Catch up on stories from the past week (and beyond) at the Slashdot story archive

 



Forgot your password?
typodupeerror
Check out the new SourceForge HTML5 internet speed test! No Flash necessary and runs on all devices. ×
AI

An Algorithm That Can Predict Human Behavior Better Than Humans (mit.edu) 84

Quartz describes an MIT study with the surprising conclusion that at least in some circumstances, an algorithm can not only sift numbers faster than humans (after all, that's what computers are best at), but also discern relevant factors within a complex data set more accurately and more quickly than can teams of humans. In a competition involving 905 human teams, a system called the Data Science Machine, designed by MIT master's student Max Kanter and his advisor, Kalyan Veeramachaneni, beat most of the humans for accuracy and speed in three tests of predictive power, including one about "whether a student would drop out during the next ten days, based on student interactions with resources on an online course." Teams might have looked at how late students turned in their problem sets, or whether they spent any time looking at lecture notes. But instead, MIT News reports, the two most important indicators turned out to be how far ahead of a deadline the student began working on their problem set, and how much time the student spent on the course website. ... The Data Science Machine performed well in this competition. It was also successful in two other competitions, one in which participants had to predict whether a crowd-funded project would be considered “exciting” and another if a customer would become a repeat buyer.
This discussion has been archived. No new comments can be posted.

An Algorithm That Can Predict Human Behavior Better Than Humans

Comments Filter:
  • by Ungrounded Lightning ( 62228 ) on Monday October 19, 2015 @01:26AM (#50756529) Journal

    It might be interesting to try it at predicting future violent behavior of individuals.

    Those who study such behavior have come up with the aphorism "The only effective predictor of future violent behavior is past violent behavior." Let's see if the

    Shrinks who try to make predictions about individuals come out WORSE than chance - which implies that there may be SOME prediction possible - but the current paradigms have it backward.

    This, by the way, is ONE of the reasons the pro-gun crowd pooh-poohs mental health tests for gun ownership or purchase. Another is the observation that people with mental illnesses are, on the average, far LESS likely to cause harm to others than the average of the population. (They may harm THEMSELVES, but suicide rates don't change if guns aren't available: Instead the suicidal switch to less effective and usually more painful means, averaging more tries before they succeed.)

    • (The Lenovo Trackpad ate my homework again. B-b )

      Those who study such behavior have come up with the aphorism "The only effective predictor of future violent behavior is past violent behavior." Let's see if the program can identify any other indicators.

      • by gweihir ( 88907 )

        There is one: Membership in the biological group "Homo Sapiens". Anything more accurate is just trying to stick labels on people to make them "different" and push them out of society. Some people like doing that to elevate their pathetic selves by stomping on people they perceive to be even more pathetic.

    • by gweihir ( 88907 )

      The problem is that most people are violent deep down. The question thus becomes one of self-control and to what limit it gets tested. That is mostly an external effect, i.e. the level of provocation (sometimes real, but usually perceived only) the subject is exposed to. As you cannot measure the level of self-control, what you get is people where their normal environment often or sometimes exceeds their control and they get violent. As most people with lower control are unable to adjust that control by the

      • The Marshmallow Test has been used to try and assess innate levels of self-control in young kids, and I believe it strongly correlates to things like academic success. https://en.wikipedia.org/wiki/... [wikipedia.org]
        • by gweihir ( 88907 )

          This is an interesting thing. It does not seem to apply to violence though as violence does not occur in regular situations, but a stressful ones. But maybe a modified version would work and finding kids with problems when making decisions under stress and tending towards violence then could be identified early on and given extra skills in self-control under those circumstances.

          This would need to be done without giving them any stigma or permanent record though, so it cannot really be done in the US these d

        • by KGIII ( 973947 )

          I am reminded of the Emo Phillips skit about the chocolate Easter Bunny. I went and found it for you.

          https://www.youtube.com/watch?... [youtube.com]

          If you like highbrow humor, he's your man.

      • by HiThere ( 15173 )

        I believe that there's also a learning effect, where is someone learns that they will usually get their way if the threaten violence, they will be both more likely to threaten and more likely to engage in violence. And when you practice at something you usually get better at it, and so find it more rewarding than those that do not so practice.

        Lots and lots of feedback loops, also feedforward loops.

    • An Algorith That Can Predict Boring Human Behavior Better Than Boring Humans
      There, FTFY
    • The way to protect yourself from data sprongers is to use their spronging against them. Control what they see so they do what you want. If you're smart enough, you won't get caught, and they will never get data that people like you commit crimes because people like you don't get caught. Be the kind of person that doesn't get caught.

    • by G00F ( 241765 )

      (They may harm THEMSELVES, but suicide rates don't change if guns aren't available: Instead the suicidal switch to less effective and usually more painful means, averaging more tries before they succeed.)

      That is very false. Suicidal rates drop with measures preventing people from obtaining guns.(see the book More Guns Less crime) It's really the only statistic that gets better after passage of gun laws.

      In fact this is true with all methods people use for committing suicide. Even minor inconveniences like

      • Suicidal rates drop with measures preventing people from obtaining guns.(see the book More Guns Less crime)

        I have a strong respect for Lott's work - but haven't read all of it. This is the first time I'd heard that he was putting forth that assertion.

        It would be interesting to examine how much of the effect, if any, was an actual reduction in the suicide rate and how much, if any, was that some of the suicidal simply went to another location.

