Fickle preferences...

I still haven't yet finished that old paper by Gary Becker, but I hope to get back to it soon, and see whether or not I treated his ideas fairly in The Social Atom. In the paper, he's trying to argue that economists can make much weaker assumptions about the rationality of people, and still derive some of the key theorems of economic theory. I have my doubts, but I'm not quite ready yet to lay out why in detail.

One of the problems, I suspect, may hinge on the notion of fixed preferences. Economists tend to assume that people have fixed preferences -- say, for deciding between apples and oranges. If those preferences are indeed fixed and independent of any situation, then it's easy to understand if you see lots of people eating apples and few eating oranges; what you see in the group, in the aggregate, is merely the summation of the preferences of the individuals. There's a technique that theorists use to formalize this intuition, called the Representative Agent. If you see that 80% of the people are eating apples, then you can infer that 80% prefer apples to oranges. (Effectively, you could also treat the response of the crowd as if it was the response of a single "representative agent," who likes to mix apples and oranges in a four to one ratio; in mathematical theories, this is done for convenience.)

Trouble arises, however, if one assumes that this representative agent trick can always be carried through. And it clearly cannot if people interact (which they do!), and if what one person prefers depends on what others prefer, or on what they actually choose to do (which it often does!). In an extreme case, for example, most people might (on their own) prefer apples to oranges, but also have a strong preference to eat what they see most other people eating. If one person starts out by eating an orange, and a second follows suit, so may all the rest, leading to an outcome that doesn't reflect the "innate" preferences of the group, but the way the dynamics of influence can easily twist peoples' preferences away from their independent values.

Economists have long tried to get by with this representative agent trick, as it greatly simplifies the task of going from some knowledge of what individuals want to how a group will behave. Unfortunately, it is a very serious mistake, as it effectively cuts the interactions of people out of the loop. And interactions count.

As one intriguing example, some Chilean and Columbian economists, working at the Santa Fe Institute, have recently been studying the way people manage limited resources. In experiments, they had people "extract resources" from a mythical reservoir (think grasslands, minerals, fresh water, etc). Each person could choose how much to take. The more they take, the more they get, in general, except there's also a cost -- if everyone takes a lot, the resources gets depleted and then everyone suffers. The study is fairly involved, and I won't get into the details, but one fascinating outcome is that the fact that peoples' preferences can be influenced by what others do can have a great effect.

Many individuals feel guilty if they take too much (more than an accepted social norm). In some of the experiments, they found that this guilt (particularly when backed up by the small threat of a fine for taking too much), led to a stable situation in which the people extracted resources in a healthy, sustainable way. However, many people only feel guilt in overextracting as long as others aren't overextracting too; if many others do begin to take out too much, many individuals lose their guilty feelings and will join in. As a result, even under the same conditions, the group's behavior can end up in two different states, pushed into one or another by some initial accident -- not unlike the example of the apples and oranges.

I think this emphasizes the major theme of The Social Atom pretty well; people cann't be understood in isolation, and then summed together, to get social reality. It emerges inherently from the collective patterns born of their interactions. Not really an earth-shaking thought, but one that many scientists have tried to suppress for many years.

Our hapless media...

The ever-active Glenn Greenwald cites yet another example of the sad failing of our media's "finest" to render poltical events in anything like a "fair and balanced" manner. As I tentatively suggested last week, our media figures could do a great service to our democracy -- not to mention to the standing of journalism itself --if they tried a bit harder to act as intelligent and independent minds in their own right, and to peer behind the themes and images given them by political figures and operatives, so as to help Americans know and understand the world, insofar as that might be possible. The Washington Post's Anne Kornblut clearly hasn't yet mastered the requisite skills, relatively simple though they may be. An excerpt from Tuesday night's Hardball:

KORNBLUT: It remains, especially in Democratic crowds, the number-one issue. There is no applause line that gets a bigger response when you're out with Senator Hillary Clinton, Barack Obama, John Edwards, than when they say the first thing I'm going to do is I'm going to start ending this war in Iraq.

