Slashdot Mirror
The Limits of Big Data For Social Engineering
An anonymous reader writes "In his new book, Social Physics, MIT data scientist Alex 'Sandy' Pentland argues that by analyzing data from smartphones, social media, and credit-card systems, we'll soon be able to have a mathematical understanding of 'the basic mechanisms of social interactions.' Social scientists will be able to understand and predict the interactions of people the way physicists understand and predict the interactions of objects. That will, in turn, enable governments and businesses to create incentive systems to 'tune' people's behavior, making society more productive and creative. In a review of Pentland's book in Technology Review, Nicholas Carr argues that such data-based social engineering 'will tend to perpetuate existing social structures and dynamics' and 'encourage us to optimize the status quo rather than challenge it.' Carr writes, 'Defining social relations as a pattern of stimulus and response makes the math easier, but it ignores the deep, structural sources of social ills. Pentland may be right that our behavior is determined largely by social norms and the influences of our peers, but what he fails to see is that those norms and influences are themselves shaped by history, politics, and economics, not to mention power and prejudice.'"
What could possibly go wrong?
Confucius say, "Find worm in apple - bad. Find half a worm - worse."
This is for some BA forum. Not a BS forum like this!
This is great and all, but how do we dumb the model down enough that it's usable? What is the social equivalent of a spherical cow?
On a serious note, this could really be a double-edged sword. Not conforming to the model of behavior? Into the gulag with you, citizen.
Yes, because surely governments want to make society more productive and creative instead of giving themselves more money and power.
Ah, Asimov was ahead of his time.
I think I wrote that back in middle school
10 wake up
20 go somewhere public
30 regret it and go home
40 sleep 28800
50 GOTO 10
Big Data wants to be Multivac, Google wants to be R. Daneel Olivaw, and now Hari Seldon has sufficient data to begin his work. Yet, as in Foundation and Earth, is a benevolent dictatorship by remote overlords truly the answer, or is there still something missing that could doom us all...?
I read that as 'we would like to change your opinion by using tricks'.
This strikes me as mass manipulation.
If you want to change someones opinion you need to show them why it is bad.
For example yesterday I demonstrated to a waitress how her feeding of the jukebox was a bad idea. All of her fellow employees were saying 'dont put any more money into that it is a waste of money'. 'But its only 2 dollars a day'. I spoke up with 'what if you had say 500 dollars right now what would you do?' 'OH I would pay off some bills'. 'That is about what you spend per year on that box.' 'oh had not thought of it that way'.
Honestly, I manipulated her. I was a bit tired of her choice of music. But I used a monetary manipulation to make her stop.
This sort of manipulation is used to force others into your opinion. A difference of opinion is not a bad thing. However, there are those out there would would see you dead for not thinking exactly like them (and this is more groups than the ones you are thinking of).
Many of us technical types would love for this line of inquiry to be fruitful. But to have a 'physics of people' you have to know the values of all the parameters needed to specify the current state of a person and you need to know all interactions of that person with the rest of the universe. Phrased like that you can see how ludicrous it is to dream of using the methods of physics for social science. Physics works because the fundamental constituents of the universe happen to be only a small number of particles whose interactions are amazingly simple. For example all electrons are exactly identical and interact via only 3 forces (with some uncertainties about effects on scales larger than galaxies and energies higher than trillions of electron volts). The hope for a theory of sociology is a false hope. The hope for a useful phenomenology might be more reasonable and big data can help.
seem to me to be two separate things. Statistics tells you what happens when. Understanding tells you why things happen when they do. Statistics can give you data, but not reasons. Understanding requires both. Even with all the data in the world, we're still going to need some kind of interpretative framework to make sense of it all, and creating that framework is the one thing Big Data doesn't make easier.
