A Wall Street Journal article by Harvard professor of government Harvey Mansfield claims that the social sciences and humanities are inferior to the sciences.  The article implicitly urges undergraduates to major in science.  From the article:

 

“Science has knowledge of fact, and this makes it rigorous and hard.”

 

“Others try to imitate the sciences and call themselves ‘social scientists.’ The best imitators of scientists are the economists. Among social scientists they rank highest in rigor, which means in mathematics... Just as Gender Studies taints the whole university with its sexless fantasies, so economists infect their neighbors with the imitation science they peddle. (Game theorists, I'm talking about you.)”

 

Do you agree with this?  As a game theorist I probably have a rather biased view of the situation.  It's certainly true that the ideal of the scientific method is vastly better than the practice of economists, but I think that majoring in economics provides better training for a rationalist than majoring in any of the sciences does.

Economics explicitly considers what it means to be rational.  Although it infrequently considers ways in which humans are irrational, I'm under the impression that the hard sciences never do this.  Furthermore, because economists can almost never perform replicable experiments we have to rely on what everyone in the profession recognizes as messy data; therefore we’re far more equipped than hard scientists to understand the limits of using statistical inference to draw conclusions from real world situations.  Although I have seen no data on this, I bet that a claim by nutritionists that they have found a strong causal link between some X and heart disease would be treated with far more skepticism by the average economist than the average hard scientist.

New Comment
40 comments, sorted by Click to highlight new comments since:
[-]knb220

I confess that I may be biased, but....

Why do people think that Economics is more scientific than Psychology?!?

I suspect that what happens is people realize that economists use a lot of math, and assume because of that, they are "scientific" (science = math stuff). Psychologists actually perform scientific experiments and disprove hypotheses. Admittedly Psychologists are less numerate than Economists, but math skills != science.

If "science" is held to be adherence to the scientific method, economics is mostly non-scientific. Most economists don't perform experiments, but rather apply assumed axioms. There are experimental economists, but the field is rather new, and basically a subfield of experimental psychology.

It's certainly true that the ideal of the scientific method is vastly better than the practice of economists, but I think that majoring in economics provides better training for a rationalist than majoring in any of the sciences does.

I disagree. Insofar as their body of expertise is sound, economists are able to mount sound criticism against ideas that contradict the well-established knowledge in their profession, but this is not different from any other area. I don't observe that they are exceptionally rational when it comes to questions that can't be readily answered using their standard knowledge and methods.

What's more, when it comes to questioning the soundness and justification of ideas that are widely accepted within their profession, economists are definitely far worse than natural scientists -- especially when it comes to ideas that they use to support their own ideological positions. When you ask physicists how they justify their physics, you get direct, detailed, and sound answers that go to the heart of the matter. In contrast, on many topics, economists are apt to react with a smug, dismissive, and stonewalling attitude instead. (With depressingly few honorable exceptions.)

Now, it does happen that there are a great many unsound ideas that sound nice and attractive to economically illiterate general public and intellectuals (often including smart natural scientists), and an economist is likely to be capable of discussing these much more rationally. On the other hand, there are many matters of widespread agreement or even near-consensus in modern economics that would not stand up to rational scrutiny, which are sometimes little more than a sophisticated fig-leaf for ideology. In these matters, various fringe low-status contrarians and even naive folk economics can be much closer to reality, however much credentialed economists tend to scoff at them. Clearly, a training in economics may well cause one to become less rational on such topics.

On the other hand, there are many matters of widespread agreement or even near-consensus in modern economics that would not stand up to rational scrutiny, which are sometimes little more than a sophisticated fig-leaf for ideology.

Examples would help (if they don't contain much mind-killing potency).

One example are discussions of free trade, which is strongly favored by an overwhelming majority of economists. The standard justifications for this position are based on models whose connection to reality is tenuous at best, and which ignore a host of potentially relevant factors. (Also, I often see arguments that are clearly unsound even if the models used are assumed to be adequate, as well as naive intuitive arguments that violate elementary logic.) Now, for all I know, the majority opinion of economists on this issue could be mostly correct (though I am sure that it's wrong on some particulars), but whichever way it is, their existing justifications are far from adequate. Once you start questioning them about this, you are likely to quickly run into sneering, stonewalling, moral posturing, and what looks like ideologically induced blockheadedness.

