How “Not Interested” Affects Recommendations
Not Interested is perhaps the clearest example of explicit recommendation feedback. The current Phoenix model includes a not_interested prediction, and the ranking system treats it as a negative signal. (Phoenix source)
The basic idea is: “This user doesn't want this type of recommendation.”
Not Interested vs. No Engagement
These should not be treated as identical:
The second gives the recommendation system a much clearer preference signal. The current architecture explicitly models not_interested rather than relying only on the absence of positive engagement. (X Algorithm repository)
Comparing the Four Negative Signals
| Signal | What it tells X | Main distinction |
|---|---|---|
| Block | User rejects the author | Strong author-level negative action |
| Mute | User doesn't want the author's content | Author-level preference |
| Report | User flags content/author for potential violation | Can connect to safety/visibility systems |
| Not Interested | User doesn't want this recommendation | Direct recommendation preference |
All four are explicitly represented in the current Phoenix prediction architecture. (Phoenix runner)
Negative Signals Don't Work Like a Public Penalty Table
The current open-source architecture uses predicted action probabilities multiplied by configurable weights. The weights aren't a public creator-facing penalty chart. (X ranking scorer)
And even this is only a simplified representation of the broader recommendation pipeline.
Under The Hood Takeaway
Block: A negative author-level signal and a filtering consideration for the affected user.
Mute: A negative author/content preference that can affect recommendations for the affected user.
Report: A negative prediction that can also intersect with separate safety and visibility systems.
Not Interested: An explicit recommendation-preference signal telling the system that the user doesn't want that type of content.
The key point is: These signals don't prove a universal penalty against your account. They provide information about whether content is appropriate or desirable for particular users, while some actions can also trigger separate filtering or safety mechanisms.
How X Ranks and Recommends Posts
Official Open-Source References
- • X / xAI — For You Algorithm repository — current architecture, Phoenix, Thunder, filtering, scoring and candidate pipeline. (GitHub)