Racial profiling is a good concrete example of this. Crime statistics are often used to rationalize race-based profiling, but it is key to realize that by taking this step you are moving beyond passively understanding the present, into using that information to shape the future, thus perpetuating a problem that should instead be solved.
the Republicans will keep abusing these tools. Also, these tools make it easy to determine which people to fuck over because so many people will either fill-out their race or give-out enough data to make it obvious. For Suckerberg, News Corp w/ myspace, and the rest of their kind, hurting minorities has been their number one goal. Just look at how bad myspace is and the demographic that still uses it. Murdoch is punishing African Americans with that horrific site. It is horrible that he is allowed to do that.
Most notably, all the data being considered is wholly commercial in nature. For these people, an interaction between human beings that doesn't involve a third party cannot possibly exist.
Industry over-extends usage of fad and is disappointed when it can't produce magic.
Gee, would've never seen that coming.
Table-ized A.I.
>we'll soon be able to have a mathematical understanding of 'the basic mechanisms of social interactions.'
Wall Street has claimed to do that countless times, and countless times, they have been proven wrong with a monumental market crash, around every 8 years or so. We should be due for one soon.
What are they going to do when they find out it doesn't work for smart people or people who make conscious decisions to alter their behavior based on their own research? They will just be ignored as outliers because they don't fit in to the statistical modeling. How does the machine learning algorithm model a learning human unless it knows where they're going before they do? What if people inside the model start computing social interactions based on a different model? Do we prohibit these people from evolving their behavior because they don't live inside the machine's conception of how they should act?
Social scientists will be able to understand and predict the interactions of people the way physicists understand and predict the interactions of objects
Since social scientists are completely unable to quantify anything in their field of study, I somehow doubt that they will ever come anywhere close to doing "real" science in the way that physicists can: with their "laws", measurements and equations.
But maybe this person doesn't really have much idea how proper scientists do their work?
politicians are like babies' nappies: they should both be changed regularly and for the same reasons
Nope. And even more: NOPE!
What it will actually give are the ways society CAN change to make people more productive,
creative, and happy. These changes will not occur. They will be counter to the desires of the CEOs
and Politicians who want more productive WITHOUT the changes that might hurt or decrease their power.
I've found that most people don't appreciate it when they realize they are being 'handled' and it usually ends badly.
The question is ... how long will it take the population to realize they are ... if ever?
Someone was supposed to apply the PAX before they published this article
Persistent Volume manager for Kubernetes - https://github.com/dwimsey/openshift-pvmanager
He's not impressed by your MIT data scientist.
... typically think they can predict individual events from statistics. That is not how it works. Statistics lets you predict statistical things, e.g. how the average of, say, 1000 people will react given a specific situation, when you have an observation how the average of 1000 other people reacted to that situation, but only if the selection was random and there are not many possible reactions. It tells you exactly nothing about how a specific individual will react.
Most ACs are not even worth the keystrokes to insult them. Be generically insulted by this and ignored otherwise.
good comment...I came here to post a violent rebuttal to the notion that:
but then I saw your comment and realized that i'm not the only one who thinks this research is absolute shit ...srsly..."just like physicists!"
Thank you Dave Raggett
> data-based social engineering 'will tend to perpetuate existing social structures and dynamics' and 'encourage us to optimize the status quo rather than challenge it.
That's a good thing. Social interactions are not just about behavior, but also genetics. We're programmed to work a certain way, why challenge it? Because it MIGHT result in a net gain from disrupting successful genetic patterns?
I'd like to suggest History is more determined by the ones of us too stupid to understand the social norms were supposed to bow down to.
Don't complain about syntax, grammar, or spelling. There is no.hell like input on android.
fuck yeah!
also: it's more than just those jerk-offs you mention
it's like we have a huge asteroid that is a "planet killer" followed by dozens of smaller "city killer" sized ones headed *straight for earth*...we obv need to take out the first one but the job isn't over at that point
Hillary Clinton...wtf...I want to think that if she gets elected she wont be beholden to corporatists, but that's not her typical M.O.
there are many others...but Democrats will fall in line to a progressive agenda if either one brave person pushes the limits and is rewarded *or* someone like Hillary gets defeated by an Elizabeth Warren type
Thank you Dave Raggett
Nicholas Carr argues that such data-based social engineering 'will tend to perpetuate existing social structures and dynamics' and 'encourage us to optimize the status quo rather than challenge it.'