Another example is the question of how meaningful various economic figures and quantities are. This includes various economic statistics that are often used in arguments and calculations with multiple digits of precision, even though the way they have been obtained clearly makes even the first digit, and sometimes even the order of magnitude, highly questionable. (Oskar Morgenstern's On the Accuracy of Economic Observations presents a good book-length critique of this phenomenon. This book has been conveniently ignored and forgotten by economists, even though the issues it raised have never been addressed, to my knowledge.)

Even worse, however, are various artificial quantities such as price indexes and the "real" figures based on them, which are typically reified and treated as if they were objectively measurable properties of the real world, whereas in reality they are arbitrary constructs that could be defined with as much (or rather as little) justification in different ways to yield wholly different figures. Again, with a few honorable exceptions, trying to discuss these issues with economists usually leads to frustrating unsuccessful attempts to explain what the problem is, often followed by smug dismissals. You'll practically never see anything like that from natural scientists.

I generally agree with you; however, some comments:

Even worse, however, are various artificial quantities such as price indexes and the "real" figures based on them, which are typically reified and treated as if they were objectively measurable properties of the real world, whereas in reality they are arbitrary constructs that could be defined with as much (or rather as little) justification in different ways to yield wholly different figures.

I'm sure you'll agree that long term economic numbers should be adjusted for inflation to be more meaningful, whether this is happening in a reasonable way is another question.

You'll practically never see anything like that from natural scientists.

Well, in physics you have things like friction and air-resistance that, at least at first, were fudge factors to explain why objects don't obey Newton's laws. In fact looking at physics, in order to make progress it was necessary to disentangle the "fundamental behavior" of objects, i.e., how they behave on "frictionless surfaces", when hung from "strings of negligible mass" and other unrealistic assumptions, from things like friction that cause real objects to behave slightly differently. The same will undoubtedly be necessary in economics.

That having been said, I have no idea whether existing attempts to do this are any good. Physicists, of course, had the advantage that they could do experiments much more easily.

I'm sure you'll agree that long term economic numbers should be adjusted for inflation to be more meaningful, whether this is happening in a reasonable way is another question.

The problem is that "adjusting for inflation" makes sense only as a rough and vague heuristic, not as an exact measure of objective value which can be computed up to three, four, or more significant digits. The reality is that when you compare the purchasing power of money in different places and times, the differences in a myriad of relevant factors are often so great that it makes no sense at all to express them with a single number, except perhaps for the purposes of some extremely rough, Fermi problem-style calculation.

The idea that you can define some objective and scientific "price index" and then do calculations that will tell you that some "real" values differ by 4.83%, or invent models that will predict such figures by capturing real insight, is completely detached from reality. This is not a simplification of a problem to make attacking it easier; it's not even an instance of unjustified simplification -- rather, it means dreaming up complete fantasies and giving them a pseudoscientific dressing. I simply see nothing comparable among natural scientists, who are generally capable of dissolving the concepts they work with and avoiding getting lost in such elaborate fantasies built of reified artificial concepts.

(From this it also follows that most discussions of economic growth are nonsense on stilts, since they require combining the nonsensical notions of "real" values with GDP and similar figures, which have their own host of problems that normally go unacknowledged, and which are also usually impossible to discuss rationally.)

[-][anonymous]00

The reality is that when you compare the purchasing power of money in different places and times, the differences in a myriad of relevant factors are often so great that it makes no sense at all to express them with a single number

That is too sweeping a criticism. We have to model the world somehow when making decisions, and quantitatively. That goes for policy makers also.

Are you familiar with Jared Diamond's lay account and defense of operationalization, here? I'd be interested to hear what parts of it you object to.

That is too sweeping a criticism. We have to model the world somehow when making decisions, and quantitatively. That goes for policy makers also.

With all due respect, this is exactly the sort of stonewalling, discourse-killing response I commonly see from economists. If you can't justify your model, then the correct and honest thing to do is to admit that you don't have a model and try to deal as best you can with your ignorance -- not to continue using models divorced from reality with the excuse that you don't have anything better. An astrologer can use the same excuse, with equal justification.

However, clearly, admitting ignorance in these matters also means admitting that "scientific" policy is in many instances just pseudoscientific dilettantism, which in turn hits all sorts of ideological hot buttons. Of course, some centuries ago, a court astrologer would also likely stonewall if questioned too much about the validity of his astrology, which is an almost perfect analogy.