Republican crowds are a little different. They still want to be supporting the troops.

Notice the logic, or rather the lack thereof. She might as well have said "Republicans are a little different, they still hold America in high esteem and do not want to dishonor the dead." Or, "Democrats want to start ending this war, but Republicans are a little different - they love their children and see hard work and responsibility as a good thing, not something to be ashamed of." Etc.

This is indeed a very crafty way to use, or misuse, language. "Republican crowds are a little different. They still want ..." sets us up to take whatever comes next as roughly the opposite of what Democrats want. The speaker then has free reign to say anything, knowing that the association is implicitly made.

I find it very hard to believe this is mere incompetence. Okay, speaking on the fly on television isn't easy, but a "political analyst" with any skill and experience ought to be able to avoid a juvenile error like this. It ought to be almost automatic; second nature. I cannot imagine myself, in a context more close to my own knowledge base, saying something like "Classical physics attibutes a precise position and velocity to a particle, such as an electron. Quantum physics is a little different. It views the world as something that can be explained, often in terms of elegant mathematical theory."

Indeed, I found it quite difficult even to construct that example, as my brain strongly demanded that I follow "It views..." with some statement about whether quantum theory attributes precise positions and velocities to particles. But someone reading could easily come away thinking that "classical physics" doesn't see the world as something that can be explained.

Alas, however, this kind of mushy and slippery thinking and speaking routinely passes as "analysis." Why even bother educating journalists? Greenwald summarized the point quite clearly, referring to the potentially useful role of "analysts":

One of the principal functions of political reporters ought to be to dissect and dispense with misleading political sloganeering, but instead, they fulfill the opposite function: they are the most enthusiastic and effective disseminators of these cliches.

...Instead, this sort of cheap sloganeering entirely engulfs the analysis itself and becomes the only way many political journalists can think and talk about political issues.

Rational or irrational?

A few weeks ago, Dr James B. Burnham of the Donahue Graduate School of Business at Duquesne University sent me a polite email chiding me for my discussion, in The Social Atom, of the work of economist Gary Becker. Becker is famous for his association with the so-called Rational Choice school of economic thinking, which tries to model a great deal of human behaviour -- including things like crime, individuals' choices to have children and so on -- in terms of people making fully rational choices. Becker's Nobel Prize Lecture from 1992 offers a nice synopsis of the principle ideas and some of their applications.

I portrayed Becker as being rather strongly committed to the rationalistic viewpoint, and Burnham quite rightly points out that I overlooked a classic work of Becker's, Irrational Behavior and Economic Theory, published in the Journal of Political Economy, February 1962. In the paper, it seems, Becker tried to establish some of the central "theorems" of classical economics without insisting on full individual rationality. Dr. Burnham was also kind enough to send me a copy of the paper, and I'm in the process of reading it. Hopefully I'll be able to comment tomorrow on just how fairly or unfairly I treated him.

But I came across something today, not quite as old as the overlooked Becker paper, that I suspect may shed some light on the question of whether people are rational or not, and when economic rationality can be assumed without fear of getting things rather completely wrong. The paper in question is a short review from last year by economists Ernst Fehr and Colin Camerer, looking at the question of how the collective logic of a situation can strongly influence human behaviour -- sometimes even making rational people act "irrationally," with important consequences.

Contrast the following two situations:

1. Suppose you'd like to try to win some money by playing the University of Iowa Presidential Election Market for the 2008 election. The idea is that you can buy a contract that will pay off if, when the election takes place, the numbers turn out a certain way. The contract you buy should reflect what you think is really going to happen, and accuracy will make it pay out more. Suppose you're an extremely well informed political analyst, more knowledgeable and with access to more information than just about anyone. Then you should act rationally, making your best guess. If everyone else playing the market goes half mad and starts buying futures for some completely implausible candidate, so be it; you'd have no incentive to follow. The more crazies out there, the more your cool wisdom will pay off.

