Topic: AI replacing academics in the social sciences

[Andrew Clunn]

So two topics on this board got me thinking.  I have huge issues with the social sciences because I see them as unable to overcome the biases of individual researchers, often use bad methodology resulting in poor reproducability, and largely aren't predictive.  But advances in AI (namely in neural networks where machines rewrite their own code to discover patterns... look into linear dichotomizers for a simplified form if you're interested in learning more) coupled with vast amounts of readily available data online are making large scale big data analysis more viable in a way that overcomes these biases.  Cladistics largely made taxonomy obsolete.  Will AI soon make short work of a social science in the next decade?  Well obviously academia won't let it officially do so (in the sense that tenured professors aren't going anywhere), but will it make research by individuals in a social science field moot?  This is as much a train of thought that I wanted to put out there as anything else.

[SkeptiQueer]

As I said in the other thread, how the AI is programmed and what information it's given will still produce a bias.

[Andrew Clunn]

As I said in the other thread, how the AI is programmed and what information it's given will still produce a bias.

Yes but sampling bias potentially exists in all research (and is something that can explicitly be managed).  Whereas bias (in the more colloquial sense) leading researchers to interpret their data in alignment with their own values or preconceived notions, would be eliminated.

[SkeptiQueer]

As I said in the other thread, how the AI is programmed and what information it's given will still produce a bias.

Yes but sampling bias potentially exists in all research (and is something that can explicitly be managed).  Whereas bias (in the more colloquial sense) leading researchers to interpret their data in alignment with their own values or preconceived notions, would be eliminated.

Again, bias will exist based on how it's programmed to interpret data.

[Harry Black]

Besides which someone still has to bridge the gap between the AI and the public/policy.
Its like thinking computers make mathematicians irrelevant. It becomes a tool for humans to get more information to male decisions on.
So, for example, if this AI suggested that it was more efficient to simply pay certain individuals to stay home and out of trouble, we would still need to decide if we want to live in a society that does that and perhaps that area is where the human element of social science would move to?

[Andrew Clunn]

Besides which someone still has to bridge the gap between the AI and the public/policy.
Its like thinking computers make mathematicians irrelevant. It becomes a tool for humans to get more information to male decisions on.
So, for example, if this AI suggested that it was more efficient to simply pay certain individuals to stay home and out of trouble, we would still need to decide if we want to live in a society that does that and perhaps that area is where the human element of social science would move to?

The fact that this conflation of social science with public policy making is so wide spread (hint, that is not science) is disturbing.  The social sciences are SUPPOSED to exist outside of political activism for particular policies.  Of course in practice they often do not, but that's more a topic for the other thread I started where I talk about why I'm so dismissive of the social sciences, not this one.

[Harry Black]

Im not saying thats the only reason they exist. Im saying its a thing that has to happen with any new information we get that could affect the policys we make.

[Andrew Clunn]

The point is that the policies are an afterthought.  Physics is physics regardless of whether you're trying to launch a rocket or not.  Taking a consequentialist view of science automatically will make one bias.

[Harry Black]

All Im saying is that removing humans from the field, even given a perfect AI, doesnt seem likely.

[SkeptiQueer]

Besides which someone still has to bridge the gap between the AI and the public/policy.
Its like thinking computers make mathematicians irrelevant. It becomes a tool for humans to get more information to male decisions on.
So, for example, if this AI suggested that it was more efficient to simply pay certain individuals to stay home and out of trouble, we would still need to decide if we want to live in a society that does that and perhaps that area is where the human element of social science would move to?

The fact that this conflation of social science with public policy making is so wide spread (hint, that is not science) is disturbing.  The social sciences are SUPPOSED to exist outside of political activism for particular policies.  Of course in practice they often do not, but that's more a topic for the other thread I started where I talk about why I'm so dismissive of the social sciences, not this one.

The sciences will never exist outside political activism, nor have they ever done so. I don't know why you think they can, could, or should.

[Dr Havoc]

Besides which someone still has to bridge the gap between the AI and the public/policy.
Its like thinking computers make mathematicians irrelevant. It becomes a tool for humans to get more information to male decisions on.
So, for example, if this AI suggested that it was more efficient to simply pay certain individuals to stay home and out of trouble, we would still need to decide if we want to live in a society that does that and perhaps that area is where the human element of social science would move to?

