Before the scaled sort was introduced, I had hoped it would provide a solution similar to the “Hot” or “Top” filters but without so many memes and political posts. Unfortunately, the scaled sort seems ineffective as most posts appear with a single vote, making it practically the same as the “New” sort. Although I’ve banned the largest communities, I still want to see some of that content occasionally.
The developers have closed all issues related to the scaled sort, even though it fails to address the issues raised in several discussions:
- Rework “Hot” sorting to show posts from more varied communities
- The rank of a post in the aggregated feed should be inversely proportional to the size of the community
- Is there any way to reverse degrowth of the niche communities on Lemmy?
- I hate to say it but I haven’t been very active on lemmy, but I want to be
Personally, I believe the best way to address this issue is through the implementation of tags and custom feeds. With post tags and custom feeds, users could create separate feeds tailored to their preferences by subscribing to a few communities and blocking specific tags or keywords. However, this would require an incentive system similar to imageboards like Safebooru, with a leaderboard to encourage accurate post tagging by users, as also mentioned in The Great Monkey Tagging Army: How Fake Internet Points Can Save Us All!
Do you have any ideas or suggestions on how Lemmy could better surface content from smaller communities?
I just want a superintelligent AI that always talks to me about the things I like to hear. That’s a bit far-fetched, but I also thought it was possible before: Algorithm Marketplace
This is not possible because sorting is done in the database, so adding a new sort option requires a database migration with new indexes, columns and updated queries. Not something that can be done with a simple plugin.
Ah that’s a shame
Actually scores are regularly calcuated from a scheduled task which runs in Rust. Yes the score caculation is currently implemented in SQL, but it could also be changed to a Rust implementation or a plugin. This would probably need some optimization so that the plugin only calculates scores for recent posts, not every single known post. In any case it would need someone with the time and motivation to implement it.