2. Suppose instead that you decide to play the stock market. Again, being a bright and well informed individual, you scour the financial pages and identify one stock that you're sure is undervalued. Maybe it's IBM. You know they're in healthy condition, with diversified interests and good leadership, even though many irrational investors, lured by sexier stocks in smaller companies, have ignored IBM and pushed its value down unreasonably. So what do you do? Apparently, it seems, you could follow a similar strategy to the example above. Trust your own superior knowledge, buy up lots of IBM stock, and then sit back and wait for the price to return to it proper, reasonable value. Then you can sell it for a killing, making easy money.

The problem, as Fehr and Camerer point out, is that these two situations, though superficially similar, are actually totally different.

In the first case, other people being irrational gives you no incentive to follow along. The collective logic -- and the fact the election will take place on a specific date, at which point all contracts will be settled -- means that rational strategies work best. But in the second case, there's no guarantee. You may in all sound wisdom buy up the undervalued IBM stock, and then find over coming weeks and months that the irrational mob only acts to devalue it more. The herd may push its price down to ridiculous levels and leave it there for years; keeping your money in the dumps while it could have been earning interest elsewhere. In the latter situation, the rational can easily be forced to follow along with the irrational herd.

Fehr and Camerer argue that these examples illustrate a general pattern:

Individuals who violate the assumptions of economics may create powerful economic incentives for Economic Man to change his behavior...

This is indeed one of the major shortomings of any theory that assumes rational behavior; rational action, in a sense, is often just not possible, and people have to rely on other means for making decisions. Other people acting irrationally -- stampeding across a bridge, for example -- may quickly make stampeding across a bridge the only wise thing to do. What happens in the aggregate of many people is often in no way simply a summation of what they would have desired a few moments ago.

What, if anything, this might have to do with the attempts of Gary Becker to relax the assumptions behind theorems of classical economics, I don't know. I suspect there may be some holes in his reasoning, and they may center on whether the actions of individuals are independent, or whether what one person does has consequences for what others do. But we'll see...

Rove's legacy

I couldn't agree more with this assessment of the legacy of Karl Rove by Juan Cole. If there was ever a time when the rift between the left and right in the US could have been temporarily healed, and our collective energy harnessed for genuine improvement of our nation and the world, it was in the immediate aftermath of 9/11. Instead, Rove casually sacrificed that historic opportunity on the alter of momentary political gain. As Cole puts it,

... his most tragic legacy lay in taking something that happened to all Americans, the murderous attacks of Sept. 11, and attempting to turn those calamities into a stick with which to beat his Democratic opponents. In so doing, he desecrated the nearly 3,000 dead for petty factional gains, and wrought enormous injustices on genuine war heroes such as Max Cleland, George McGovern and John Kerry. Long after his permanent Republican majority is forgotten, Rove will be remembered for using his rhetorical gifts to divide instead of unite.

Rove's tactics were deplorable, damaging and, in my view at least, thoroughly un-American. One thing you cannot call them, however, is novel. He's clearly a fine student of history, and the historical lessons of dividing a people against themselves, especially in times of apparent danger when they are most easily manipulated. As Ian Welsh pointed out a while ago, Rove's insights have been used before, even more effectively, and with far greater consequences...

...voice or no voice, the people can always be brought to the bidding of the leaders. That is easy. All you have to do is tell them they are being attacked and denounce the pacifists for lack of patriotism and exposing the country to danger. It works the same way in any country."- Herman Goering

Torturing light...

This post really has little to do with The Social Atom, but I've had no time to blog anything today, as I've been occupied with a last-minute article on some recent physics research. Physicists are getting ever closer to building material systems that will coax photons, the quantum particles of light, to interact with one another. Ordinarily, that never happens; flashlight beams, for example, just pass through one another. But it turns out that with some really tricky quantum optics, it should be possible to effect (which quantum theory says is possible).

You may ask -- why even try? I'll get to that.

