read More!
Or, as Joe noted, the relations that we observe may be entirely superficial and coincidental and the real variables that govern the behavior of the phenomena we seek to understand could yet be entirely unknown to us.

In either case we must recognize that technically my arguments apply to all knowledge. All human knowledge is fundamentally uncertain. Thus we must fall back upon practical certainties. For the phenomena that we most thoroughly understand we cannot imagine any variables which we have failed to consider, and regardless of whether or not we fully understand the true basis of the phenomena we are at least able to make virtually perfect predictions as to the outcomes that the phenomena in question result in. Though the earlier arguments apply to our understanding of gravitation we can still use our knowledge of it to weave the paths of space probes in intricate patterns through the gravitational fields of the bodies of our solar system with uncanny precision.

With regard to what we may practically label as soft sciences this is not true. In the case of these sciences we can easily conceive of variables that may be significant but for which we are not accounting. Likewise in these sciences we regularly experience failures in prediction. Thus I would say that although in a strict sense all science is soft we might with a fair degree of certainty arrange them into the practical categories hard and soft.

I would also pause to note that the traditional classification of various types of science are largely arbitrary and, well, merely traditional. There are certainly fields of study that we yet call physics because they have grown out of the earlier fields of physics but which deal with such complexity that we must call them soft. I would also agree with Joe in seeing the medical sciences as existing on both sides of the boundary. I would not say that this is the argument that applies to string theory and advanced cosmological theories; these are merely speculation and as such are not science. I think their popularity largely arises from the same causes as that of the social sciences. This is not to say that there is no place for such theories or that their proponents are fools. Merely that they are speculations the validation of which is yet pending and thus are not posited (at least, not legitimately) in any way as scientific facts. They are neither hard nor soft.

As to the boundary of complexity itself this discussion should be illuminating. As we can never be certain that we are accounting for all of the significant variables, we can never be certain as to how complex an event is. In this strict sense there is no barrier. All phenomena are potentially too complicated for us to arrive at certain conclusions with regard to them. But of course we must consider the barrier not in the absolute sense but in the practical sense. Thus it is the line that I have already defined between hard and soft. I do not pretend that in the marginal regions this line is well defined. I do however maintain that the majority of cases are not marginal. Most of the principles of physics are predictive in an absolute sense; they will describe the future states of the system in proportion to the detail of the input data. For these phenomena we cannot conceive of any variables for which we are not accounting. We can say with just as much conviction that these things are false with regard to the social sciences in general.

I will conclude this section with a caveat that bears within itself the seed of a related argument. I have said that the social sciences are, in general, soft. Sometimes it is true that we may arrive at principles with regard to social phenomena which are relatively absolute. I will posit a general principle with regard to human behavior: Human beings abhor the touch of flame and they will strive to prevent themselves from being consumed by it. As much as any statement we may make with regard to human behavior as it occurs in the world this is certain (of course the exceptional phenomena of self-immolation is troublesome to this postulate; it would be a serious blow to the credibility of physics if objects occasionally fell up). Yet the question remains, of what use is this fact to us? The questions which we wish to solve with regard to social life involve the full complexity of human social interaction: what will be the interest rate on ten-year federal bonds 3 years hence? What political party will best enhance the well being of the people? How do we avoid social strife and physical violence? The acknowledgement of the initial fact does not help us to understand the more complex phenomena of which it is only a small part. We may understand the behavior of gravity and this may allow us to predict the path that a projectile will follow, but only if that path involves minimal interference. A cannon ball that is affected only by the wind will follow a relatively predictable path. An elastic rubber ball bouncing downhill on a roughly paved street in heavy traffic will not. Regardless of the fact that we know that the ball will follow roughly parabolic paths between collisions we are helpless to predict its actual overall path with any semblance of accuracy. Our fact about the behavior of men would likewise be useful if we were placing bets upon whether the man we are about to light on fire will attempt to extinguish the flames or rather quietly submit to their consumption of his flesh. Yet it is of little use in solving more broadly the questions related to human action. This point was implicit in my earlier arguments. For instance, I did not claim that the individual assertions upon which Matthew Rabin's paper was based were necessarily false in and of themselves, merely that in the context of the wider problem they were pitifully insufficient.

