https://sebastianspicker.github.io/tags/opinion-dynamics/ Recent content in Opinion-Dynamics on Sebastian J. Spicker https://sebastianspicker.github.io/og-image.png https://sebastianspicker.github.io/og-image.png Hugo -- 0.160.0 en Sat, 24 Oct 2026 00:00:00 +0200 https://sebastianspicker.github.io/posts/who-raises-whom-recommender-feedback/ Sat, 24 Oct 2026 00:00:00 +0200 https://sebastianspicker.github.io/posts/who-raises-whom-recommender-feedback/ De Kai’s publisher calls the recommender systems that watch us “tweens”. A feed is less like a child than like a loop: it is trained on behaviour it has itself shaped. This post follows that loop through performative prediction, a small opinion-dynamics model with a toy simulation of my own, the field evidence on feeds, filter bubbles and YouTube, chatbot sycophancy, and model collapse. The mechanism is real, and its measured effects on opinion are small or mixed. The most explicit documented case of a model shaped by its users’ approval is a chat assistant, though I know of no study that has watched such a loop close over time.