Wednesday, July 16, 2008

Educational Policy and the Extended Mind

Here’s an apparent inconsistency: it seems that, given a century of psychometric study, that there are innate factors to intellectual performance that are rigid vis-à-vis enviornmental inputs and settings. No matter what you do or where you put them, some students will predictably excell at school, and others will fail. Better teachers, smaller classrooms, expenditure per pupil, a healthy diet from pre-natal development onwards, a nurturing home enviornment—these all have important effects, but even considered collectively, those effects are bound by what seems to be innate, even genetic capacities.

On the other hand, it is undeniable that we are all a lot smarter than our counter-parts a hundred years ago. This is not only true in terms of literacy rates, basic mathematical competence, graduation rates and college attendance. Even our IQ’s have been improving (by what is known as the Flynn effect).

The first observation suggests that we really ought to be quite a bit less ambitious when it comes to public education policy, lest we waste a lot of resources and energy for negligible marginal benefits, incurring high opportunity costs. Prudent public education policy would replace the goal of making everyone smart (a pie-in-the-sky or ‘romantic’ view), and orient itself towards finding ways to make the incorrigibly dim nonetheless productive workers. On the other hand, the second observation suggests that there are in fact important and measurable returns to investment in public education—witness the fact that most Americans can now read, and know enough arithmetic at least to fill out their tax forms.

Let me call these respective positions ‘What’s the Point?’ (WTP) and ‘Yes We Can’ (SSP). WTP often responds to SSP in the following way: yes, certain metrics like literacy, basic mathematical competence, graduation rates, even a base-line IQ, have improved over the centuries, most notably the past one. But this achievement has been merely to allow the full exploitation of a natural capacity, and we are fast approaching the time when marginal returns on education investment fast diminish. In other words, whereas some prudent social policies have enabled increasing numbers of citizens to achieve their natural potential, we have not affected that natural potential itself, and once we reach it, there is not much more that policy will effect. Furthermore, for many students, we have already past whatever natural potential they have, and are now expecting results that simply are not achievable.

WTPers are fond of an analogy between innate mental capacities and innate physical capacities. Take running. Just about every human being can run, and some can run faster and farther than others. Surely some of that is due to training, diet, confidence, dedication--but in the end, a defnite limit is reached, and an innate distribution of ability becomes evident. Thus, (so the WTP argument runs) it is just as much folly to expect every child to learn calculus and to quote Shakespeare as it is to expect every child to run a 7 second 100m or a 4 minute mile.

But let’s consider this analogy a little further. Observe that ‘innate’ capacities, like running, operate within what are in effect artificial constraints. To measure one’s ‘innate’ running ability, we require (for example) that aids like drug enhancers, bionic legs, superhero lung transplants, roller skates, and so on, are verboten. But we do allow for scientific nutrition regiments, super-tech training aids, the use of all sorts of biometric technology. Without any of these artificial constraints, the relevance of ‘innate’ ability becomes not only specious, but moot. With superhero lungs and bionic legs, who knows, maybe I could run a one-minute mile. The point being, if we refuse to abide by artificial constraints, ‘innate ability’ becomes not only irrelevant, but almost incoherent. (Consider this question: what is the ‘innate’ life-span? If the physicalists are correct, and brain transplants become one day possible, and new bodies can be grown ‘brave new world’ style, then….you see the point).

I wonder why we shouldn’t consider ‘innate’ intelligence along the same lines. The equivalent of Olympic-criteria for measuring intellectual performance in the United States is the National Assessment of Educational Progress (NAEP). Like the olympic sports, the NAEP sets up artificial constraints on the measurement of intelligence. The NAEP assesses skills in reading, math, science, writing, history, civics and geography. Verboten for students taking the NAEP are instruments like calculators, spell check, wikipedia, maps, and so on. But why do we insist on these constraints? It is hard to imagine very many scenarios where anyone in a ‘real life’ situation would not be able to avail themselves of any one of these technologies, so what is it exactly that we are measuring, and why?

