This is an attempt to mitigate spam on the platform by focusing the recommendations a user receives on content that is relevant within their network. This way, users generating spam should have a harder time reaching potential victims since they are unable to infiltrate the victim's social network.
Basically the system should work like this:
- The "victim" follows a user that they know and trust. This user will be handled as a "trust source".
- The trust source then interacts with content, which increases the chances of it being recommended to the victim.
- Whenever the victim logs into the application, the system generates a list of recommendations that it can then use to suggest new content to the victim.
- This could also be repeated with a layer up or down from the trust source, so that content can 'skip' a level.
The system should consider content only if:
- The interactions are fresh and do not exceed a certain threshold.
- The content is fresh. This can be a relative freshness (maybe something like 3-6 months so the page does not become stale).
In order to prevent the user from feeling out of control, this could be moved to a "discover" page, where content they might like is aggregated.