Chaos and the Transit of the Complexity Barrier

Here I will talk a bit about complex and chaotic systems. I am not an expert but I believe that I understand enough to cover the points that are relevant to our discussion. I would encourage you to engage this subject with confidence, as I don't think that it is at all beyond the average person to understand these things. As I stated in my first paper, the essence of mathematical ideas is best expressed in plain English; we need resort to equations only for the purpose of calculation (and as we don't need to make any calculations we won't need equations except as conceptual examples).

We must begin with systems that are neither complex nor chaotic. These rare systems are those for which we may discover exact solutions.

An example here would be two spherical objects moving through space exerting gravitational forces upon one another. We can write out the force acting upon each body exactly. The force will always be in the direction of the other body, it will be proportional to the product of the masses of the two objects, and it will be inversely proportional to the square of the distance between them. If we have an equation for the acceleration (the force equation that I just mentioned) we can integrate it to find an equation that calculates the velocities of the objects. Once we have the velocity equation we can integrate again to get an equation that will tell us the positions of the objects at future times. This position equation will allow us to plot the exact trajectory that the objects in question will follow given any initial setup (combinations of masses, initial positions, and initial velocities).

The preceding scenario was only made possible by our ability to integrate the equations in question. This is far from given. Equations that we can integrate are actually quite exceptional in real world problems. In the case above it is possible to do the integrations, but if there had been three objects instead of two it would not have been so. In taking such a step we move from the realm of exact solutions to the realm of approximate solutions.

I will step briefly aside to illustrate the concepts of exact and approximate solution.

[Imagine an object being thrown upward (we must neglect air resistance, so imagine we're on the moon). As before we can come up with the acceleration equation immediately: a = -g. That is to say, gravity is always pulling down with a constant force g (g is negative as I have defined the downward direction to be negative and up to be positive). In fact the gravitational force diminishes as the object moves upward (away from the body attracting it) but we are assuming that we can't throw it high enough for such changes to be significant. By integrating we can discover the velocity equation for the thrown object: v(t) = -g x t + vi. That is to say, the object has an initial velocity vi imparted by our throwing it and over time that velocity is reduced by the operation of the acceleration g. Vi is positive, as we have thrown the object upward. Here t is meant to signify time. Thus our velocity, v is reduced by g for each unit of t. Thus after one second (assuming that our unit of time is the second) v(1) = vi - g. If our initial velocity upward (vi) was 20 and g was 4 then it would take five seconds before our actual velocity (v) was reduced to zero (this is the point when the object ceases to move upward and it is the moment before it begins to move downward, thus it corresponds to the highest point of flight). Now by integrating once again we may discover the equation for the position, or height of the object: h(t) = 1/2 x -g x t ^2 + vi x t + hi. Now we have multiplied the first term by t to get t^2 and we have multiplied it by one half (or divided it by two). We have done this in order to calculate the change in position that results from the acceleration g. -g x t tells us what change in velocity g has resulted in (and this is why we have -g x t in the equation for v). Since initially g had had no effect this component starts at zero but by time t the acceleration g has generated a contribution to velocity equal to -g x t. So the average velocity due to g between time zero and time t is (-g x t + 0)/2 or (-g x t)/2 hence the factor of 1/2. In order to figure out how much distance was traversed due to this component of velocity we must multiply the average velocity by the time that the object has been traveling at this velocity t. Hence the additional t and thus we get 1/2 x -g x t^2. The we have vi x t which is easily understood, as the vi component of velocity is unchanging, thus we merely multiply vi times the time t to figure out what distance vi has caused the object to traverse. And finally we have the initial height hi which is the position that the object was thrown from: if t = 0 then hi = h.

This is the simplest example of something similar to the two-body problem described in the third paragraph of this section. The main purpose of it for now is to illustrate that when integration can be done we get nice equations, equations that we can simply plug numbers into in order to get exact solutions. Now I will solve the same problem imagining that we knew only the acceleration equation and that we could not arrive at the velocity and position equations using integration. This will require an approximate or numerical solution.