That's the goal. As long as there's enough bread and circuses to go around, the social engineers can make the masses dance to their tune.
Taking guns away from the 99% gives the 1% 100% of the power.
yes but not quite...there's a difference in scale & depth of corruption
as to the "BRAG" you mention...I know what you mean...it's the dipshits from Chicago "startup" that were profiled several times in places like Time Magazines under headlines like "Obama's tech wizards" and bullshit like that...
if you look at it...the 2012 election vs 2008 the notion that his 2012 "tech wizards" somehow **outperformed** the 2008 strategy is absolutely silly...if anything they fucked him over by telling him to narrow his message...aka "target"...
the whole "Obama's 2012 tech wizards" narrative was *hype from a dumb startup*
Here's the secret to getting elected: focus on registering as many people as possible and then communicating with them about what's important to them in **all the ways humanly possible** without violating privacy
it's really that simple
Thank you Dave Raggett
The irony is that at a deeper level it's utterly unpredictable what and when a single particle will do in physics. Take a lump of uranium -- it's easy to predict when how much of it will have decayed to lead. It works all the time, always the same. But look at a single uranium atom and there's no way to predict when it will decay. It may be the next second or in a thousand years. All you have is probabilities but these work out into cold, hard predictable facts if what you're dealing with is a lump large enough.
Psychology works very similar. You can't predict what an individual person will do, but look at enough of them and you'll be able to predict what will happen if you have good enough data. YOU may have "free will" and the freedom to do what you want but as a mass we may still follow strict laws, like everything else in nature.
You may feel insulted by that or you may see such things as great tools for better understanding of social dynamics.
He wants his deeply unlikely fictional premise back...
It was called "the Soviet Union." Adding smart phone and Facebook data wouldn't have made that clusterfuck of genocidal failure any better...
I haven't read the book, but this summary sounds a lot like the new left's cultural marxism, with the latter half referencing political correctness as justification. Whether it is or not, most people don't like being 'gamed' in this way, and when they find out, the backlash can be far worse than the desired outcome and/or the original status quo.
Asimov predicted that you'd be able to predict what the masses will do, not what an individual will do. There's too much uncertainty with an individual. But with enough of them and a good enough statistical model, you can get some good extrapolation . . .
Anyway, we're a LONG ways from that being anywhere close to accurate. More importantly, I don't know about you, but I don't want the government that can't even balance it's own checkbook to try and balance my individual life. It's already engaged in enough social engineering with detrimental effects.
Except in Physics, ALL particles will follow the same basic laws and behavior over time. People are more complex systems and don't follow the same behavior. A lot of people often don't even follow sensible behavior. But those are the easiest to manipulate.
He'd have the mathematics of psychohistory all worked out for us, by now.
"That will, in turn, enable governments and businesses to create incentive systems to 'tune' people's behavior, making society more productive and creative."
HOW STUPID DO YOU THINK I AM. ThIS IS INSULTING!!
And who shapes our perception of history, politics, and economics, power and prejudice?
I'm still waiting for the media to publicly question the cause of WTC tower 7 collapse as hard as they've questioned any heart-tugging story about someone that went missing.
In a pistachio nutshell:
Their paradigm is based on that human behavior is deterministic and as such predictable. To them it is purely the result of chemical and electric phenomena. We believe there is a ghost in the machine, they do not. We will find out which holds true. However the "people behind the curtain behind the people behind the curtain" certain know better. Their hope their current crop of Kurt Lewins et al., is to limit human expression by further surpressing the ghost in the machine.
To lean more start here:
http://www.tavinstitute.org/pr...
Judea Pearl and others have done a lot to work that out. It's the science of answering "what if". It's not impossible.
weinersmith
...is the maximum amount of money big business can line it's pocket with. More and more rapidly, there's becoming no distinction between big business and government.
