{"slug": "social-media-algorithms-accumulated-aesthetic-artifacts", "title": "Social media algorithms: accumulated aesthetic artifacts", "summary": "Social media algorithms function as software agents that build personalized aesthetic profiles through user interaction, creating unique digital artifacts representing individual tastes. The author argues these accumulated preference vectors demonstrate the need for data ownership, as corporations profit from user-generated aesthetic data.", "body_md": "A few years ago, before the agent hype was quite so feverish, I had\nbegun to work on a blogpost about how social media algorithms map quite\nneatly onto the older “software agent” concepts of the 1990s ([and\nbefore](https://en.wikipedia.org/wiki/Software_agent#History). Essentially, the idea of those earlier agents is that\nthey’re a self-contained, interactive objects with internal state and\nthe ability to communicate.\n\nWhich is really just the same as a social media algorithm - there’s two-way communication (e.g. you like, linger on, or share a piece of content (which tells the algorithm you like it) and then it shows you more), they’re relatively self-contained, they’re interactive, and they have state.\n\nThat state gets improved over time through re-inforcement - there is a special set of numbers, vectors that represent your own historical behavior and preferences. This is the meat-n-potatoes of a recommendation algorithm. Those vectors get updated over time, and eventually, they accrete into an artifact.\n\nThat artifact is what so intrigues me!\n\nSomewhere, on Instagram’s servers, there’s a magic packet of numbers that describes my aesthetic interests. I want it! Let me see it - there’s something for me to discover there.\n\n*Aside* from my own selfish interest, I bring this up because\nI’m fascinated by the way social media might have changed our brains. I\nwas thinking earlier today about the profound difference in aesthetic\nmovements between the 1990s and the current day, mostly because I think\n(based on no particular real *data*) that the average person\ntoday is much more aesthetically consistent.\n\nI *don’t* mean that people today are more fashionable than\npeople in the past, or that everything looks the same. Rather that on an\nindividual level, people are more aligned with a particular aesthetic\nphilosophy than in the past. Social media and online shopping has raised\nthe threshold by which people are evaluated on their aesthetic taste,\nand reduced cost of manufacturing has made it feasible to make extremely\nniche products targeted at specific aesthetic cohorts.\n\nThe social media algorithms encourage a sort of aesthetic echo-chamber, where-in people are exposed to a gradually narrowing conception of what is fashionable. At the end of it all, you have a crystalline shell of user embeddings that represents a distinct set of aesthetic interests - an artifact that describes what you like.\n\nI think this is maybe one of the most compelling arguments for data-ownership. If I can’t have that data, why should a mega-corporation, one almost certainly aligned against my interests?\n\nsome links that have been percolating to produce this post:", "url": "https://wpnews.pro/news/social-media-algorithms-accumulated-aesthetic-artifacts", "canonical_source": "https://henderson.lol/posts/202607-social-media-algorithms-accumulated-aesthetic-artifacts", "published_at": "2026-07-19 01:03:48+00:00", "updated_at": "2026-07-19 01:21:07.753907+00:00", "lang": "en", "topics": ["artificial-intelligence", "ai-ethics", "ai-agents"], "entities": ["Instagram"], "alternates": {"html": "https://wpnews.pro/news/social-media-algorithms-accumulated-aesthetic-artifacts", "markdown": "https://wpnews.pro/news/social-media-algorithms-accumulated-aesthetic-artifacts.md", "text": "https://wpnews.pro/news/social-media-algorithms-accumulated-aesthetic-artifacts.txt", "jsonld": "https://wpnews.pro/news/social-media-algorithms-accumulated-aesthetic-artifacts.jsonld"}}