Are you familiar with Jared Diamond's lay account and defense of operationalization, here? I'd be interested to hear what parts of it you object to.

There is in principle no problem with operationalization, as long as the models based on it are sound logically and epistemologically. In particular, it is important that the operationalized concepts can be traced back to how exactly they correspond to observable reality, and that the limitations of models are understood and recognized by those who employ them. What I'm arguing is that many concepts used by economists don't satisfy these basic criteria of validity.

The crucial practical problem here is that attempts to tackle scientifically issues that are relevant for politics, power, and ideology are more likely to lead to the emergence of ideologically-driven pseudoscience than to a real scientific clarification of contentions issues. Clearly, modern economics and other social sciences have not been immune to this problem, which is probably the main reason for their present awful state.

[-][anonymous]30

There are two questions here. One is whether economists have a good influence on policy. Another is whether they even understand anything. I don't have any illusions about the first question, but my interest is in understanding.

We observe that prices tend to go up, and wish to quantify that observation, maybe to test ideas about its relationship to money supply. So we have to operationalize "prices go up." We do it by sampling commonly purchased things and keeping track of their prices over time. That seems to me to be as innocuous as operationalizations come in social science. If you don't think we learn anything by doing it, I don't see how you think we can learn anything about human affairs at all.

You might object that in the details the way inflation is measured and reported is very political. I don't disagree but that's not the same thing as saying we don't know anything.

The crucial practical problem here is that attempts to tackle scientifically issues that are relevant for politics, power, and ideology are more likely to lead to the emergence of ideologically-driven pseudoscience than to a real scientific clarification of contentions issues.

That's a problem for all domains of knowledge including the hard sciences.

We observe that prices tend to go up, and wish to quantify that observation, maybe to test ideas about its relationship to money supply. So we have to operationalize "prices go up." We do it by sampling commonly purchased things and keeping track of their prices over time. That seems to me to be as innocuous as operationalizations come in social science. If you don't think we learn anything by doing it, I don't see how you think we can learn anything about human affairs at all.

Well, yes, if you just want to operationalize the vague statement that "prices went up," it makes sense to do it as you describe. What doesn't make sense is postulating the existence of some reified "real" value of money, and then claiming that it has changed by 2.61% or whatever since last year. (Not to even mention the even more outlandish attempts to evaluate this change in "real" value across decades and centuries, or to compare them across vastly different places, where even the set of things available on the market is largely different, as well as all sorts of further reifications such as the "real" GDP.) However, vast edifices of both theory and practical policy have been built under the assumption that you can do such things, and if one attempts to discuss them by dissolving these nebulous concepts, one encounters confusion and stonewalling.

That's a problem for all domains of knowledge including the hard sciences.

The problem is far more severe where the questions studied have some bearing on ideology, politics, and power. It's a matter of incentives, after all.

[-][anonymous]30

What doesn't make sense is postulating the existence of some reified "real" value of money, and then claiming that it has changed by 2.61% or whatever since last year. (Not to even mention the even more outlandish attempts to evaluate this change in "real" value across decades and centuries, or to compare them across vastly different places, where even the set of things available on the market is largely different, as well as all sorts of further reifications such as the "real" GDP.)

I would like to hear a more detailed criticism of the "real value of money" concept specifically. Human height varies over the course of a lifetime, over the course of a day, and geographically, and yet it is fairly clear that people are taller on average now than during the middle ages. And by different operationalizations we can measure how much taller. Isn't it also clear that people are richer on average now that during the middle ages? Are you arguing that any attempt to measure how much richer is doomed to be misleading?

Statistics about human height are not a good analogy, since the concepts are simple and straightforward enough that they can be readily dissolved and their connection to reality re-evaluated if necessary. But many other "social science" numbers are indeed as bad as those found in economics, in the sense that even a casual rational evaluation of these numbers will reveal critical problems that are nonchalantly ignored in the regular "scientific" practice in these areas. (Though it would probably be hard to find anything as perverse as the epistemic rat's nest that economists have built around various numbers they operate with.)

