Stories
Slash Boxes
Comments

SoylentNews is people

posted by martyb on Monday August 24 2020, @06:45AM   Printer-friendly
from the department-of-unwanted-hyperfocus dept.

Researchers at the Cornell and the Technische Univerität Berlin and Cornell have studied the problem that more popular items get priority in search results, creating a positive feedback loop that unfairly deprecates other, equally valuable items.

Rankings are the primary interface through which many online platforms match users to items (e.g. news, products, music, video). In these two-sided markets, not only the users draw utility from the rankings, but the rankings also determine the utility (e.g. exposure, revenue) for the item providers (e.g. publishers, sellers, artists, studios). It has already been noted that myopically optimizing utility to the users – as done by virtually all learning-to-rank algorithms – can be unfair to the item providers. We, therefore, present a learning-to-rank approach for explicitly enforcing merit-based fairness guarantees to groups of items (e.g. articles by the same publisher, tracks by the same artist). In particular, we propose a learning algorithm that ensures notions of amortized group fairness, while simultaneously learning the ranking function from implicit feedback data. The algorithm takes the form of a controller that integrates unbiased estimators for both fairness and utility, dynamically adapting both as more data becomes available. In addition to its rigorous theoretical foundation and convergence guarantees, we find empirically that the algorithm is highly practical and robust.

Journal Reference:
Marco Morik, Ashudeep Singh, Jessica Hong, and Thorsten Joachims. 2020. Controlling Fairness and Bias in Dynamic Learning-to-Rank. In Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR '20), July 25–30, 2020, Virtual Event, China. ACM, NewYork, NY, USA. DOI: https://doi.org/10.1145/3397271.3401100

Maybe this, if deployed widely, can help reduce the tendencies for discourse to develop isolated silos.


Original Submission

 
This discussion has been archived. No new comments can be posted.
Display Options Threshold/Breakthrough Mark All as Read Mark All as Unread
The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.
  • (Score: 4, Insightful) by maxwell demon on Monday August 24 2020, @04:37PM (7 children)

    by maxwell demon (1608) on Monday August 24 2020, @04:37PM (#1041170) Journal

    Results are popular because they are by definition the result people want more often than not.

    Not necessarily. If an item shows up on the first search page, more people will click on it simply because it is readily there, not because it is really the one they would like best. It's quite possible that they would have preferred one from the second page, if they had been shown it immediately. But the algorithm just sees that you clicked on that link, therefore it concludes that this was indeed the right thing to show, and thus it gets better rank, which means it will get displayed even more prominently on searches by others.

    --
    The Tao of math: The numbers you can count are not the real numbers.
    Starting Score:    1  point
    Moderation   +2  
       Insightful=1, Informative=1, Total=2
    Extra 'Insightful' Modifier   0  
    Karma-Bonus Modifier   +1  

    Total Score:   4  
  • (Score: 2) by The Mighty Buzzard on Monday August 24 2020, @05:06PM (4 children)

    Yup, there should be a way to say "no, that appeared useful but was not."

    --
    My rights don't end where your fear begins.
    • (Score: 0) by Anonymous Coward on Monday August 24 2020, @06:51PM (1 child)

      by Anonymous Coward on Monday August 24 2020, @06:51PM (#1041272)

      This is why they want tracking embedded, so the “no” can be inferred by later actions. A users opinion on usefulness means nothing to marketeers.

      • (Score: 2) by acid andy on Monday August 24 2020, @07:28PM

        by acid andy (1683) on Monday August 24 2020, @07:28PM (#1041289) Homepage Journal

        A users opinion on usefulness means nothing to marketeers.

        That's why a good search engine would only be funded by old-fashioned banner ads and its ranking should have nothing whatsoever to do with marketing. Nah, I don't think anyone's going to do it anymore either.

        I know Duckduckgo isn't supposed to track users but it uses results from some search engines that do so, I don't know.

        --
        If a cat has kittens, does a rat have rittens, a bat bittens and a mat mittens?
    • (Score: 1, Insightful) by Anonymous Coward on Tuesday August 25 2020, @02:00AM (1 child)

      by Anonymous Coward on Tuesday August 25 2020, @02:00AM (#1041432)

      Definitely. It pisses me off seeing all those links to resources where I have to pay money or join the site in order to see the full result. Expertsexchange and quora are particularly egregious examples. Not to mention the various random linkfarms that just exist to capitalize on random accidental clicks. Getting those demoted down to the bottom of the list would be great for everybody else.

      • (Score: 0) by Anonymous Coward on Tuesday August 25 2020, @05:16AM

        by Anonymous Coward on Tuesday August 25 2020, @05:16AM (#1041493)

        "Expert sex change", now there's a naming disaster with recipe for disappointment.

  • (Score: 2) by darkfeline on Monday August 24 2020, @09:03PM (1 child)

    by darkfeline (1030) on Monday August 24 2020, @09:03PM (#1041325) Homepage

    >the algorithm just sees that you clicked on that link

    See, that's where you're wrong. Real Search Engines also track engagement (it's partly why they've sunk their tracking tentacles into everything). If the site you visited is using Search Analytics, Search Ads, Search Like Button, etc, then it can track how long you spent on the page, whether you scrolled or moved the mouse, etc. Each search page also has tracking, so if you go through multiple links with an interval in between, it knows the first one you clicked wasn't useful.

    If you add Machine Learning and a huge number of data points, you get a pretty good signal of what's the Best, although keep in mind that people don't know what good for them. Of course, if the Search Engine tries to decide what's Best for them, then it becomes a Political Agenda. You really can't win.

    --
    Join the SDF Public Access UNIX System today!
    • (Score: 0) by Anonymous Coward on Tuesday August 25 2020, @05:14AM

      by Anonymous Coward on Tuesday August 25 2020, @05:14AM (#1041492)

      They just need to listen on your computer microphone for the inevitable "fuck this site pisses me off", "now that was a waste of time", "fucking pop ups!!!", "why the fuck is this site in the first page", or "there's 10 minutes I'll never get back".