The fact that this conflation of social science with public policy making is so wide spread (hint, that is not science) is disturbing.  The social sciences are SUPPOSED to exist outside of political activism for particular policies.  Of course in practice they often do not, but that's more a topic for the other thread I started where I talk about why I'm so dismissive of the social sciences, not this one.

The sciences will never exist outside political activism, nor have they ever done so. I don't know why you think they can, could, or should.

The sciences should certainly exist outside political activism. The objective is to generate knowledge. Political activists can do whatever they want with the knowledge and perhaps political activism frames some of the questions that science investigates, but science can never have a vested interest in a particular answer. Political, religious or other kinds of beliefs have no place in science. A scientist should never set out to prove that what they believe is right. They can however set out to investigate IF what they believe is right. Only the former is political activism and only the latter latter is good science.

AI as it stands can replace physicists and mathematicians a lot more readily than it can replace social scientists. We still haven't even conceived of an idea that can give rise in the foreseeable future to a system capable of replacing human scientists or engineers.

[Johnny Slick]

Political activism may have no place in science, but science absolutely has a place in politics, unless by "[t]he sciences should certainly exist outside political activism" you mean that politicians should just accept the general consensus of science as fact when making, enforcing, and adjudicating laws. A United States of America where scientists and science-minded people do not get involved in political activism is a United States of America that "teaches the controversy" of creationism in schools, bans stem cell research because babies, and denies that climate change is occurring. Well, I guess 2 out of 3 ain't bad...

[Shibboleth]

So two topics on this board got me thinking.  I have huge issues with the social sciences because I see them as unable to overcome the biases of individual researchers, often use bad methodology resulting in poor reproducability, and largely aren't predictive.  But advances in AI (namely in neural networks where machines rewrite their own code to discover patterns... look into linear dichotomizers for a simplified form if you're interested in learning more) coupled with vast amounts of readily available data online are making large scale big data analysis more viable in a way that overcomes these biases.  Cladistics largely made taxonomy obsolete.  Will AI soon make short work of a social science in the next decade?  Well obviously academia won't let it officially do so (in the sense that tenured professors aren't going anywhere), but will it make research by individuals in a social science field moot?  This is as much a train of thought that I wanted to put out there as anything else.

Social sciences use double blind experiments quite often and other tools to avoid bias.

[Andrew Clunn]

So two topics on this board got me thinking.  I have huge issues with the social sciences because I see them as unable to overcome the biases of individual researchers, often use bad methodology resulting in poor reproducability, and largely aren't predictive.  But advances in AI (namely in neural networks where machines rewrite their own code to discover patterns... look into linear dichotomizers for a simplified form if you're interested in learning more) coupled with vast amounts of readily available data online are making large scale big data analysis more viable in a way that overcomes these biases.  Cladistics largely made taxonomy obsolete.  Will AI soon make short work of a social science in the next decade?  Well obviously academia won't let it officially do so (in the sense that tenured professors aren't going anywhere), but will it make research by individuals in a social science field moot?  This is as much a train of thought that I wanted to put out there as anything else.

Social sciences use double blind experiments quite often and other tools to avoid bias.

Example of such a study perhaps?

[Dr Havoc]

Political activism may have no place in science, but science absolutely has a place in politics, unless by "[t]he sciences should certainly exist outside political activism" you mean that politicians should just accept the general consensus of science as fact when making, enforcing, and adjudicating laws. A United States of America where scientists and science-minded people do not get involved in political activism is a United States of America that "teaches the controversy" of creationism in schools, bans stem cell research because babies, and denies that climate change is occurring. Well, I guess 2 out of 3 ain't bad...

Scientists are free to get involved in politics, but a scientist should never see an experiment as an opportunity to provide support for a political argument. Scientists must ask the question without any prior bias and report the result, even if it supports the argument of their political opponents. That is pretty rough to do when you are a political activist. That is why scientists who can't separate politics from science should choose one and give up the other. Science-minded people can do whatever they like, because they aren't doing any science.

It is oh-so-easy to go ahead and publish results you know are not airtight if you are a political activist with powerful backers. In political activism the ends often justify the means, but in science the end is irrelevant and the means are everything.