As for measuring how much richer people are than in the past, you can provide information that will leave readers with more or less correct intuitions about such things, by describing at length how much people in various occupation had to work and in what conditions, what they could afford with typical wages, etc., etc. But the idea that you can describe this with a single scalar number and then treat this number as a real physical quantity that features in models and theories is obviously complete nonsense.

[-][anonymous]00

But the idea that you can describe this with a single scalar number and then treat this number as a real physical quantity that features in models and theories is obviously complete nonsense.

It's not obviously complete nonsense. It's obviously part nonsense and part sense, and without further argumentation it's not obvious how large a part nonsense and how large a part sense.

Just because a quantity is defined in a rather arbitrary way, that doesn't mean it carries no information. You can take the position that an informal analysis would incorporate the same information and do it better, but you'd have to actually compare the biases inherent in these two approaches. (I haven't read the Morgenstern book that you've linked; maybe it does so.)

[-][anonymous]00

There are two questions here. One is whether economists have a good influence on policy. Another is whether they even understand anything. I don't have any illusions about the first question, but my interest is in understanding.

We observe that prices tend to go up, and wish to quantify that observation, maybe to test ideas about its relationship to money supply. So we have to operationalize "prices go up." We do it by sampling commonly purchased things and keeping track of their prices over time. That seems to me to be as innocuous as operationalizations come in social science. If you don't think we learn anything by doing it, I don't see how you think we can learn anything about human affairs at all.

The crucial practical problem here is that attempts to tackle scientifically issues that are relevant for politics, power, and ideology are more likely to lead to the emergence of ideologically-driven pseudoscience than to a real scientific clarification of contentions issues.

That's a problem for all domains of knowledge including the hard sciences.

I simply see nothing comparable among natural scientists, who are generally capable of dissolving the concepts they work with and avoiding getting lost in such elaborate fantasies built of reified artificial concepts.

While that's mostly true today after several centuries of work, Newton would have been hard pressed to explain what a force was without resorting to something that sounds like mysticism. Also calculus was only placed on a firm mathematical footing by Weierstrass two centuries after Newton had invented it and based his physics on it.

That's not a good comparison. Every physical theory postulates the existence of some fundamental constituents of reality that cannot be reduced further. (And a practically useful theory may well treat that way concepts that we in principle know how to reduce to something more fundamental, but it would be impractical to do so.) The concepts in economics that I'm attacking are a completely different case: they consist of quantities defined in arbitrary ways, whose arbitrariness is then forgotten -- leading to their treatment as objective properties of reality that should be calculated and studied in their own right, far beyond the limits of their actual usefulness, and without the ability to work back through the logic of their use as a reality check.

I can't think of anything comparable in the history of modern physics, from Galileo till present day. I simply can't imagine a physicist similarly incapable of grasping with the logic of physical theories and their connection to reality.

Every physical theory postulates the existence of some fundamental constituents of reality that cannot be reduced further.

I've heard one that doesn't. It was agnostic on the subject and speculation on what the reason was for having indefinite layers of 'more fundamental' was outside the scope of the theory (like a 'first cause'). It did qualify as a physical theory (but certainly isn't mine.)

Honestly, this discussion is backwards (in addition to being mostly a form of tribal warfare). You don't study a subject in order to become more rational; you become more rational in order to more effectively study a subject.

It goes both ways, just like with e.g. lifting heavy things.

While that's generally true, I'd also point out that most people in lower-level academia aren't studying, say, sociology in order to become sociologists. This is even true to some extent for a lot of technical degrees, though moreso for some than others.

A student aiming for a field that has a lot to do with instrumental rationality and little to do with the stated aims of any common undergraduate degree would be well served by picking whatever undergrad degree does the best job of teaching instrumental rationality. Plenty of those fields exist -- like law and entrepreneurship, and to a lesser extent business, off the top of my head -- so I'd say it's a sane question to be asking.

Almost everything you learn when you study "hard sciences" as an undergraduate consists of already established results. How to solve problems that nobody has solved before isn't really part of the curriculum, as far as I can tell.

[-][anonymous]20

I don't think there is a general method for solving problems that nobody has solved before, but if you happen to know of one I'd love to hear it. :D

Speaking as an undergraduate physics student, the issue of already established results being the main focus of the curriculum is mainly solved through the inclusion of practical labs. These undergraduate labs are supposed to demonstrate that what we're learning works in practice and to illustrate specific phenomena that are counterintuitive. In practice, my freshman and sophomore labs didn't really do that; underclassmen physics labs are sometimes notorious for not working when inexperienced undergrads are working with old equipment.