The bizarre things researchers are developing to achieve this illustrate the way science leads in directions that no one could possibly imagine beforehand; truth really is stranger than fiction. One prominent proposal, for example, involves lasers illuminating thin films of diamond, within which specially-chosen atoms have been embedded amidst a precise pattern of tiny, drilled microholes. Without getting into the details, the quantum physics of this weird system leads is an array of cavities which can trap single photons. With the cavities arranged in precise geometric arrays, the photons in neighboring cavities will interact fairly strongly with one another, and so what you get is a system akin to a collection of interacting photons. It hasn't been done yet in the lab, but the principles are so well understood that it almost certainly will be done in a few years.

Now back to the why. Why do this? There are a couple of reasons, besides it's being just good fun and a challenge leading to new physics (which I think is probably the most important reason, by the way). First, there's good reason for expecting that interacting photons would be exceedingly useful for doing quantum information processing; i.e. as parts of quantum computers. But well before that, physicists also expect they'll be able to use these photon systems as "quantum simulators" to do virtual experiments mimicking just about any kind of particle system you like.

When we think of simulation we ordinarily think of simulation by computer; but if you think about it, simulation really has nothing to do with computers. It really means using something that is simpler or faster or easier to use to study the behavior of something else, under conditions when you have confidence that the two things should work similarly, so you can learn about the one from the other. Theorists have shown that interacting photon systems could be tuned to dial up just about any physics you like -- any details about the interactions between particles -- and so could be used to do virtual experiments for all kinds of hypotheticl materials, even ones that don't exist.

Curiously, I just came across a review in Nature, written by my former colleague Phil Ball, of two new books on simulation in the social sciences. As the review notes,

..these two books are part of an important trend in the social sciences. Both argue for the value of agent-based modelling (ABM) in social science. This approach involves "growing societies from the bottom up", as Epstein has put it, rather than devising analytically airtight theorems from first principles that are tractable but transparently wrong in what they assume and imply about human behaviour.

The aim of ABM is to study whether the macroscopic patterns or regularities that we observe in society, such as price equilibria or the appearance of behavioural norms, can be generated from decentralized, local interactions between collections of agents.... Agent-based models may not describe reality, but they can show how interaction and nonlinearity produce social outcomes that could not be predicted simply by inspecting the behavioural rules.

Of course, all this is simulation, in this case using computers, which help us go beyond what our human brains can foresee in situations in which many factors begin working off one another. For the moment, quantum simulators are tuned specifically to atomic systems, and I'm not sure how far they might be generalized in the future. I can't imagine that quantum simulators in thin diamond films will ever be used to shed light on important social questions, by modeling social interactions, but then, who knows -- and why not?

The higher (or lower?) language

It's quite clear by now that much of our national discourse on topics ranging from gay marriage to Iraq to global warming takes place not in terms of facts but in the more powerful language of emotional associations, in terms of ideology. Framing counts for more than facts (as Republicans seem to understand better than Democrats, for some reason), in part because of the power of modern media to spread simple messages and sound bites far and wide, and their inability or unwillingness to act as honest interpreters of framing strategy.

In a New York Times essay that I never got around to writing, I had intended to argue that the proper role for the media in today's world ought to be as "censors" of a sort, or exposers of framing. For example, rather than merely providing a forum and promulgating Republican claims that war opponents "hate America" or are "helping the enemy," the press ought to provide independent assessment and comment on how such attacks aim to frame the war debate in a way favorable to Republicans. And vice-versa for Democrats. When Hillary Clinton attacks Obama for saying nuclear weapons should be off the table in pursuing terrorists, the press shouldn't merely repeat her words that "Presidents since the Cold War have used nuclear deterrents to keep the peace, and I don't believe any president should make blanket statements with the regard to use or non-use." Comment ought to focus on the obvious Clinton strategy of trying to paint Obama as being "unpresidential" and inexperienced. Reporting (sadly quite a lot of it) that quotes political figures without commenting on the frames they're attempting to establish really is mere stenography.

