Why are "soft" sciences soft?

[Surlethe]

A while ago, we had a thread about defining science*. I postulated that hard sciences are chiefly quantitative, while soft sciences are chiefly qualitative. Now, I was thinking the other day: why are softer sciences, soft? There seems to be a category of sciences which are not historical and rely entirely on 'natural experiments'; they include psychology, sociology, and economics. Because they are not historical, it is possible to devise controlled experiments to test them, unlike history or archeology.

It occurred to me that the barrier which prevents greater progress in these sciences seems to be ethical. We cannot both morally and knowingly take random people and subject them to a battery of tests to map out what weird things happen to the human mind under stress, just as we cannot morally and knowingly mess with various parameters in the economy (what happens with 500% interest rates?) just to see what happens. Now, to be clear, I do not advocate removing ethical barriers. But at the same time, are these not major barriers to understanding? How far can they be pushed?

*Sorry, Alyrium, I didn't see your response before it dropped off the front page of SLAM.

[sketerpot]

A big part of it is that mathematical models often work so well in the hard sciences. Mathematically, we can describe the position of a particle with a complex-valued wave function. We can combine those, and then find the resulting probability density by squaring the magnitude of the resulting wave function. And it turns out that this is a really, really good model of reality for what I've been told are good, "obvious" mathematical reasons. When was the last time you saw something that elegant describing, say, psychology?

[Darth Wong]

A while ago, we had a thread about defining science*. I postulated that hard sciences are chiefly quantitative, while soft sciences are chiefly qualitative. Now, I was thinking the other day: why are softer sciences, soft? There seems to be a category of sciences which are not historical and rely entirely on 'natural experiments'; they include psychology, sociology, and economics. Because they are not historical, it is possible to devise controlled experiments to test them, unlike history or archeology.

That's half the problem. The soft sciences are soft not only because they lack experimental control (although that is obviously a major problem) but because historically, their practitioners have been far too prone to questionable reasoning, particularly of the a priori variety.

It occurred to me that the barrier which prevents greater progress in these sciences seems to be ethical. We cannot both morally and knowingly take random people and subject them to a battery of tests to map out what weird things happen to the human mind under stress, just as we cannot morally and knowingly mess with various parameters in the economy (what happens with 500% interest rates?) just to see what happens. Now, to be clear, I do not advocate removing ethical barriers. But at the same time, are these not major barriers to understanding? How far can they be pushed?

*Sorry, Alyrium, I didn't see your response before it dropped off the front page of SLAM.

Ethics do not seem to stop people from conducting experiments with the economy; consider Reagan's experiment with supply-side economics

[Zixinus]

I am not sure whether this is appropriate question, but here it is anyway: science is a four step process, right? Observe, hypothese, experiment till the hypothesis gets it right, theory. So, when you can't experiment, you observe and just gather data and work with those.

Is that how it works or am I missing something? Because I gather that's the way these "soft" sciences work, because of the ethical limits.

[Hawkwings]

Would astronomy be a soft science then? You can't experiment, so you just observe. But there are no real ethical problems in the way of astronomy.

[starslayer]

Would astronomy be a soft science then? You can't experiment, so you just observe. But there are no real ethical problems in the way of astronomy.

This is true, but I would still consider astronomy (really, astrophysics) to be a hard science. This is because the observations and models can be checked against already verified physical models and principles; i.e., explanations of astronomical phenomena must be backed up by physics, and they must make physical sense. This cannot be done in the soft sciences.

The ethical problems Surlethe is describing come from putting people in controlled conditions and then doing something to them, or wildly changing the economy and potentially destroying people's livelihoods. By their very nature, astronomy and astrophysics do not involve throwing anything into a controlled situation; as you noted, we can only observe from afar, or send probes to gather more data.

[Feil]

I think it's worth observing that the very softness of the soft sciences perpetuates their softness: hard sciences have high barriers to entry, requiring would-be practitioners to have advanced mathematical knowledge, logic skills, and experience with the scientific method. Soft sciences, on the other hand, require the practitioner to know how to read books, analyze paradigms, and write essays. Useful skills, but hardly ones that will weed out everyone except the people who will push the field in useful directions.

[Alyrium Denryle]

The reason the softer sciences are soft is for three primary reasons. The first is lack of experimental controls. We simply cant directly test a lot of psychology for example. We cant test what is going on in someone's head. We can only look at proxy measures. But the tests we devise cannot necessarily distinguish between two competing theories. For lack of a better analogy, they lack resolution. You can rule out some stuff (most of freud) but other stuff there is no way to falsify.

The second reason is that the fields are so fucking broad. Some branches of these fields are actually rather firm. Neuropsych, social psych, and Evolutionay Psych for example. These are pretty much on the level of the biological sciences. They produce firm easily testable models with high resolution. Hell, evolutionary psych unifies the neuro/biological psych people and social psych with biology. As far as I a concerned they are one large discipline.

