Some things are incredibly appealing to everyone and also bad for society. We have to treat those things responsibly.
Recommendation algorithms can be useful, to assist you in discovering content. But only as a tool that you can choose to use. If I can select a person that I like listening to and get a list of other people who I may be interested in (assuming that the algorithm is simply matching me to similar peers and not also adding in some “also Elon/Bezos/whoever really wants you to see these guys” skew)… that would be a useful tool.
However, the recommendation algorithms should not be used to make the second-by-second decision about what you see next. The next item in your feed should always be there because of a decision that you make, not as a means of “maximizing engagement” + whatever skew the owner wants to add.
Of course people like these features, these algorithms are literally trained to maximize how likable their recommendations are.
It’s like how people like heroin because it perfectly fits our opioid receptors. The problem is that you can’t simply trust that the person giving you heroin will always have your best interests in mind.
Recommendation algorithms are a useful tool but, only when used in moderation. Attaching a recommendation algorithm directly to your brain via a curated content feed is incredibly unhealthy for both the individual and society.
Treating it responsibly in this case would mean actually offering a recommendation algorithm that is free of corporate interest, then. To go along with your own simile, you can’t really go up to a junkie and say “Hey, you should really consider giving up heroin and having a salad instead. It’s better for you.” and expect it to be a convincing argument. Which is why Bluesky is succeeding and Mastodon isn’t.
It also means decoupling the recommendation system from people’s feeds.
Having a “you may like this” section is a lot less abusable than “the next item in your doomscroll is <recommendation>”.
Bluesky is just another Twitter. Everything that happened to Twitter can happen to Bluesky. It’s not fundamentally changing anything except trading Elon for a different owner.
It’s not a bad change, people want Twitter after all… but it isn’t fixing any problems in the underlying incentive structures or algorithm control.
The core problem is that curated feeds allow the owner to substitute their recommendations in place of recommendations that would interest you.
Until the owner can’t do that, the social network is always one sale away from being the next Twitter/Truth Social.
Bluesky is fixing social media by changing the owner, Mastodon/ActivityPub is fixing social media by getting rid of the owner.
I think the latter is the better choice for how to structure these things.
Some things are incredibly appealing to everyone and also bad for society. We have to treat those things responsibly.
Recommendation algorithms can be useful, to assist you in discovering content. But only as a tool that you can choose to use. If I can select a person that I like listening to and get a list of other people who I may be interested in (assuming that the algorithm is simply matching me to similar peers and not also adding in some “also Elon/Bezos/whoever really wants you to see these guys” skew)… that would be a useful tool.
However, the recommendation algorithms should not be used to make the second-by-second decision about what you see next. The next item in your feed should always be there because of a decision that you make, not as a means of “maximizing engagement” + whatever skew the owner wants to add.
Of course people like these features, these algorithms are literally trained to maximize how likable their recommendations are.
It’s like how people like heroin because it perfectly fits our opioid receptors. The problem is that you can’t simply trust that the person giving you heroin will always have your best interests in mind.
Recommendation algorithms are a useful tool but, only when used in moderation. Attaching a recommendation algorithm directly to your brain via a curated content feed is incredibly unhealthy for both the individual and society.
Treating it responsibly in this case would mean actually offering a recommendation algorithm that is free of corporate interest, then. To go along with your own simile, you can’t really go up to a junkie and say “Hey, you should really consider giving up heroin and having a salad instead. It’s better for you.” and expect it to be a convincing argument. Which is why Bluesky is succeeding and Mastodon isn’t.
It also means decoupling the recommendation system from people’s feeds.
Having a “you may like this” section is a lot less abusable than “the next item in your doomscroll is <recommendation>”.
Bluesky is just another Twitter. Everything that happened to Twitter can happen to Bluesky. It’s not fundamentally changing anything except trading Elon for a different owner.
It’s not a bad change, people want Twitter after all… but it isn’t fixing any problems in the underlying incentive structures or algorithm control.
The core problem is that curated feeds allow the owner to substitute their recommendations in place of recommendations that would interest you.
Until the owner can’t do that, the social network is always one sale away from being the next Twitter/Truth Social.
Bluesky is fixing social media by changing the owner, Mastodon/ActivityPub is fixing social media by getting rid of the owner.
I think the latter is the better choice for how to structure these things.