If you have a spare hour, I highly recommend this podcast of an entertaining and deeply informative lecture at Xerox PARC by Geoffrey Nunberg of the University of California, Berkeley. Entitled The Paradox of Political Language, this is the abstract:

There's a paradox in modern attitudes about political language. Left and right may disagree as to which expressions count as deceptive packaging and which are merely effective branding, but both sides acknowledge that the American public is particularly susceptible to linguistic manipulation. Yet it's also fair to say that there has never been an age that was so wary of the mischief that language can work or so alert to the dangers of political euphemism and indirection. How did we come to this point? Are political and public figures really more mendacious than they used to be, or does it reflect a changing media role or an increasingly polarized political climate? Why is widespread sophistication no impediment to the misleading use of language, and why do many of the most successful linguistic maneuvers pass our radar undetected?

Actually, I hadn't really intended to post on this subject, but it came to mind this morning when reading a spirited opinion piece by Nicolas Kristof in the New York Times. If you don't think framing and metaphor can help alter the terms of a debate, read this:

If we learned that Al Qaeda was secretly developing a new terrorist technique that could disrupt water supplies around the globe, force tens of millions from their homes and potentially endanger our entire planet, we would be aroused into a frenzy and deploy every possible asset to neutralize the threat.

Yet that is precisely the threat that we’re creating ourselves, with our greenhouse gases. While there is still much uncertainty about the severity of the consequences, a series of new studies indicate that we’re cooking our favorite planet more quickly than experts had expected...

That's effective framing.

Sex by numbers

Basic biology in many species pushes males and females toward different reproductive strategies, with males in some cases tending to seek as many partners as possible, and females -- because they invest so much more physically in producing offspring, and could even die in the process -- seeking one stable partner. The folklore in humans is that men tend to be more promiscuous than women, and its the biology that makes them do it.

In surveys, sure enough, this is what you find -- men report having, on average, more sexual partners than women. In one fairly recent study, sociologist Frederick Liljeros and colleagues found this pattern clearly among men and women in Sweden (they also found that the network of sexual contacts is highly skewed, with a few men and women having far more partners than most others). But this light article in the New York Times raises a fair question. Excluding homosexual contacts, the total number of times that females have sex has to be identical to the number of times males have sex; hence, the average number of sexual contacts for men and women has to be equal, simply by mathematics.

Liljeros et al. noted this in the paper I mentioned, and suggested that it might have something to do with males over-reporting their numbers, "bragging" about their impressive success. But I suspect there is probably also significant female under-reporting, especially as traditional norms in most societies take a dim view of female promiscuity. (One point the article doesn't explore is whether men and woman might interpret the word "sex" differently, leading to some encounters being viewed as sex by one partner, and non-sex by the other...but I'm guessing careful studies control for that by using some strict and unambiguous definition).

This systematic deviation of what people report from the underlying reality probably tells us something about mating strategies themselves; perhaps, for example, that females have a greater incentive than males to keep their promiscuity hidden. I came across an interesting essay on the topic, not by a biologist, oddly enough, but by Eric Raymond, a fellow linked in some deep way to the Open Source movement. No matter, it's a nice essay anyway, and informative. As Raymond argues:

Actual paternity/maternity-marker studies in urban populations done under guarantees that one's spouse and others won't see the results have found that the percentage of adulterous children born to married women with ready access to other men can be startlingly high, often in the 25% to 45% range. In most cases, the father has no idea and the mother, in the nature of things, was unsure before the assay.

These statistics cry out for explanation -- and it turns out women do have an evolutionary incentive to screw around. The light began to dawn during studies of chimpanzee populations. Female chimps who spurn low-status bachelor males from their own band are much more willing to have sex with low-status bachelor males from other bands.

That turned out to be the critical clue. There may be other incentives we don't understand, but it turns out that women genetically "want" both to keep an alpha male faithful and to capture maximum genetic variation in their offspring. Maximum genetic variation increases the chance that some offspring will survive the vicissitudes of rapidly-changing environmental stresses, of which a notably important one is co-evolving parasites and pathogens.

So maybe neither sex is the "more" promiscuous.