Other subfields of just psychology though are about as soft as certain tasty french cheeses.

Sociology, explanatory criminology (as opposed to the observational stuff), things of that nature suffer from the same problem. They are over-broad.

The last problem is that it is very easy to build and reinforce an ideology with these sciences. Because they lack experimental resolution, it is very easy to fall victim to confirmation bias and interpret data how you want, in order to support your pre-existing notions. Committing a sort of reverse-naturalistic fallacy. An ought, becomes an is, rather than the other way around.

[Darth Wong]

Would astronomy be a soft science then? You can't experiment, so you just observe. But there are no real ethical problems in the way of astronomy.

You can conduct experiments in order to confirm astrophysical theories. You are falling prey to the creationist myth, in which any phenomenon must be tested directly. In reality, you need only test the underlying physics mechanism.

For example, you can conduct an experiment to confirm the Doppler shift mechanism, which is, in turn, used to analyze images from astronomy.

[Thanas]

I think it's worth observing that the very softness of the soft sciences perpetuates their softness: hard sciences have high barriers to entry, requiring would-be practitioners to have advanced mathematical knowledge, logic skills, and experience with the scientific method. Soft sciences, on the other hand, require the practitioner to know how to read books, analyze paradigms, and write essays. Useful skills, but hardly ones that will weed out everyone except the people who will push the field in useful directions.

That is generally true, however some soft sciences would most likely qualify as exceptions. For example, law schools have quite often high entry requirements and drop out rates.

Furthermore, some soft sciences have low entry requirements, others do not, for example ancient history and archeology. Several universities I know expect you to be fluent in two modern languages, and expect you to know a third modern + two ancient ones, both at least good enough to translate fluently. And I am not talking about "I took a course in High School once."

Obviously it depends more on the university instead of the general difficulty of the program, something that is different from the hard sciences. Generally, the better respected a university program is, the harder it will get. However, this is another thing that varies - I know a professor who once flunked over 95% of his students. Granted, this is somewhat unusual, but it happened.

The real challenge IMO in the soft sciences is when you get to the postgraduate programs - then the real weeding out begins.

[Guardsman Bass]

Do any of these have the potential to be "hard" sciences, if you had the resources and (possible) lack of ethics to pull it off?

[Thanas]

Do any of these have the potential to be "hard" sciences, if you had the resources and (possible) lack of ethics to pull it off?

^I am not sure I understand the question, presuming it is even directed at me. If not, please disregard the rambling below.


Can those programs be hard sciences? No, since as people mentioned above, you cannot check the conclusions by performing experiments, which is a defining characteristic of hard sciences. That is a luxury you do not have in the soft sciences where you have to convince people instead of doing the right equations.

Are those programs I mentioned close in dropout rates to hard sciences? Law is, with only 10-20% reaching the exam phase (and it has the worst grade average of every university department, with only about 10% of those being accepted to doctorate studies). Note that the law school also only accepts less than 5% of all applicants.

History, which doesn't really force you to quit once you get accepted, uses a different model. Out of about 800 students, about 30-50 manage to get accepted to the postgraduate programs. In ancient history, less than two dozen attend the special program for advanced students every year. Obviously, almost everyone who gets accepted into the general program passes them, but it is the advanced programs that really make a difference.

Are they as challenging as hard sciences? I do not know. I can only relate personal experience here. I know I generally spent about 70-80 hours a week working for the university during my first 7 semesters. But a lot of this was personal initiative. My course load was something about 38 hours per semester, though I am hardly the best model for the course work since I practically studied two subjects at once. A general student will get about 26 hours of course work per semester. Which is far less than the "hard sciences", but he will of course have to work extra hours in order to learn languages, homework assignments (usually 3-4 20ish pages per semester in addition to presentations, talks and discussion panels) etc. I know people who have quit hard sciences and studied soft ones because they thought hard sciences were too tough and vice versa (though the latter are in the minority). In general, I wager it depends on the type of person you are and what works best for you. Some professors take the time and effort to weed out bad students, others do not, for varying reaons ranging from being overworked to plain laziness or ideology.

[Guardsman Bass]

What I meant was, is there potential for hard science style experiments to be set up in some of these fields if you had a lot of resources and virtually no qualms about using and abusing human beings to find out your ends (i.e., assuming that you can effectively use a certain number of people as "slaves" in your experiments)?

I recognize that it some of these, it's probably impossible anyways - there's no good way to set up a control.

[Alyrium Denryle]

What I meant was, is there potential for hard science style experiments to be set up in some of these fields if you had a lot of resources and virtually no qualms about using and abusing human beings to find out your ends (i.e., assuming that you can effectively use a certain number of people as "slaves" in your experiments)?

