Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

To Fix the Social Sciences, Look to the “Dark Ages” of Medicine

thereader.mitpress.mit.edu

Post reply on HN
Print view View original

31–40 of 71 posts

Re: To Fix the Social Sciences, Look to the “Dark Ages” of Medicine

#31

Post by jftj »

malvosenior wrote: post #28
AlotOfReading wrote: post #27

Earlier quoted context omitted.

The "objective" approach was never actually objective. It had just as many biases and issues as everything after it, but everyone sort of pretended it didn't. The change was recognizing that the same data wasn't objective and that researchers couldn't make it so. Instead, theory shifted to analyzing those inherent biases and studying how they affected research.

If your data is not good enough, get better data. Don't abandon objectivity. I think this is why a lot of people don't take the social sciences seriously, there's an extreme resistance to just doing objective, data driven research (and the lack of reproducibility that comes with that).

It's not that "the data is not good enough". It's that it's easy to obfuscate results by hiding bias in the data, even if unintentionally. Even in the "hard" sciences. Or have we forgotten about some of the awful research being done in machine learning? Remember the one about identifying future criminals from photos?

Re: To Fix the Social Sciences, Look to the “Dark Ages” of Medicine

#32

Post by danieltillett »

ralusek wrote: post #15

For those interested in this, take a look at the work regarding "grievance studies" by Peter Boghossian, James Lindsay, and Helen Pluckrose. They made intentionally unscientific studies with the intent of publication, and actually got themselves published on a few of their works. The objective was to demonstrate that not only does academia in the social sciences operate according to a completely different standard th…

They would have got a lot more published except their experiment was sprung by some reporter.

The social sciences (along with most of the humanities) appear to be beyond saving at this point and it might be a good idea to make a hard break with the past and start again with people trained outside the field.

Re: To Fix the Social Sciences, Look to the “Dark Ages” of Medicine

#34

Post by naasking »

ralusek wrote: post #15

For those interested in this, take a look at the work regarding "grievance studies" by Peter Boghossian, James Lindsay, and Helen Pluckrose. They made intentionally unscientific studies with the intent of publication, and actually got themselves published on a few of their works. The objective was to demonstrate that not only does academia in the social sciences operate according to a completely different standard th…

I think the main takeaway from the Sokal Squared hoax was that reviewers in this field are either a) extremely biased and approve any results that agree with their beliefs, or b) anti-science, or c) scientifically illiterate, because they couldn't even spot the painfully clear methodological problems in the hoax studies that were inserted on purpose to test exactly the quality of review.

Re: To Fix the Social Sciences, Look to the “Dark Ages” of Medicine

#35

Post by paganel »

hogFeast wrote: post #4

Sounds like, unsurprisingly, the author hasn't spent that much time considering any of these questions themselves...they just have all the answers for those that do (thanks buddy!). Some questions just deal with issues that are just not quantifiable (even the question he poses about immigration is just not completely quantifiable). And even questions that are quantifiable are often affected by beliefs (i.e. what happ…

Very well put. Regarding history and its own past battles with this illusory search for "correctness" and mathematical-like precision, I'm really surprised that the author seems to have completely ignored Popper's pretty well known "The Poverty of Historicism" [1].

I'd also strongly recommend the author to also check Raymond Aron's "Introduction to the philosophy of history : an essay on the limits of historical objectivity." [2], a book first published in French in 1938 and translated into English in 1948.

I find it sad that some people still open up this discussion about trying to make history more "exact", more physics-like, I thought we had already proved that that is an impossible task.

[1] https://en.wikipedia.org/wiki/The_Poverty_of_Historicism

[2] https://www.amazon.com/Introduction-philosophy-history-histo...

Re: To Fix the Social Sciences, Look to the “Dark Ages” of Medicine

#36

Post by ajb »

malvosenior wrote: post #28
AlotOfReading wrote: post #27

Earlier quoted context omitted.

The "objective" approach was never actually objective. It had just as many biases and issues as everything after it, but everyone sort of pretended it didn't. The change was recognizing that the same data wasn't objective and that researchers couldn't make it so. Instead, theory shifted to analyzing those inherent biases and studying how they affected research.

If your data is not good enough, get better data. Don't abandon objectivity. I think this is why a lot of people don't take the social sciences seriously, there's an extreme resistance to just doing objective, data driven research (and the lack of reproducibility that comes with that).

It took 200 years for the hard sciences to figure out the right way to 'do' objectivity, and arguably it's still evolving (see 'Objectivity'[1] by Daston and Galison, nice summary here: https://scienceobjectivity.weebly.com/daston--galison.html). The social sciences have additional problems, saying 'just copy the hard sciences' isn't reasonable.