        However:

        Studies have shown that well-designed suicide barriers no

  • by Anonymous Coward

    Just what we need right? More tracking of people and their (possible) habits. That surely won't be abused, used to pre-maturely judge / fire / expel / etc someone based on their social media pages (or lack thereof)? It won't be used to serve us ever more intrusive ads by monetizing every last detail of our lives and never giving us a penny? It could never be used to create a pre-crime system ala Person of Interest's "the machine"?

    For everything these analytics programs do, I don't see much benefit from them

    • Re: (Score:2, Insightful)

      by gl4ss ( 559668 )

      it's just what seems rather simple data analysis.

      I don't get it though how it would get the answer better than a human who should be using the same resources anyways, that is, using something like this as a resource.

      not sure what good it does for them to know if someone is dropping out of an online course or not though, like, I suppose it just checks if someone has advanced in the course in the past 3-7 days at all

      • by Kjella ( 173770 )

        not sure what good it does for them to know if someone is dropping out of an online course or not though, like, I suppose it just checks if someone has advanced in the course in the past 3-7 days at all

        If they're paying tuition or to take the exam then sending some kind of trigger message is obviously good for business. See also general customer retention, freemium games and so on. Everybody wants to know what customers are likely to leave them soon.

  • "MIT News reports, the two most important indicators turned out to be how far ahead of a deadline the student began working on their problem set, and how much time the student spent on the course website. "

    IOW, the computer program used a more applicable approach/pertinent data and had humans had used the same algorithm/data, they would have reached the same or better results.

    • I've seen several plugins and 3rd party services for popular learning management systems that alert teachers and/or admins to potential drop-outs before they happen. As we've seen from the MIT article, it's not actually that hard to do.

      The research on this goes back several years so this part really isn't news.

  • Overblown headline (Score:5, Informative)

    by binarstu ( 720435 ) on Monday October 19, 2015 @01:44AM (#50756555)

    The headline of the Quartz article and the Slashdot summary, "An algorithm can predict human behavior better than humans", is, not surprisingly, hugely overblown.

    What these researchers actually did was develop a system for automatically taking a massive data set with a huge number of variables, identifying the subset of variables or new combinations of variables that are most likely to be useful for predicting a particular response, and then formulating a predictive model. (This is an extremely simplified summary.) That is really cool, but to present it as some sort of general "algorithm for predicting human behavior" is silly. It's no more an algorithm for predicting human behavior than are automated statistical methods for building a predictive model from a massive dataset.

  • Good! Maybe AT&T will finally use the system to realize we are sick and tired of their telemarketing calls and that we hate their slimy guts for their billing gimmicks.

    Right now they seem clueless, as if a bunch of PHB's in a room stick a wet finger in.....the air to make a GUESS that we like being annoyed and that annoying us somehow makes them more profitable.

    Robots couldn't do worse than the current marketing idiots there. Hell, give chimps a try also.

  • No surprise (Score:5, Interesting)

    by gweihir ( 88907 ) on Monday October 19, 2015 @01:55AM (#50756575)

    First, most humans are not that smart, but do not know that. It is called the "Dunning-Kruger" effect and it is well-established. Apparently at least the OP is unaware of it, possibly making him a subject of the Dunning-Kruger effect.

    Second, a majority of humans do not use what smarts they have effectively, but rather do decide "emotionally" when it comes to important decisions or understanding important situations. That obviously works rather badly, just look at what politicians get voted into office, or what life-choices people make. The problem here is that the whole "emotional decision" apparatus is a rather primitive left-over from caveman-times that cannot handle even situations of moderate complexity well. The second problem is that most humans never find that out, as the skill for self-reflection is also rather scarce and hence cannot actively compensate.

    So give an arbitrary group of people an analysis or decision problem that somehow "touches" them (like asking students to predict whether other students fail at being students), and suddenly most of them turn into morons (or rather do not stop being morons in many cases), and even a simplistic statistic predictor does a lot better than they do.

    • It is called the "Dunning-Kruger" effect and it is well-established.

      Otherwise known as the "You kids get off my lawn, unless you have more letters after your name than I do" effect, or the "How to be a condescending expert" effect. There's a reason they won the Ig Nobel Prize in 2000 for the paper. The fundamental presumption is that, even knowing of the effect, you will not be able to avoid it, and therefore your opinion should be discarded.

      • by gweihir ( 88907 )

        Seems you have not understood at all what the research says. (No surprise. Likely you are on the left side of their plot.)

        The Ig Nobel Prize is not only for bad research. Quite a bit of good research has gotten it.

  • It is impossible to behave as humans better than what humans can do. As far as predicting is basically simulating the given behaviour, there is nothing which can predict human behaviour better than what humans can do.

    Translation of the previous paragraph to a more-appealing language (c-biased pseudo-code):

    Function main()
    {
    behaviour = doBehaviour(true);
    predAlgorithm = doBehaviour(false);
    predHuman = doBehaviour(true);
    countAlgorithm = 0;
    countHuman = 0;

  • I don't think Hari Seldon is too worried for his job.

  • based on student interactions with resources on an online course.

    Clearly the human brain is good at interpreting a different kind of behaviour (instantaneously and visually received), not temporal behaviour of interaction with a machine presented as a spreadsheet. Given the type of information and unintuitive presentation for the human brain, the question is: did the humans have as long with the training data as the ANN? I'm guessing not, but i suppose the competition does at least prove it's utility over an untrained human rather than to purely serve as publicity.

    This i

  • We had long known that given a feature set, computers can automatically fit a pretty good model. Now, can they automaticall extract this feature set and data without humans?

Lend money to a bad debtor and he will hate you.

Working...