I recognize that it some of these, it's probably impossible anyways - there's no good way to set up a control.

Yes. And if you separate out sub-disciplines some of them already are

I love it when I answer a question... and then have to repeat myself*

*Note: I am a bit testy tonight....

[Thanas]

What I meant was, is there potential for hard science style experiments to be set up in some of these fields if you had a lot of resources and virtually no qualms about using and abusing human beings to find out your ends (i.e., assuming that you can effectively use a certain number of people as "slaves" in your experiments)?

I recognize that it some of these, it's probably impossible anyways - there's no good way to set up a control.

In some, like history and law, it would be impossible not only because of that, but because of the variables involved. How would you, for example, try to figure out what happened at the battle of Mursa if there are so many unknown variables that the number of experiments one can do would be in the millions? You would need unlimited resources and time. And even then you wouldn't know what the correct answer would be. You would just end up with a group of results that would be more likely than the others.

In other fields, like sociology, one can do experiments. But then you run into the problem of individualism - an atom will always behave like an atom, humans won't. So you are left with broad conclusion and models. Very advanced models, sure, but nowhere near as definite as you can get in the hard sciences.

Then, as Alyrium noted, there are subdivisions, but these are IMO more the exception instead of being the rule.

[Block]

I think it's worth observing that the very softness of the soft sciences perpetuates their softness: hard sciences have high barriers to entry, requiring would-be practitioners to have advanced mathematical knowledge, logic skills, and experience with the scientific method. Soft sciences, on the other hand, require the practitioner to know how to read books, analyze paradigms, and write essays. Useful skills, but hardly ones that will weed out everyone except the people who will push the field in useful directions.

So you're saying that for example, a Neuro-psychologist, who's expected to know the physiological reasons for a number of brain disorders, possibly even do research into new ones, has less training than a chemist? or a Biologist? and that the entrance requirements for the last two outstrip the requirements for the first one? Or for a psychiatrist, who goes to med school, and usually isn't finished his training until the vey late 20's or early 30's?
Economists, yeah you'd be right for the most part, since marketing data isn't really hard to figure out for local areas, though as you go into higher levels of that you're dealing with math, logic and all sorts of statistical analysis that if you do wrong pretty much instantly costs you a job as well as any chance of being employed again.

I know the hard sciences are worshiped around here, but come on, you really think that there aren't a ton of really brilliant people that choose other paths?

[Darth Wong]

Funny how people keep citing neuropsychology as an example of hard science, when its greatest contribution to medicine has been the anti-depressant drugs that were hailed as a breakthrough for a new kind of psychological treatment, only to have studies show up ten years later demonstrating that they don't really have a net-positive effect at all.

The problem with any kind of psychology research, even if it starts with biochemistry, is that the point where the rubber meets the road is still human behaviour. Altering brain chemistry alters human behaviour, but the interaction is still so poorly understood that the side-effects and long-term outcomes are still basically unpredictable.

I know the hard sciences are worshiped around here, but come on, you really think that there aren't a ton of really brilliant people that choose other paths?

The fact that a brilliant person does something does not make it more scientific. The most brilliant people in your average university are the theoretical mathematicians, but that doesn't mean math is a science.

[Block]

Not really what I meant. His post implies that the reason people go into soft sciences is that they're not mentally capable of advanced levels of thinking.

[Block]

I also wasn't claiming neuropsych as a hard science, I'm simply saying it takes a great deal of intelligence and learning to become a competent one, which again the post I quoted seems to imply isn't the case.

[Darth Wong]

Not really what I meant. His post implies that the reason people go into soft sciences is that they're not mentally capable of advanced levels of thinking.

He suggested that they might lack "advanced mathematical knowledge, logic skills, and experience with the scientific method". You need all three to do hard science. Two of three makes you a mathematician. That is an "advanced level of thinking", but it is not a scientific way of thinking.

[Alyrium Denryle]

Funny how people keep citing neuropsychology as an example of hard science, when its greatest contribution to medicine has been the anti-depressant drugs that were hailed as a breakthrough for a new kind of psychological treatment, only to have studies show up ten years later demonstrating that they don't really have a net-positive effect at all.

That is a methodoligical and cost problem... If you are referring to MAOIs they had side effects that did unpleasant things (but did keep people from being depressed). Serotonin reuptake inhibitors are better...

It is the problem with medicine in general (as opposed to pure research) you have to field test before you necessarily know if the treatment actually works in the long term.

[Satori]

You can have "hard" quantitative figure in the social sciences, but only statistically... With a sufficiently large group of subjects behavioral patterns can be approximately graphed as a normal curve on basis of set criteria. The problem is that this only works, and fuzzily so, on the large scale.

The distinction between soft and hard science is somewhat blurry at times.