The what is a very prickly issue, and innatists will get upset if you seem at all puzzled about it. But I think I can partially answer the ‘why’. We are still wed to a fundamentally Cartesian, fundamentally classical understanding of intelligence. According to this model, to ‘know’ something is to be a certain state, rather than to possess a certain ability. Thus, to ‘know’ that 98 + 113.5 = 211.5, or that the slope of a curve equals ∆x/∆y, is to have an intuitive insight into the nature of number, or in the nature of m. But consider for a moment: why is it that today, almost any decently educated 5th grader will be able to determine that 98 + 113.5 = 211.5, and any decently educated 8th grader will be able to solve for a slope-intercept? Before the development of a base-10 arabic numeral system, it would have been difficult for almost anyone to solve for the first, and before Descartes, to solve for the second. I am suggesting of course that there is a strict analogy between the use of a calculator and the use of a base-10 numeral system. Both are artifical, yet testers for the NAEP consider one’s ability to use the first still somehow ‘innate,’ while the latter is ‘artificial’—indeed, cheating. Why not then, instead of trying to measure some suspect ‘innate’ faculty, we instead measure ability—not under no constraints, but under constraints that are plausible and ‘realistic.’

Overall, then, I am suggesting that ‘innate’ intelligence no longer makes a whole lot of sense—although it still makes some sense, just like running—once we accept an externalist, “extended” theory of mind. In other words, we should take to heart the theory developed by Clark and Chalmers in their famous paper and apply it to the debate over innatism and educational policy. Returning now to the issue between WTP and SSP, we can at least partially explain the discrepancy noted at the beginning by recognizing that, because of technologies like a base-10 numeral system, even someone of ‘average’ intelligence can now solve for problems that, half a milennia ago, only the most educated and ‘innately’ intelligent could solve. That is to say, tests like the NAEP are in fact somewhat anachronistic, and I am sure that, if we did for instance allow for the use of graphing calculators, and the internet, that we would see marked improvements in test scores and therefore ‘average intelligence.’

Some caveats: this makes most sense when applied to mathematics, and to a lesser degree, skills like geography and history. That’s because the gains from technology (including symbol-systems) demostrably extend by orders of magnitude cognitive capacity. It’s not clear what role any such technology plays in writing and reading. I have some thoughts on this issue, but I’ll save them for the comment section if any one cares to explore the issue further.

Secondly, and less directly, I’m still not convinced that even on the innatists own ground and under their conditions that there is anything obvoiusly being measured. This is because I suspect that ‘innate’ ability, to whatever extent the concept makes sense, is influenced as much if not more by factors such as focus, attention, and motivation as by any raw capacity. The problem might be fitfully compared to the issue of indeterminacy. Whatever it is one is measuring by standardized tests, it will remain inscrutable whether performance results from raw capacity or from motivation, and so far, we have no reliable way (as far as I know) of controlling for one or the other. To see just how this issue informs the debate, check out this discussion.


  1. Nice. One question, though: Can we distinguish innate vs external--at least to some extent--on the basis of constitutive features? For example, working with a base 10 numerical system is constitutive for solving some problems, though not others. The same is true for Cartesian geometry. Could we perhaps say that "innate" intelligence involves the ability to solve problems that are constituted by a system one has mastered? Well, that's a messy way of putting it, but I'm feeling lazy at the moment.

    By the way, the average increase in IQ is due almost entirely to a single feature: greater capacity for hypothetical thinking. This raises the entire issue of relativism in IQ testing: we live in a culture that prizes hypothetical thinking. Not surprisingly, such a culture treats the ability to think hypothetically as an important part of intelligence. Also not surprisingly, people immersed in this culture are likely to have better hypothetical thinking skills. It's likely that, for most features considered by a culture to be intelligence-bestowing, those features will be more marked in that culture than in those that don't care about these features.

    And another point: a LOT goes into the extended mind. For example, the aforementioned culture. I take it as a fairly clear implication of the "existence precedes essence" thesis that our cultural backgrounds are constitutive of our minds.

    About writing and reading: those ARE technologies. And new technologies are modifying them. The suggestion that internet reading, which involves flitting around, tends to undermine the ability to do protracted deep reading of a single piece, for example, gets discussed a lot. Some students might do better on reading tests if they're tested for internet-reading tasks rather than book-reading tasks. Which of these we consider to demonstrate "innate" intelligence is likely to depend on which we, as a culture, consider the more important ability.

