End User Authoring of Personalized Content Classifiers: Comparing Example Labeling, Rule Writing, and LLM Prompting
arxiv.orgSaved by Sam Liebeskind
End User Authoring of Personalized Content Classifiers: Comparing Example Labeling, Rule Writing, and LLM Prompting
Instead of asking whether or not a platform optimizes for engagement, it can be more helpful to ask how explicit or implicit these engagement signals are, and provide users with more controls to allow them to tweak their online experience.
I don’t know how to currently form my thought around these trend I’ve notice of new social / media apps (those that are social and media focused but not content slop). Basically social medias that are build in the same vein as are.na and substack.com
Some users may not want to see any potentially abusive content at all, while others may, for example, need to sift through banal personal attacks in order to track credible death threats. The point, however, is that users themselves should be able to exercise significantly more agency over their own social media experience.