We know that we have an object beginning at a height hi traveling at an initial velocity vi and being accelerated downward at an acceleration g. We wish to know where this object will be at a future time t. Well, we know that at time t = 0 its position h will be h = hi. We also know that its velocity at t = 0 is v = vi. Thus we could assume that, for discrete periods of time, everything in the problem remains constant. Let us use intervals of one-tenth of a second. Thus we presume that for the first tenth of a second v = vi. If this is the case then at time t = 1/10 our height h will be h = hi + vi x 1/10. Our new velocity will be v = vi + -g x 1/10 and now this velocity will be constant for the next tenth of a second. So after two tenths of a second (t = 2/10) the object's height h = hi + vi x 1/10 + (vi + 1/10 x -g) x 1/10 and so on for each consecutive tenth of a second. If we wish to figure out where the object is 4 seconds after it is thrown we must execute 40 iterations of this process to find the answer. This is obviously much more difficult than simply plugging numbers for t, g, hi, and vi into the exact equation: h = 1/2 x -g x t ^2 + vi x t + hi. Furthermore, it is less accurate than doing so. The numerical answer will only be approximately accurate whereas the exact solution will be as accurate as the values of the inputs. You can, of course, increase the accuracy of the numerical method by reducing the length of the intervals between iterations but you must exchange the time required to make more calculations for the accuracy gained. ]

Now that we may use the concept of exact and numerical solutions we can proceed with the first step away from the most simple phenomena--those that yield to exact solutions--and toward more complex phenomena--those that require numerical solutions. The problem of three spherical bodies moving in space influencing one another via the force of gravity does not have an exact solution. There is no equation into which we can plug the initial conditions in order to find precisely what the subsequent states will be. We must utilize a numerical solution to solve this problem. This is manageable. Modern computers were designed for the express purpose of performing such numerical solutions (and business calculations which are different) and they greatly facilitate this procedure. But the important thing to recognize is that the transition from the two-body to the three-body problem does not involve a 50% increase in computational involvement but rather a manyfold increase.

After the transition from exact to numerical solutions we continue for some time along the complexity continuum without encountering a similar barrier. In this region additional factors will increase the overall computational complexity of the system faster than linearly but not in the exaggerated way that it jumped when we crossed the barrier of exact solution. When we move from the three-body problem to an equivalent four-body problem the computational complexity increases by more than thirty three percent and here is why.

In the three-body problem we have to do three sets of calculations per iteration, one for each body. For the four-body problem we need four sets of calculations. If they were the same calculations as before we would expect the computation to take 33% longer, but the fourth body increases the complexity of all of the computations for each of the other bodies. With three bodies we had to take into account the position of only the other two, now with four we need to take into account the position of three other bodies in order to calculate the resultant force on each body for each iteration. The inclusion of the fourth body, then, not only requires that we make an additional set of calculations each iteration but it also makes each individual set of calculations more involved. Further, with four bodies we have to make four approximations per iteration and the resulting increase in error is compounded with each iteration. Thus we are required to further reduce the intervals between iterations in order to achieve the same level of accuracy.

We can see that the more complicated a system is, the greater the complexity will be increased by an additional element.

So long as we can make a wide variety of approximations this is still not too troublesome. Experience has shown that we can engineer a wide variety of useful systems using numerical computation techniques and our powers of computation are rapidly rising; although a doubling in computing power may fall significantly short of resulting in a doubling of the complexity of the systems we may explain, progress is being made.

Now we arrive at the realm of the chaotic. It is frequently said that chaos theory is a misnomer and we must agree. Chaotic systems are not truly chaotic; that is, random and unruly. They are merely systems of such extreme complexity that numerical modeling is bound to fail because the inevitable approximations result in very significant errors. They are only chaotic in the sense that they are utterly unpredictable to us.

For our chaotic system imagine a two dimensional surface. Imagine that this is the surface of a turbulent pool of water; it contains a wide variety of flows and whirls. Now we place a bit of cork on the surface of the pool and our goal is to predict where the cork will be at some future time t. We can define the cork's location on the surface with two coordinates; let us call them x and y. We can define the flows over this surface as an equation that maps an x and a y input to a two dimensional vector. That is, for a given point on the surface (x1, y1) there will be a corresponding vector that describes the direction and magnitude of the current in this location. We can assume that the cork does not have significant momentum and that its velocity is always exactly that of the current on which it is resting. Thus at point (x1,y1) the velocity vector of the cork is identical to the current vector for the same point . We can use numerical methods to calculate the trajectory of the cork. If the cork is at (x1,y1) then its velocity at this instant is . We can multiply the velocity vector by a time interval t in order to determine the approximate displacement for that interval. We add the changes in x and y to the original position (x1,y1) to get a new position (x2,y2) = (x1 + dx, y1 + dy). We can enter the new position of the cork into the equation for the flows over the surface to find the direction and magnitude of the current at this new position. From here we may repeat the process through subsequent iterations in order to plot a trajectory for the cork.