The deterministic laws of physics comes out of quantum stochasticity, so it's easy to see why one would think the same could be applied to human behavior. An individual is unpredictable but masses of people are very predictable. But how many people must be in the group before that is the case? In physics even a very small object has an enormous amount of individual particles at play. A grain of salt has more atoms in it then their are people on the planet.
Obviously I'm not the first person to bring up Hari Seldon and psychohistory, but the same held true there. You needed a galaxy's worth of population before it became useful.
I don't think people are that simple, static or homogeneous.
It will be like squeezing down on a watermelon seed. The harder you squeeze the higher the escape velocity.
I mean holy unintended consequences.
I'm getting my Ph.D. in behavioral and institutional economics, so this is right up my ally. Carr's response, that "defining social relations as a pattern of stimulus and response makes the math easier" but misses the deeper structure is dead on, but it's more than this. Social norms may determine much of our behavioral responses, but norms vary tremendously by the institution from which they come: how your group of friends prefer to treat each other != how other's prefer to treat each other != how strangers are "supposed" to treat each other in NYC != out strangers are supposed to treat each other in a small town. Moreover, while these norms may be highly correlated with your behavioral responses, people select into institutions (friend circles, communities, neighborhoods, etc) to a large degree based off their compatibility with the institutions norms; e.g. think about outcomes of social group formation, from mostly scratch, freshman year of college.
This all matters for the article's context because the behavioral parameters they estimate only approximate social norm's suggested behavior, but the suggestions ultimately come from those who chose to adhere to that particular set of norms; trying to "tune" people in ways they don't intrinsically want will fail because they'll just reselect or simply ignore the competing suggestions in favor of those authentic to the group into which they selected. If they take into account that all norms are highly idiosyncratic to their parent institution, it may help with better targeting of products, programs, and information, but the targeting will still have to be revised as people revise their norms; an institution will not revise its norms to conform to what an outside entity feels they should be. So, yeah, I don't think catering to the current observed state of the world can keep norms and society from evolving any more than, say, de jure segregation laws catering to status quo racists/-ism can keep people from forming revising their views about the morality of racism, the laws surrounding it, and their behavioral responses to such societal "tuning", especially over years and generations.
even weirder with physics, the closer you look at the uranium, the less you are sure that it's really there at all. it looks like some sort of knot in waves of energy combining with others according to probabilities. There's no "thing" actually there to measure the size of at some point.
One of the fundamental AI issues is figuring how that can apply to social groups and individuals. It's not going to be solved. It's probably even the wrong way to look at the question.
That will, in turn, enable governments and businesses to create incentive systems to 'tune' people's behavior, making society more productive and creative for those governments and businesses.
How is this different from what the supermarkets have been doing for years with their loyalty cards?
Not that I'm a fan of prejudice. But history, economics, power, and politics are all very useful very valuable and much more to my liking than arbitrary opinions about what constitutes a social ill or a social good. WE DECIDE what is good or evil BASED ON history, economics, power, and politics, AND WE SHOULD - it crowd sources morality, as Darwin and God intended - that's what morality IS - what benefits the herd and it's members, as determined by the behavior and the resulting history, economics, power, and politics of the herd and of it's members.
Go join a Church of your choice, fool, and leave the real world to realists.
"Psychology works very similar. You can't predict what an individual person will do, but look at enough of them and you'll be able to predict what will happen if you have good enough data. YOU may have "free will" and the freedom to do what you want but as a mass we may still follow strict laws, like everything else in nature."
I think this is raw, Asimovian geek fantasy. Do you have a citation for this assertion? Social uprisings catch people by total surprise every generation. No one can predict the stock market, as much as the desire is there.
The fact that people (and thus societies) are engaged in conscious feedback loops means that they're qualitatively different from masses of dumb particles.
We know where leadership by an anti-intellectual "strongman" who scapegoats minorities and likes boisterous rallies goes
"Social scientists will be able to understand and predict the interactions of people the way physicists understand and predict the interactions of objects."
