The “downvote to disagree” thing isn’t just an attitude problem, it’s a structural issue. No amount of asking people nicely to obey site etiquette will change the fact that the downvote button is a disagree button. If you don’t want a hive mind, you necessarily need to be able to allow for things you don’t like to be amplified.

Actually, with enough interactions from different people (ie: enough data points) Lemmy should be able to determine if a comment brings value to the conversation (either positive or negative) or if it's noise that should be ignored (and prioritized lower).

If you have 4 comments:

1. Has 100 upvotes (in total)
2. Another has 100 downvote (in total)
3. Another has 50 upvotes and 50 downvote (100 in total with a 0 sum)
4. The last was a new comment with 0 votes.

It's obvious that 1 and 3 are providing more to the conversation than 2. 4 is a bit of an outlier, but probably provides more value than 2.

Regarding 3: The challenge would be that there's a low chance that there will be such a wide margin of upvotes/downvotes. Due to the hive mind, the voting will probably look like 30 upvotes and 130 downvotes. So, there would need to be a weight accordingly, so those fewer upvotes had a greater impact (in terms of sorting and scoring comments)

Reddit has a "sort by controversial" algorithm that seems to be missing from Lemmy (or maybe it's hidden in the “what's hot" - I haven't looked at the code).

It would be awesome (and resource intensive) if Lemmy could provide the federated instances with custom sorting algorithms. It would allow federated instances to be unique, provide some playful competition, and given the open source nature of Lemmy - I'm sure these algorithms would be open sourced, which would improve the entire Lemmy ecosystem as a whole.