So long as the system is not complex enough to be chaotic this method will work. Here is how we might determine whether or not it is chaotic.

With a system that yields properly to numerical computation we find that the results brought about by subsequent reductions in the time between iterations approach a final value. The paths get smoother, the final answer gets more precise and our knowledge is increased. But with a chaotic equation this is not the case. We begin with intervals of 1/10th of a second and get a tentative trajectory for the cork. When we reduce the interval to 1/100th of a second we get an entirely different trajectory. When we reduce it to 1/1000th of a second we again get an entirely different trajectory. The values do not tend toward some definite path. Now let me quote from my first commentary:

"The American Heritage Dictionary provides the following as the mathematical definition of chaotic: A dynamical system that has a sensitive dependence on its initial conditions."

Hopefully this is more meaningful now. Depending on what interval length we choose we will get a different position after the first iteration. Working from these variations on the second point (x2,y2) we arrive at entirely different trajectories. A seemingly insignificant change in the initial conditions brings about results so different that there appears to be no correlation between the initial and the later states of the system. Of course, there is a correlation. The flow equation is perfectly determinate, but, due to the impossibility of exact solution and the failure of numerical methods, we are at a loss to say what it determines. Again I quote myself:

"The formula that those economists are looking for is infinitely complex and chaotic and if they were ever able to discern it, its implications would be incalculable. Seemingly insignificant variables would form endlessly diverse combinations to trigger extremely significant events."

We know precisely the equation that describes our hypothetical pool yet the trajectory described by the cork is as mysterious as ever. Likewise, if the economists were able to derive perfectly the equation that would describe the level of the Dow Jones Industrial Average (something that I maintain is an impossibility) it would be useless to them, as its implications would remain incalculable.

Now we must discuss what precisely chaos theory is. Up to this point I have only said that chaotic systems are necessarily mysterious. If this summed up chaos theory it would not be very impressive and I would indeed be an expert. But, of course, chaos theory has more to say about chaotic systems. Chaos theory attempts to determine if there are things that we can say about chaotic systems even though we cannot predict future states with precision. With regard to our pool con bobbing cork they might be able to say that when placed in region A the cork always shoots into the upper right corner of the pool and bounces around for a while. With regard to a particular whirl there may be a region in which placed corks will always leave the whirl through the bottom half. Chaos theory then gives us an assortment of rules that we may glean from the flow equation, making it not entirely useless to us. Thus if we could arrive at equations that perfectly map the reality of economic systems then perhaps chaos theory could teach us something about the actual operation of the system even though we could not calculate actual future states.

My warning with regard to the applicability of chaos theory to social phenomena is that chaos theory, like exact solution and numerical computation, yields results only as sound as the equation that you are working from. There are many mathematicians attempting to apply the principles of chaos theory to economic problems but I do not believe that they will succeed, as the equations they are working from are insufficient (and are necessarily so as the perfect equation would require a practically infinite number of variables): garbage in, garbage out.

As to the transit of the complexity barrier, it is safe to say that this barrier is significantly mobile. It is undoubtedly true that this barrier has shifted quite noticeably in our own lifetimes. But then again, sixty miles an hour is, to us, a noticeable velocity, while the universe would end before we reached the nearest star traveling at such a speed. This is to say that though the progression of the barrier is clearly occurring it need not be significant in terms of problems which are the complexity equivalent of intergalactic distances. In the next one hundred years our computational abilities may increase one hundred trillion fold, and this would certainly allow us to solve, to our great advantage, many problems that are currently beyond us, yet our progress would still be unnoticeable in terms of ultra complex social phenomena.

I will finish with a couple of examples that ought to add heuristic weight to these arguments.