However, in my junior lab, something that really helped was that we were working with better equipment, the experiments (for the most part) worked, and we were seeing not just that established results were being confirmed, but that the experiments (photoelectric effect, Ramsauer-Townsend, electron diffraction) were well explained by quantum models and (in some cases very) poorly explained by classical models. We were able to see why scientists had shifted from classical theory to quantum theory, which I gave major rationalist points.

The curriculum could overall be improved by something like the gallery of failed atomic models, but this would take a lot of time and would probably be best offered as a supplemental, elective course that gave in-major credit.

[This comment is no longer endorsed by its author]Reply

This seems to be bad news for the social sciences:

Unfortunately, as Perry and Mace point out, many social scientists appear to acquire biased and distorted beliefs concerning evolution during their training, as evidenced by their survey finding that rejection of evolution is significantly correlated with the number of years spent studying the social sciences.

Here's a link to the abstract of the survey paper itself.

Studying an academic discipline should not make you delusional and ignorant!

I don't think that the appropriate distinction is between social and natural sciences, so much as it is between verifiable and not verifiable.

In Physics, there are lots of counterintuitive results for a truthseeker to have to accept, so it probably helps someone become more rational.

In Literature, there's very little that lets you check what you're saying, and in some circles not even asking the author what they meant counts. So there's confusion to dissolve, but it doesn't seem to be common practice.

I lean towards thinking that Economics is good rationalist training because of the messy data, emotionally-charged issues that you're investigating, and the ability to come up with useful results despite that.

In Literature, there's very little that lets you check what you're saying, and in some circles not even asking the author what they meant counts. So there's confusion to dissolve, but it doesn't seem to be common practice.

Study of literature, especially older literature, is in my opinion indispensable for what is perhaps the most difficult act of overcoming bias, namely realization that there can exist people with views and values radically different from one's own who are not delusional, stupid, or malicious.

Do you mean actually reading older literature, or academic analysis of older literature? If the latter, a quick look at the Wikipedia entry for literary theory seems to show that the field is full of confusion:

...many contemporary theorists and literary scholars believe either that "literature" cannot be defined...

and many of the schools of thought listed there seem to be about writing a bottom line, and then making whatever work of literature (whatever that is) is being examined fit.

I meant actually reading older literature, but third-party analysis and commentary can be very helpful, as long as you're not taking their claims for granted. I certainly agree that there is a whole lot of nonsense produced in the humanities in general, so caveat lector.

Do you have any books in particular in mind?

Well, that depends on what exactly interests you. On the whole, for a modern English speaker, there is a huge wealth of 18th and 19th century literature available for free online, which is probably optimal because it's distant enough to offer interesting perspective, but at the same time highly readable and not so alien as to be incomprehensible without specialized study. On almost any subject that is a matter of ideological controversy today (or that was a matter of controversy in past ages), you can have a lot of fun by reading through random titles from past centuries that come out of Google Books when you search for the relevant terms.

In Physics, there are lots of counterintuitive results for a truthseeker to have to accept, so it probably helps someone become more rational.

In the same time, physics students usually accept this by authority, and only later learn all the details which constitute the evidence implying the counterintuitive theories, so this is a double-edged sword.

(The physicist training probably con't be much improved in that. If the students had to really see that e.g. quantum mechanics is really the best theory to explain the observed phenomena, they had to go through a lot of blind alleys that physicists examined between 1900 and 1930 - and it isn't without a reason that it took about thirty years to formulate QM in a coherent form.)

Smarter/more rational/more logical people self-select themselves into the math/hard sciences to begin with (but they also have peers who aren't likely to subscribe to certain irrational belief systems). I've never seen any real evidence that math/physics majors have ever become more rational as a result of their training (they're basically taught material as if it were completely inapplicable to the real world). Transfer generally doesn't seem to happen like that.

Personally, I believe that philosophy and economics are probably the best fields to train rationalists, although there are subsets of CS and statistics (Pearl's causal modelling) that are also excellent for training rationalists. That being said, much of the philosophy curriculum does consist of ancient/medieval bullshit, so while it can definitely help a rationalist think, its material isn't the best use of a rationalist's time.