  2. Call me a WTPer. Here are some quick responses:

    “The NAEP assesses skills in reading, math, science, writing, history, civics and geography. Verboten for students... are instruments…what is it exactly that we are measuring, and why? ”

    1) We are trying to get a measure that accurately predicts performance on certain tasks, such tasks as are often encountered in jobs and in academia. It has been found that the scores on all cognitive tests are highly correlated, so it doesn’t matter much what’s “verboten” in terms of predicting performance on the tasks that we are interested in. When we say that there are low marginal returns to education in terms of test scores, we usually mean that more education does not result in significantly higher test scores, ALL ELSE HELD THE SAME, including what’s “verboten.” That’s the definition of marginal.

    Think of it like this. Your IQ score, S, is just the sum of some underlying factor g and some other factor U: S=g+u. Changing what’s “verboten” just changes u, not g, but g is what determines on the job productivity, not u.

    2) There are lots of theories about what’s the exact nature of what’s being measured. This isn’t the place to get into them. But arguments that intelligence can’t have a biological (and by implication, to some extent, genetic) component are absurd. All reasonable people admit that savants and the mentally retarded are born with different mental capacities than the ordinary person. Hence, if your theory of the mind does not allow for different mental capacities arising from biological (and hence, to some extent, genetic) causes, so much the worse for your theory of the mind.

    “I’m still not convinced that even on the innatists own ground and under their conditions that there is anything obvoiusly being measured. This is because I suspect that ‘innate’ ability, to whatever extent the concept makes sense, is influenced as much if not more by factors such as focus, attention, and motivation as by any raw capacity.”

    You’re basically saying that since all we observe is S, how can we infer a g? The past 100 years of psychometric research has been an attempt to answer that question. And the evidence for underlying capacity/capacities is overwhelming. Positing underyling factors likes g generates certain hypotheses that have been overwhelming verified by reams of data and scores of studies. Here are a few:

    1) The significant positive correlation of all IQ tests scores. An underlying factor predicts this, whereas if changing what is “verboten” meant that people had to use a completely different set of skills, you wouldn’t expect the scores of different tests to be correlated.

    2) On any given IQ test, the high scorers tend to do better than the lower scorers on the same questions. If a test just measures a suite of abilities that are independent of each other, high scorers will average better on all the questions at about the same rate. However, if some questions require more cognitive effort than others, and high scorers have more cognitive capacity, we’d expect to see the high scorers do better on the “g-loaded” questions at a higher average rate, which is precisely what we see.

    3) Contrary to your hypothesis that the tests only measure focus, attention span, or motivation, many studies have tried to control for these factors by testing reaction speeds for carefully designed tasks that are meant to minimize the effects of non-cognitive factors as much as possible. These reaction time studies have shown that reaction time correlates highly with IQ, militating against your hypothesis.

    4) Twin studies suggest that IQ scores have a substantial genetic component.

    “Even our IQ’s have been improving (by what is known as the Flynn effect)
    … [this] observation suggests that there are in fact important and measurable returns to investment in public education—witness the fact that most Americans can now read, and know enough arithmetic at least to fill out their tax forms.”

    1) There’s strong evidence suggesting that the Flynn effect has been slowing down or may have stopped some time ago (indeed, some studies even suggest that it may have reversed. Would you admit that this evidence against investment in public education?

    2) There’s no reason to think that all—or even most—of the Flynn effect is attributable to increases in public education. Indeed, there’s some evidence against this since the IQ’s of older people actually increased at about the same time as the IQ’s of younger people (the education hypothesis would posit that the effect would be concentrated among the younger, more educated cohorts):
    3) Efforts to find correlations between per pupil expenditure—as well as a host of other things related to education—and myriad measures of school performance, including IQ tests, have failed to turn up anything significant. If you don’t believe me, consider that Heckman (a Nobel Prize winning economist) was a fierce critic of the Bell Curve when it came out, but in a recent paper has more or less conceded much of what the Bell Curve argued regarding the inability of interventions in the education system to make much of a difference:

    If you are simply claiming that in different environments, cognitive performance might be different than what we observe now, you should realize that almost no one disagrees with you. The claim that you appear to bristle at—but that nonetheless is backed up by mountains and mountains of data—is that no known environmental changes substantially alter IQ performance and/or other indicators of productivity. You appear not to realize that you are responding to empirical evidence about the relative magnitudes of certain changes (e.g. education spending) on measured quantities (e.g. IQ scores/productivity) by saying that you think there might be other changes that might have different effects, and therefore the cited evidence is irrelevant. This is clearly a non sequitur.