One I have stumbled across at random. I quote:

"For a single spoonful of air we would need to know the position and velocity of 1020 molecules bumping into one another about 6 x 109 times a second. The physicist Michael Berry considered a collection of oxygen molecules at atmospheric pressure at room temperature. He imagined a single electron placed at the edge of the known universe (somewhere around 1010 light-years away). After how many collisions would a given molecule in the oxygen miss a collision with another molecule which it would not have missed had the electron not been there? Remember that the electron is affecting the oxygen only by its gravitational field, which must be so weak that we can virtually discount its effect. You might think so, but in fact Berry calculated that the oxygen molecule would miss its collision after a mere 56 collisions."

I couldn't find an original source for this, so I cannot know what calculation results in this fact. Even if Berry were off by several orders of magnitude one could not predict the position of his oxygen molecules, even a fraction of a second into the future, without accounting for every particle in the universe. This is complex interdependence. Such unpredictable interactions are undoubtedly percolating up from the microscopic world at every moment.

We may also consider again the elastic rubber ball bouncing down a roughly paved street in heavy traffic that I mentioned at the end of the previous section. Perhaps thirty meters down the hill the ball is struck by a fast moving minivan and it ricochets over a nearby building. This is a very significant event in the course of the phenomenon we wish to understand. So we attempt to trace the cause of the minivan's appearance. Its driver, a middle aged woman, is traveling to the grocery store to buy a cake. This is because the cake that she had planned to serve was ruined. This is because a vase full of flowers fell upon it. This is because her housecat, in the pursuit of a buzzing housefly, bumped into the vase. The fly had entered the house because a whiff of delicious cake odor had been wafted out the window by a gentle breeze that would not have occurred had not a star located a billion miles away gone super nova a billion years ago last summer. This, again, is the sort of complexity that I am talking about. A researcher trying to understand why the ball behaves the way it does would have a wide field of study indeed. They could write papers on different types of cars. They could develop classification systems for potholes. They could explore the motivation of house cats. They could write papers about the reasons that people drive children to school or they could contemplate the effects of deep space astronomical events on gentle breezes. Quite simply, every researcher, in every university on earth, could devote all of their time to studying the problem of the bouncing ball and the poorly maintained street, and such could remain the case for a million years, and yet, they would be nowhere near exhausting the subject of study and nowhere nearer to predicting the trajectory of the ball. I do not see how anyone could believe that the GDP of the United States in 2006 or the causes of the next world war would be any less complicated than the bouncing ball (for all we know the bouncing ball could (along with an infinite number of other factors) cause the next world war or economic collapse).

Simply put, physics succeeds because it considers unimaginably simple systems in isolated situations. The social sciences fail because they cannot explore such simple systems because the most basic element of their study, a human being, is extremely complicated. It brings into the experiment all of the experiences it has ever had, it is necessarily impressed by an infinite number of uncontrollable and unique circumstances. So far as social scientists can arrive at definite conclusions these are only tiny facets of the larger equation that explains the actual system in question (a bouncing ball researcher may be perfectly correct in asserting that a reason for a vehicle to be on the road is because its occupant may be on the way to a movie, but so what?); the equation that, even if we had it, still would not be very useful. This refers back to the illusion of progress section of my original essay; the accumulation of social scientific knowledge appears promising but in reality we could fill countless libraries with such information without getting any closer to solving such questions as: How should we educate our children? How do we reduce criminal behavior? How do we enhance the satisfaction of mankind?

Deduction in General and Ludwig von Mises in Particular

Joe rightly notes that the social sciences clearly make use of deduction. Now I did state that deduction is an essential part of the scientific method, that induction only arrives at postulates and so far as scientists ever reach new conclusions and hypothesis they must use deduction; so I don't think its fair to suggest that I did not allow that the modern social sciences use deduction at all. But I do not suppose that this is what Joe meant; rather he was referring to my argument that Homo Economicus is used for modeling and not purposes of deductive argumentation. Although I do admit that my point in this regard is a bit fuzzy (this is one of the reasons that I did not give you the Rabin paper as my original contribution; it didn't hold together that well), but I think I can at least bolster it up for the time being. Before I do though I have a quick unrelated response.