In fact, the idea that objectivity can be done in a simple mechanical way doesn't work for the hard sciences either, and was rejected by them in the early 20th, because it turns out that interpreting the evidence objectively still requires an educated eye. If you've ever had the good fortune (or misfortune, considering the circs) to see one of your own x-rays, you'll know how much you have to take the word of the radiographer for what the photo actually means. Sure, eventually, after much debate, a mechanical formula for interpreting the photo may be developed. Radiographers are probably due to be replaced by AI any day now - but their judgements will have formed the labels of the training data. At the cutting edge of the sciences, you don't have any of that yet.

[1] that book was expanded from their earlier paper which can be found here: http://cspeech.ucd.ie/Fred/docs/Galison.pdf

Re: To Fix the Social Sciences, Look to the “Dark Ages” of Medicine

#37

Post by jackcosgrove »

The article did not acknowledge the two biggest problems with the social sciences: inability to do many experiments because of ethical reasons, and the fact that these disciplines often study human social systems which have mutable "laws".

There's not much we can do about the first problem except slowly accrete knowledge through retrospective study of actual events, and maybe fill in some gaps with simulation.

The second problem is more solvable. I do believe that there are actually laws in the social sciences, but that they are far fewer in number than the phenomena people want to study. Many social phenomena are simply artifacts of existing systems, and may shed light on behavior in a particular case but are not generalizable beyond that system.

The law of supply and demand is a good example of a law that is universal, in many ways because it lies at the junction of physical systems and social systems. The physical aspects of supply and demand, for example the amount of arable land nearby as well as the caloric requirements of a population, can be measured well because of our physical knowledge and provide a decent jumping off place to expand the theory.

I disagree that some things are not quantifiable. Everything can be counted, from the number of neural connections in our brains to the number of widgets that a factory can produce. As usual, the limiting factor in science is not a lack of theory but a lack of instrumentation and thus data. I am basically an optimist about the social sciences as our measurement abilities continue to grow. However, we are in the dark ages regarding the tools we have to collect data.

Re: To Fix the Social Sciences, Look to the “Dark Ages” of Medicine

#39

Post by AlotOfReading »

malvosenior wrote: post #28
AlotOfReading wrote: post #27

Earlier quoted context omitted.

The "objective" approach was never actually objective. It had just as many biases and issues as everything after it, but everyone sort of pretended it didn't. The change was recognizing that the same data wasn't objective and that researchers couldn't make it so. Instead, theory shifted to analyzing those inherent biases and studying how they affected research.

If your data is not good enough, get better data. Don't abandon objectivity. I think this is why a lot of people don't take the social sciences seriously, there's an extreme resistance to just doing objective, data driven research (and the lack of reproducibility that comes with that).

Again, the issue isn't one of methodology or technology. It's inherent to the research process.

Let's do a thought experiment and imagine two identical research projects done in parallel universes. Everything is the same between these two universes except that the researchers are different people.

In some fields, you'd probably expect the results to be the same. The researcher shouldn't affect objective results after all. In a social science setting, the results may not be the same even if both researchers used exactly the same research design on exactly the same subjects at exactly the same point in their lives.

It should be very clear by this point in the thought experiment that "objectivity" is not really possible under those conditions. You need some theory to deal with this issue, but honestly it's manageable for the most part.

Now, the real kicker is that researchers affect their research way more than this. They actually design the project and constantly make decisions / interpretations throughout. These decisions and interpretations are subjective by definition. Even if we had a magical way to get objective data, are you really still willing to call it objective after a dirty, subjective human has so much as looked at it, let alone done the sort of filtering and interpretation inherent to academic research in all fields?

Re: To Fix the Social Sciences, Look to the “Dark Ages” of Medicine

#40

Post by nradov »

AlotOfReading wrote: post #7

I'm always amazed that someone who by all rights should be intimately familiar with the history of social science research like McIntyre can seemingly be so unaware of it. There was an empirical, modernist period in social sciences in the 40-80s of exactly the kind suggested by Lee and it was an absolute disaster. The results couldn't be related to the real world, and no one was really able to get over the inherent […

This is an intellectually lazy comment. It is by no means obvious that the gender and ethnicity of of the researcher would impact that hypothetical experiment, or what the impact would be, or that it couldn't be avoided with careful experimental design. If the social sciences had a better general track record of producing reproducible results with practical value for society then I would be inclined to give them the benefit of the doubt. But the reality is that most researchers have squandered their credibility by chasing mirages. If it were up to me I would cut funding for the entire field.
Post reply on HN