    I'll have more later,

  3. Yo Bro, so, I'll try to show with the points below what I think I know...but first, mentally adjust the enumeration so that it's points 1-9 rather than 1,2, 1234, etc....

    So, to 1) Maybe I'm wrong, but the NAEP itself doesn't purport to measure anything like g; it's just a test, a snapshot of where students are at grades 4, 8 and 12. But anyway, my general point in the passage to which you allude is just that the decision to prohibit calculators and google is more arbitrary than we tend to suppose.

    to 2) I think this point really does get the heart of my contention: If you give me a calculator, I can perform any sum almost as fast as a savant. We will be both be equally reliable, and equally fast. Now, my point in bringing in the 'extended mind' thesis is just the following: whose smarter? Our (Cartesian) assumptions would say that the savant is smarter because his 'calculator' is made of neurons and is inside his cranium, whereas mine is made of copper and silicon and is outside my cranium. But if we can do the same sums, at equal speeds, with equal accuracy, aren't we equally smart? I suggest, yes. And yet, on the NAEP the savant will be allowed his neural calculator, but I will be prohibited from use my copper one, so our equal intelligence will not be manifest on the test.

    So, is there a biological basis to intelligence? I guess what I'm trying to say is that this question is a little incoherent, and is a bit like asking, 'is there a biological basis to speed?' By examining human anatomy and our animal kin, you'd never guess that humans can go Mach 5, but we can....

    to 3-6) Sort of. I am saying that we observe 'S,' and S can equally well be explained by appeal to 'g,' or to the non-g factors I mentioned. Whether to 'g' or 'non-g,' S would be the same. But I wasn't suggesting that these 'non-g' factors are not biologically based. Indeed, it is easier for me to see how they are, as opposed to 'g.' The issue I really have in mind here is one I've seen in several places regarding the discussion of innate cognitive differences between the sexes, as in the Pinker/Spelke debate. Both Pinker and Spelke for instance acknowledge that the differences could be due to innate cognitive differences, or innate motivational differences, and they don't come to any agreement. Finally, the reaction-time tests, while interesting, are not testing 'motivation' as Pinker, Spelke and I would understand it.

    Again, maybe I should stress at this point that, while I have my problems with 'g' and IQ innatism, I don't think that the post above directly addresses those issues. Rather, I am arguing that, with the extended mind thesis, 'g', whatever its basis or status, is irrelevant. Again, it's as if one were claiming that knowing all about human musculature were relevant to answer the question, 'how fast can humans go?' There's no doubt a biological basis to the former, but it's pretty irrelevant to the latter.

    to 7) I don't dispute this

    to 8) I didn't meant to suggest that the Flynn effect was due to public education. I in any case quite doubt it. More likely due to nutrition than that. My point is that we are all getting a lot smarter, just meaning the median human can do a lot more with her brain today than her counterpart a hundred, and especially 400 years ago.

    to 9) I'm sure that we waste a lot of money in public education, but we waste a lot of money everywhere. But how else to explain that we are all smarter now? And is public education really incidental to that? I highly doubt it.

    To the final comment: I refer to my clarification for 3-6: I don't mean to be taking any position on the status of IQ, only on its relevance.

  4. So what can a smart guy do with a calculator compared to what a dumb guy can do with a calculator?

  5. carldyke,

    a very smart guy without knowledge of the base-10 arabic numeral system would probably find it very difficult to solve any modestly complex algebraic system, while a dumb guy with a calculator could easily find the slope of a given tangent line. Similarly, if you were ever to wonder how long it would take for a bowling ball dropped off the empire state building to hit the ground, you'd quickly turn to the theory of gravity, rather than running off to some temple priest with an offering. So, whose smarter? My point in the post is just that, when we consider the extended mind thesis, 'smarter' and 'dumber' become at best relative (relative to technology) and at worst simply vague concepts.