Can an Algorithm Give You Advice About Your Love Life?

By Allison Bond | June 16, 2009 6:07 pm

love Remember that water-filled Magic 8 ball you used to consult? That decision-making toy has gone high-tech, thanks to a new service called Hunch, created by one of Flickr’s co-founders. The site enters information about you into an algorithm developed by MIT computer scientists. It then formulates answers to personal questions, from what you should make for dinner to where you should take a vacation.

Here’s how it works: Just like on matchmaking sites like, users create a profile by answering questions about themselves—up to 1,500 questions, in this case.

After your profile has been created, you can ask the site a specific question. Hunch’s algorithm will lead you through another series of inquiries to filter out undesirable choices and rank those that remain. Finally, the site presents you with what is supposedly your best option.
Once the site suggests a decision, you can enter feedback about whether you think the “choice” made for you was a good one. Developers hope that input from users will shape Hunch’s algorithm, helping the site yield decisions that users increasingly find desirable.

But while the service is clever, it raises a bigger question: Can a computerized system really make decisions as well as a human can? After all, research has shown that humans’ decision-making efforts can be irrational, so capturing human nature in an algorithm could be pretty tricky. And, of course, there’s always the possibility that the algorithm could make more rational choices than we’d like.

Ultimately, we think it depends on the question: For example, Hunch would probably be a better tool for helping you decide which camera to buy than it would advising you whether to dump your boyfriend.

Related Content:
Discoblog: Google Turns “Magic Algorithm” Inwards, Predicts Which Employees Will Quit
Discoblog: Amateur Geneticists Biohack From Home
Discoblog: Ant Intelligence Could Help Us Steer Clear of Traffic Jams

Image: flickr / kjunstorm 

CATEGORIZED UNDER: Technology Attacks!
  • Al Goreythm

    Oh Werd!

  • Charles Barnard

    Frankly, humans are not very good at solving problems which have an emotional element–which is nearly every question you generate.

    Advice can come from anywhere and nowhere–as you note, the key is not whether or not a particular source can provide advice, but whether or not that advice is any good.

    If you think you are safe or you think that you are at risk, you are wrong–risk perception in particular is done poorly by humans.

    The more emotional content a decision holds, the more likely that a human will make an irrational choice.

    Computers, the stars, entrails or cards, we always have tried to foist our decisions off of ourselves, and yet, ultimately, it is our decision to accept or reject advice.

    Note that the decision to bomb Nagasaki was ultimately made by one man from a number of choices.

    He could have opted to:
    1) Drop the bomb despite being unable to see the target (what happened.)
    2) Choose yet another target (Nagasaki was 2nd choice.)
    3) Drop the bomb in a place where the Japanese military would be certain to notice, but which was uninhabited.
    4) Drop the bomb far out to sea.
    5) Fly to an airstrip which was above the 1800 foot trigger altitude, and have it disarmed.
    6) Descend to detonation altitude and commit suicide.

    According to his orders, options 2-4 were actually given, and option 1 was to be avoided–primarily because the military wanted a clear view of the explosion and damage. Since the mission had left the only camera operator at base, there was no photographic record, which would have been the main reason for the requirement of a visible target.

    Computerized systems can only make decisions based upon the data they are given, but humans are usually presented with decisions which must be made without sufficient data.

    Can a program do that? Yes. Can it do so “better” than a human. It is likley that the answer, as you say, is dependent upon the question, and the data used.

    As with any decision, the most important single factor is the phrasing of the question, very often, any particular question, as posed, automatically eliminates many possible answers.

    For the instance cited (dumping your boyfriend,) the question takes what is a very complex situation with many variables, and brings it to a forced choice–and implies other questions which will need answering before n actual course of action can be taken.

    As a society, we routinely spend far too little time preparing to make a decision by deciding what the appropriate question is and far too much time justifying the decision after we have made it.

    For instance:

    Second-hand smoke has been a major issue in my state for some time, but the solution (banning smoking) is based upon the question: Should we ban smoking because second-hand-smoke is a health risk to nearby non-smokers?

    But the correct question is: How do we ensure good indoor air quality? Because smoke is only one of many different factors which affect the quality of the air. You can ban smoking, and the air quality may still result in more deaths than necessary.

    Setting indoor air standards for public places, which could then be monitored would solve the stated problem (smoking) as well as the unstated problem (air quality.)

    Note that while prohibiting smoking is now law, we still permit manufacturers to dump hundreds of tons of toxic chemicals into our local air.

    The very first phase of any decision is to determine the proper question. This is much more difficult than most appreciate.

    In the case of smoking, as with speeding, it is far more convenient to enforce laws against smoking rather than laws specifying air quality, just as it is much easier to enforce a speeding law than a law regarding vehicle spacing–despite the fact that vehicle spacing accounts for far more traffic accidents than speed per se does….

    You are 8 times more likely to be involved in fatal vehicle accident if you are under 1 second behind the vehicle ahead of you–regardless of speed. The issue is reaction time, which has both fixed and variable components.

  • Ubu7

    “Charles” is all over the place, from the a-bomb to vehicle spacing…

    He must be an algorithm.

    The robocalypse is nigh!

  • Pingback: From Discover Magazine: Can an Algorithm Give You Advice About Your Love Life? · eHarmony Blog()

  • Pingback: What’s Your Risk of Dying Next Year? Wanna Find Out? | Discoblog | Discover Magazine()


Discover's Newsletter

Sign up to get the latest science news delivered weekly right to your inbox!


Quirky, funny, and surprising science news from the edge of the known universe.

See More

Collapse bottom bar