What would also work, is being a CNS (or economics) major at Caltech, or a Brain & Behavioral Sciences major at MIT

[-]knb10

Personally, I believe that philosophy and economics are probably the best fields to train rationalists, although there are subsets of CS and statistics (Pearl's causal modelling) that are also excellent for training rationalists.

I think Psychology, especially heuristics & biases, cognitive psychology, and social psychology should be included.

That's certainly true, but those are not the mainstays of most psychology curricula (what also hurts, is the types of people who will be your peers). So unless you're pursuing it at MIT/Caltech, I'd steer away from a psychology major (unless you can go straight into upper-division psych courses, which you can't at my school)

Personally, I believe that philosophy and economics are probably the best fields to train rationalists, ...

Rationality training should not be based on disciplines full of confusion. Philosophy is practically defined as a set of questions (so far) so confusing that they cannot be properly addressed by some branch of science. Economics is closely tied to politics and has (because of many practical reasons) a very limited possibility for empirical testing. They may be good disciplines to work on for an already advanced rationalist - and there are aspects of rationality which one can't illustrate well on examples from physics - but one has better to master basic rationality before one comes to studying philosophy and economics.

I've never seen any real evidence that math/physics majors have ever become more rational as a result of their training.

Is there an evidence for other fields of study, namely for philosophy and economics?

Rationality training should not be based on disciplines full of confusion. Philosophy is practically defined as a set of questions (so far) so confusing that they cannot be properly addressed by some branch of science. Economics is closely tied to politics and has (because of many practical reasons) a very limited possibility for empirical testing. They may be good disciplines to work on for an already advanced rationalist - and there are aspects of rationality which one can't illustrate well on examples from physics - but one has better to master basic rationality before one comes to studying philosophy and economics.

Ah yes, that's definitely true. I was really only pointing out the fields that could provide the best training for those who were LessWrong users. Which doesn't take that much - it's a lot easier to detect the bullshit in philosophy/economics than to learn the rationality inherent within them.

Is there an evidence for other fields of study, namely for philosophy and economics?

Perhaps not. All I know is that a substantial number of rationalists end up pursuing examples from philosophy and economics

But in general, the best training for rationalists cannot come out of most college programs (outside of Caltech/MIT).

Well i know for instance that you get no statistics as an undergrad in the harder sciences (at least its not required), typically no research methodology classes , you also get help recognizing how your brain goes wrong (at least in psychology).

If you want to use the beliefs of the different degrees then the closer you get to the social sciences the higher the atheism. Psychology being the major with the highest amount of atheism. Another thing is that going through basically any undergrad school increases belief in paranormal things but grad school decreases beliefs. (if you want the studies i read i will have to try to remember what book i read that cited the studies and explained the results)

If you want to use the beliefs of the different degrees then the closer you get to the social sciences the higher the atheism. Psychology being the major with the highest amount of atheism. Another thing is that going through basically any undergrad school increases belief in paranormal things but grad school decreases beliefs. (if you want the studies i read i will have to try to remember what book i read that cited the studies and explained the results)

If you could do that, I would appreciate it. I'm pretty certain I've read the opposite with regards to rates of atheism by scientific discipline.

I could not remember the original book i read it in but i did a little research and here ,here agree with me, here seems to agree with the previous ones though mentions a theory that says that it is distance from the study of religion that predicts religiosity (that this study doesn't support), and this article seems to support the idea.

Reminds me i should write up a post about the supposed mental health benefit of religiosity.

If you want to use the beliefs of the different degrees then the closer you get to the social sciences the higher the atheism. Psychology being the major with the highest amount of atheism. Another thing is that going through basically any undergrad school increases belief in paranormal things but grad school decreases beliefs. (if you want the studies i read i will have to try to remember what book i read that cited the studies and explained the results)

Not necessarily true. Psychology has lots of atheists, but the other social sciences have far more religious people (especially compared with biology and physics). There was research on this linked on Wikipedia some time ago, but the Wikipedia article is gone now for some reason.

We need approaches for developing useful models of systems that are too complex to both model and control with complete fidelity. The divergent ways that, for example, biology, meteorology and economics have developed to accomplish this is all very useful to rationalists.

Suppose we wanted to study this question you have posed; what sciences would we look towards to suggest experimental approaches?