I would like to say that I did not suggest that "Rabin's paper [was] purely a quixotic exercise attacking a long-defunct economic model." Indeed I think the fact that the majority of modern economists utilize this principle was clearly implied in my paper. My writing "This [assumption that men rationally strive toward the maximum fulfillment of stable and well-defined preferences] is not, therefore, the basis of any sound systems of economic reasoning" does not imply that the majority of modern systems of economic reasoning do not rely upon this principle; it merely states that such systems are unsound. Nor does the statement, "The presumption that man acts always to maximize his rational self-interest is undoubtedly the single most frequently and most easily criticized tenet of any economic theory" imply that this tenet is not integral to the dominant economic theories of our age; it is only meant to say that these theories are frequently and easily criticized, and rightly so. I was not trying to suggest that Homo Economicus was a straw man in the sense that no one used him in their arguments; I was suggesting that he was a straw man in the sense that he is admittedly faulty and his explosion in no way affects the conclusions of proper (though perhaps not popular) economic theories.

In order to defend my characterization of Homo Economicus I will need to deal with some finer points of the definitions involved. In doing so I will draw upon Mises, so this section will also deal with the criticisms of Misesian/Austrian Theory that arise in Joe's response number III.

The definition that will be most important to us here is that of the term 'rational'. As Joe rightly points out Mises relies upon something that appears quite similar to Homo Economicus. The difference between what Mises proposes and the rational actor Homo Economicus is quite subtle. In fact, virtually all of the ways in which Mises differs from the classical economists are quite subtle, as they must be, for the classical economists were no fools. But it is most often Mises' most brilliant insights that are most quickly criticized by those who cannot help making the same mistakes that the classical economists have made. It is certainly true that on page nineteen of Human Action Mises states:

"Human action is necessarily always rational. The term "rational action" is therefore pleonastic and must be rejected as such. When applied to the ultimate ends of action, the terms rational and irrational are inappropriate and meaningless."

It is also true that this statement is absolutely fundamental to the entire body of work that is Human Action, but it is not true that the behaviorists can successfully criticize this statement. It is a definitional, not an objective, statement. His goal is to define human action, and here we must consult what he means by 'rational' as is elaborated on page twenty:

"When applied to the means chosen for the attainment of ends, the terms
rational and irrational imply a judgment about the expediency and adequacy
of the procedure employed. The critic approves or disapproves of the method
from the point of view of whether or not it is best suited to attain the end in
question. It is a fact that human reason is not infallible and that man very
often errs in selecting and applying means. An action unsuited to the end
sought falls short of expectation. It is contrary to purpose, but it is rational,
i.e., the outcome of a reasonable—although faulty—deliberation and an
attempt—although an ineffectual attempt—to attain a definite goal. The
doctors who a hundred years ago employed certain methods for the treatment
of cancer which our contemporary doctors reject were—from the point of
view of present-day pathology—badly instructed and therefore inefficient.
But they did not act irrationally; they did their best. It is probable that in a
hundred years more doctors will have more efficient methods at hand for
the treatment of this disease. They will be more efficient but not more
rational than our physicians."

This is to say that, by rational, he means that action is considered and supported by reasons. He is contrasting conscious action with unconscious physical response. When one ingests a toxin one does not weigh reasons and costs in determining whether or not to use one's biological defenses to attempt to break down the poison. You do not consider the pros and cons before blinking when an object comes flying at your face. These are not actions in the sense that Mises is speaking of. The principles that he demonstrates in Human Action are only true of conscious, volitional, and, in this sense, rational action. They are absolutely true with regard to action, as such, and they are of no consequence with regard to unconscious response. This is what Mises means when he says " Human action is necessarily always rational." In this sense the statement truly is pleonastic. The statement understood in this way cannot be refuted by behavioral observation.

It is in using a very different definition of rational in a very similar context that the employers of Homo Economicus make a concession to modeling. Although I must admit that this error was originally made by the classical economists who reasoned largely in a deductive fashion. It could only have been necessary for the development of objectively true predictive models. That is, the definition used by Mises is actually indisputable and hence, is an excellent building block for a deductive argument. The alternative, which we have labeled as Homo Economicus, is easily disputable and as such is not a suitable premise for deduction. But it does have other merits. If we assume rational to refer to a specific set of behaviors, those ideally suited to the achievement of specific goals, we can model human behavior. We can say that given goal A, there is only one behavior B, which is ideally suited to the achievement of this goal. If we then assume that man is, in this sense, rational, we can, given his preferences, predict exactly his behavior. The assumption that man is rational in this sense is indisputably false.

