I realised after a while that one of the main reasons I go online is to expand my horizons. I'm actively looking for something new or interesting or surprising that I might not have encountered or thought about before. I am also looking for perspective - give me a broad picture of how the world is so I can calibrate and introspect where my own perspective differs and could be updated. I want to see the things that don't fully align with my ideas.
So it turns out personalised feeds are actively curating against what I'm looking for. I'm curious if others are like me or is this just something weird about me. I think it is very much why I like Hacker News format. It's in the broad areas of my interests but it is anything but curated just for me.
The vast majority of signal is from a plurality of opinions, and the outcome is an average of those inputs, usually tweaked a bit to make them more addictive.
Your thought that “the algorithm doesn’t serve me personally because I have exceptional taste” is not a unique observation. Everyone who consumes these systems has the same thoughts. Everyone is a rounding error, whether you’re in a knitting community, or you’re a worlds leading expert in deep learning.
I suspect what's going on is we're experiencing the long-term effects of AI recommendation algorithms getting plugged into everything. I'm not anti-AI. I think neural nets are incredibly powerful tools and I really like using them as an aide for my coding work. But they clearly do only a first or second level of pattern matching(asterisk), and I think that has a fundamental impact on their overall usefulness and desirability.
I can't explain this any better than Melanie Mitchell in her "AI for Thinking Humans," so I'd generally recommend that book to really understand the issues. But one way of thinking about it is that the AI keys in on patterns that are indeed there in its training data, but very often don't pick up on the same patterns as a human or treat patterns "not in the same way" as a person would. So you get the uncanny valley effect where the recommendations share some features of things you like and find interesting, but often miss the "soul" of the thing.
A good example of this is I recently watched the movie "X-Men" (2000). I then got a bunch of superhero movie recommendations. On the surface, this makes total sense. I watched a superhero movie, I get more superhero movies. But the reason I like that movie is it portrays racism/nationalism in a way that isn't totally stereotyped while still being PG-rated and easy to watch, without the seriousness of something like "American History X" (it was also a Patrick Stewart tribute). It's very unlikely that an AI can make that level of inference. I actually would enjoy watching more movies like that (as opposed to your point about wanting variety and something more exploratory), but what are the odds it's going to recommend some vague theme like that instead of "more superheroes"?
I also agree with your point that the underlying assumption we all want "more of the same" is a big thing that makes the modern internet feel more narrow. So you put those 2 together and it feels like we're being put into a narrow Skinner box that approximates aspects of things we like, but is really more of a confined environment designed to extract our time and attention while only fulfilling the shallowest, barest minimum of our wants and desires.