The fact that this is a century-old Asimovian fantasy that's gone nowhere aside (in the late 80's I was being taught that chaos theory had killed that hope; consider a hundred thousand attempts at predicting the stock market)...
Do social scientists even know how to do math? I was in a scholarly seminar a few weeks ago (the only STEM person in the room, everyone else was social scientists), and was nearly shouted out of the room when I did a spit-take on an sample published paper held up that involved a sample size of 8 sociology students keeping journals for two weeks. One of the other participants said out loud that she had know idea what the point was of another paper because it was quantitative (i.e., involved numbers) instead of qualitative (i.e., subjective opinions by the researcher). As far as I can see recently the whole discipline appears to be a "null field".
We know where leadership by an anti-intellectual "strongman" who scapegoats minorities and likes boisterous rallies goes
Hail Hydra
"That will, in turn, enable governments and businesses to create incentive systems to 'tune' people's behavior, making society more productive and creative."
Mmm... there's this thing that most societies have, it's called education. It can be highly effective at 'tuning' people's behavior, making society more productive and creative. But something tells me that's the last thing on the USA's rulers' minds...
This'll get you up to speed on what they're doing to education in the US: http://billmoyers.com/episode/public-schools-for-sale/
Public Schools for Sale?
March 28, 2014
Public education is becoming big business as bankers, hedge fund managers and private equity investors are entering what they consider to be an âoeemerging market.â As Rupert Murdoch put it after purchasing an education technology company, âoeWhen it comes to K through 12 education, we see a $500 billion sector in the US alone.â
Education historian Diane Ravitch says the privatization of public education has to stop. As assistant secretary of education under President George H.W. Bush, she was an advocate of school choice and charter schools; under George W. Bush, she supported the No Child Left Behind initiative. But after careful investigation, she changed her mind, and has become, according to Salon, âoethe nationâ(TM)s highest profile opponentâ of charter-based education.
On this weekâ(TM)s Moyers & Company, she tells Bill Moyers, âI think whatâ(TM)s at stake is the future of American public education. I believe it is one of the foundation stones of our democracy: So an attack on public education is an attack on democracy.â
Diane Ravitch is Americaâ(TM)s preeminent historian of public education. Her newest book is Reign of Error: The Hoax of the Privatization Movement and the Danger to Americaâ(TM)s Public Schools.
Good luck with this... it's not like I can have scantily clad H to help the Uranium decay faster. Unfortunately it'll be hard to figure out why I tip a waitress, without seeing how far out of my league she is but show much cleavage and how much she pretends to like me... all of these are subconscious factors but there's now way in hell I'll tell admit to this. Mrs. Coward would cut off my balls.
The WAR has been started for quite a while already!
The Beatles (rise to fame, uh, and hysteria), The Tea Party, sushi becoming normal food for yuppies, long hair in the 60's for young male middle and upper class youth, long hair in the 70's for blue collar and redneck male youth, etc. etc.
Similar in a fashion to this story....
Some geographers tried to use the math model that describes the migration of particles between atoms to try to describe population migrations between population centers. This (if I recall) was called the Gravity Model. And it was a huge attempt to make the softer sciences a bit harder by 'mathing them up'. Beyond that, it was trying to shoe-horn the data to fit the theory.
It failed and fell into just disrepute.
I have a feeling that much of the 'amazing power' of Big Data is going to turn out to be a) not as useful as anyone thinks and b) even where it is statistically accurate, it will tell little or nothing about individual instances, thus limiting utility for many purposes.
Big Data can do some things. I can collate and correlate vast data samples and perform other useful descriptive statistical tasks. When they start to move off into inferential activities, I suspect things start to break down.
I don't feel like the data about what I do with my smart phone will usefully tell anyone much about me that they could not have obtained otherwise. And its value as a tool to sell me things is fairly dubious.
-- Mal: "Well they tell you: never hit a man with a closed fist. But it is, on occasion, hilarious."
You will be assimilated.
and they want their Big Data back!