Now moving on to the criticisms of Human Action not related to Homo Economicus. We arrive at the quote:

"Concrete value judgments and definite human actions are not open to further analysis. We may fairly assume or believe that they are absolutely dependent upon and conditioned by their causes. But as long as we do not know how external facts--physical and physiological--produce in a human mind definite thoughts and volitions resulting in concrcete acts, we have to face an insurmountable methodological dualism"

Again only by misinterpreting this quote can it be criticized. It comes from the section explaining that we must consider human action to be an ultimate given. This is true because we do not understand its causes. We cannot penetrate the barrier of human action in our economic reasoning so we must acknowledge this fact. We cannot say, based upon our analysis of the firing of neurons, whether an individual will buy a Ford Taurus or a Honda Accord. We cannot predict whether a person will choose to work overtime or go golfing. Thus "concrete value judgments and definite human actions are not open to further analysis", and they will remain so "as long as we do not know how external facts--physical and physiological--produce in a human mind definite thoughts and volitions resulting in concrete acts". I did not think that the behaviorists had achieved such things but perhaps I am just ill informed. Indeed, the sentences that complete the paragraph from which the above quote was taken make it quite clear that Mises does not necessarily consider understanding the underlying causes of human behavior as "futile" (as I do), but merely as a task yet to be accomplished:

"Concrete value judgments and definite human actions are not open to
further analysis. We may fairly assume or believe that they are absolutely
dependent upon and conditioned by their causes. But as long as we do not
know how external facts—physical and physiological—produce in a human
mind definite thoughts and volitions resulting in concrete acts, we have to
face an insurmountable methodological dualism. In the present state of our
knowledge the fundamental statements of positivism, monism and
panphysicalism are mere metaphysical postulates devoid of any scientific
foundation and both meaningless and useless for scientific research. Reason
and experience show us two separate realms: the external world of physical,
chemical, and physiological phenomena and the internal world of thought,
feeling, valuation, and purposeful action. No bridge connects—as far as we
can see today—these two spheres. Identical external events result sometimes
in different human responses, and different external events produce sometimes
the same human response. We do not know why.

In the face of this state of affairs we cannot help withholding judgment
on the essential statements of monism and materialism. We may or may
not believe that the natural sciences will succeed one day in explaining
the production of definite ideas, judgments of value, and actions in the
same way in which they explain the production of a chemical compound
as the necessary and unavoidable outcome of a certain combination of
elements. In the meantime we are bound to acquiesce in a methodological

Even if the behaviorists where to explain the underlying causes of human behavior the act would render Mises' work incomplete rather than false.
Before going, for a moment, on the offensive I will cover the final Mises quote in III:

"whether or not the means chosen are fit for the attainment of the ends aimed at." (p. 21)

This statement is not, originally, used in reference to human behavior or human rationality, but with regard to the science of human action itself. Putting the quote into context is again quite useful:

"In this sense we speak of the subjectivism of the general science of human
action. It takes the ultimate ends chosen by acting man as data, it is entirely
neutral with regard to them, and it refrains from passing any value judgments.
The only standard which it applies is whether or not the means chosen are fit
for the attainment of the ends aimed at."

In fact, if one were to assume that human action where always perfectly suited for the attainment of the ends aimed at, this statement would be nonsensical, as the science of human action would be entirely superfluous. Mises is not an idiot. He is saying that the science of human action is neutral with regard to ends and that its object is to help man choose means that are suitable to attain his ends whatever they might be.

Now I would like to take a moment to analyze the following statement from III:

"It is precisely the claim of the behavioralist that an individual's actions do not rationally correspond to that individual's own value judgments. It is not a matter of imposing outside values (which Von Mises here argues against), but merely questioning the validity of assuming the rationality of action as defined by consistency with the desires and value judgments of the actor."

