Algorithmic Prediction Loop / 算法预判循环

Algorithmic prediction loop is the pattern where a platform predicts what the user wants, observes the user’s reaction, then uses that reaction to revise the user model and future recommendations. 159.算法的六副面孔：它是如何从处理数据，变成定义我们是谁的 describes this as the second face of algorithms: the user has a body in front of the screen and a data-made “digital twin” inside the system.

The source’s important correction is that users are not simply training the algorithm. The interface also trains users by deciding which feedback buttons exist, where they appear, what counts as liking, skipping, staying, or disinterest, and which behaviors become stable signals. Prediction therefore compresses a person into measurable behavior while nudging the person to behave in more measurable ways.

Key Claims

• Personalization is not pure understanding; it turns complex people into operationally useful predictions.
• Feedback buttons, watch time, search history, location, timing, and similar-user patterns are part of one behavior loop.
• The loop can create Algorithmic Desire Preemption / 算法欲望预支 when feeds show wants, lifestyles, or routes before the person has chosen them.
• The “choice disappearance” problem appears when highly rated or highly visible options become the default answer, as in the source’s Iceland travel example.
• Prediction loops can make platforms feel intimate while still being shallow about the full person.

Connections

• Public Relevance Algorithms / 公共相关性的算法 — umbrella category.
• Recommendation System Productization — product system that converts prediction into a feed or recommendation experience.
• Personalization As Social Identity — related pattern where usage data becomes a story about the user.
• Algorithmic Desire Preemption / 算法欲望预支, Attention Industrialization, and AI Consumer Decision Shaping — downstream attention and desire effects.
• Algorithmic Diversity Dividend / 算法多样性红利 and Feed Curation — user-side responses when prediction becomes too narrow.