• jarfil@beehaw.org
    link
    fedilink
    arrow-up
    16
    ·
    1 year ago

    Mastodon does have an algorithm that makes people see their posts: it’s called following someone. How does one make others follow them? Well, maybe post something they’d like to see more of.

    • ShadowPouncer@kbin.social
      link
      fedilink
      arrow-up
      5
      ·
      1 year ago

      Mastodon absolutely does have a weakness of making it more difficult to find people that you want to follow based on what you have already engaged with.

      And from a purely user perspective, that is a weakness.

      But it’s also a very distinct choice. Because having enough data to be able to meaningfully make such recommendations means having a central database of every user interaction by every user.

      And it also means making choices and value judgements which, almost by definition, can not be value neutral.

      If the creators of the algorithm are good, they will actually be aware of the choices and value judgements being made, if not, well… They will still be making them, just not in nearly as educated of a way.

      On the whole, I really hope that we eventually come up with answers to these problems that make it possible for a user to make those choices, and to have the amount of recommendations that they want, while somehow not having anyone have the huge database of user interactions. I’m not sure if that’s even possible, most especially if you assume that there will be entities on the fediverse that are fudging their data to get recommended in ways that other users don’t want.

      But it sure would be interesting to try.

      • jarfil@beehaw.org
        link
        fedilink
        arrow-up
        2
        ·
        1 year ago

        Ah, recommendations… hm.

        Just thinking out loud: what if, anyone who wanted to, could take a hash of their username, and make public all their interactions by listing the hashes of the usernames they interacted with. Maybe store it in a distributed database. So everyone could make a graph of anonymous hashes to run a recommendation algorithm in any way they wanted, but each one would only know their own username’s hash, so they could find out which hashes they could find personally interesting according to their chosen algorithm. Then, have people who wanted to be discovered that way, publish their user along their hash, so someone who found their hash through the graph of anonymous hashes, could find out which user it belongs to, and see them as a recommendation.