How do the behaviorists determine what a person's value judgments are? I suppose they must ask the individuals. But then, responding itself is an act. How can the researchers be certain that it is in the response to the question "What are your value judgments?" that an individuals true interests are represented rather than in their subsequent actions? Is it not possible that a person's value judgments change with time and that their actions are consistent with their valuations at the moment of action? Could the truth not also be that people are not generally capable of (or willing to) accurately identifying their own value judgments? If a man says that he values, above all else, his income, and then, on Sunday morning, he goes to church is he acting irrationally? Or is he simply more clearly aware of his desire for income than his desire for a reduction in his fear of death and uncertainty, or perhaps his valuation of social approval and conformity. If a man (who states that his objective is to maximize the quality of the VCR he purchases relative to his dollar cost) buys the more expensive of two VCRs when, had a third option, cheaper than the other two, been available, he would have chosen the middle alternative, is he truly acting irrationally or, is it possible, that he is optimizing preferences that he did not know how to express. Perhaps he should have said that he would also prefer not to have to spend too much time learning about VCRs. In accordance with such a preference he may have made his decision based upon a rule of thumb that is different in a case where three options are present than if two are. Perhaps one rule is that one should never buy the lowest cost alternative for such a product is generally shabby and inferior and the ownership of such a thing is popularly regarded as shameful. The second rule of thumb is that one need not purchase the most expensive option, as this is extravagant and wasteful. In a case where only two alternatives exist the two rules cannot be satisfied simultaneously and the first rule is a more significant value for this individual than the second. Thus he buys the more expensive VCR. In the case with three options he may satisfy both rules and thus he buys what in the first case was the least expensive VCR. One cannot say that, within the context of these explicit valuations, his decision was irrational. One might say that it is irrational to hold the contradictory values of trying to obtain the finest product for the least money and making the purchase while learning as little about the product as possible, but this would require the observer to impose a value judgment upon the actor.

Indeed as a heuristic exercise view yourself introspectively. In every action are you not optimizing your own satisfaction? If you tell others that you are striving to succeed in a given course and then you skip class, or ignore an assignment, have you done so out of a failure to recognize that the action will bring about results contrary to your goal? Or is the case merely that you have experienced a preference that overrode your initial desire to succeed in the course, such as a desire for leisure, a desire to spend time with a girl, or a desire to go drinking with your friends? Is it reasonable to expect that, asked the question, "What are your preferences?" you will accurately enumerate all of your actual preferences and be able to order them in such a way that will be universally true? It seems far more rational to say that if in circumstances A I choose to skip class and stay in bed the action itself represents my true preference. At the moment that I declared my intention to succeed in my class I may have been absolutely sincere, but I am equally sincere in my valuation of sleep over the class when I turn off my alarm and roll over.

Returning to the case of the VCRs, we have seen that to choose two different VCRs in what are technically the same circumstances can be a rational action. There is no objective way to determine which VCR was a better value in an absolute sense, and from the perspective of the buyer the cases were substantially different; he operated in accord with his actual values. A person may act differently in a crowd than they would if they were alone, but are they violating their preferences or are their preferences for conformity and acceptance overriding the preferences that would have been dominant in the absence of the crowd? To declare that a person's actions contradict their values and desires presumes that the critic understands the actor's values and desires; a premise that I hope I have demonstrated is highly suspect. It is just this sort of false certainty that deludes the followers of the social sciences. Mises' work is truly scientific because of the extraordinary care that he takes not to assume the veracity of what is not certain.

Post Script

Thank you again for taking the time to consider my ideas. I understand that I have been running a bit long but it is a great pleasure for me to put these ideas into writing. I hope that you will continue to spend some time in critical response. There is no better catalyst for clear thought than external challenge and review.

I would briefly like to advertise the work of Ludwig von Mises. As Joe indicated, his primary work Human Action (and many secondary ones) is available for download at It is a bit long and you might prefer to buy a hard copy but it is well worth the read. In the context of my experience I would say that it is the finest philosophical work of the twentieth century. You may disagree, but if you are a logically minded person you will not be able to stop yourself from admiring the rigorous consistency and conceptual breadth of his work.


Joe said...

Without making any substantive comments, I'd just like to note that your explanation of chaotic systems and chaos theory is phenomenal!

vox humama said...

I am out here in the hinterland with the rest of the "rational" population, so reading this was the equivalent of mental weightlifting. I am happy to state that I not only got it done but actually feel stronger for it. As a member of the over 50 age group, my eyes were not well served by the experience. (BIGGER PRINT WOULD BE APPRECIATED!)
It would be hard for me to make any comments of a sufficient intellectual rigor to feel worthy of this discussion. But my question would be, Is the academic world, particularly or especially the "disciplines" of the social or soft sciences, engaged in this kind of discussion? Or is it the elephant under the proverbial rug?