[zeusflight]

Reputation scores risk turning the rooms into an echo chamber due to karma mining. It's good for forming consensus, but bad for debates. This can be bad for some communities like FOSS groups where new users and people with fresh ideas frequent.

[komali2]

Matrix will apparently invent ui to warn you if your filters are creating an echo chamber but I have no idea how that would look or work.

EDIT: Fucking obnoxious post throttling. I don't know why HN doesn't want me posting anymore, so I guess I'll just abuse the edit feature to get my voice heard.

In reply to zuesflight:

> What if the feed manipulation was per-user?

My understanding from Matrix's blog post is that is half of the filtering equation. 1. Users and 2. Servers can filter 3. other users via reputation. It sounds like these filters will be inspectable, so, I would imagine that if you don't like how a server is filtering, you'd just not use it. I don't really have a strong understanding of Matrix (is it a protocol, a company?) so this is just me firing from the hip from the blog.

I also think the filter "Merge" thing is all about making "multiple points of attack."

None of this is implemented from what I understand, all they have now is a sort of binary banned/not banned list they share with mozilla. Very pie in the sky but if it works as they write that'd be cool!

[zeusflight]

I missed that point earlier. Thanks! However, it still sounds complicated. Granted it's open source. But I feel that the fundamental reason why we are in this social media conundrum is the on-cloud platform-wide message feed manipulation. What if matrix and other social media were just channels that didn't manipulate the feed in anyway (no filtering or sorting)? What if the feed manipulation was per-user? On client-side like a personal rspamd, or as per-user on server profiles like sieve? There could be several competing filters, all of which could be adjusted by user to varying degree. Some filters could be learning ones and some could be for special cases (like child filters). There would be no single point of attack, and therefore harder to do social engineering by manipulating the feeds.

[Arathorn]

> What if the feed manipulation was per-user?

> Or client-side like a personal rspamd

> Or as per-user on server profiles like sieve?

> There could be several competing filters, all of which could be adjusted by user to varying degree.

> Some filters could be learning ones and some could be for special cases (like child filters). There would be no single point of attack, and therefore harder to do social engineering by manipulating the feeds.

This is precisely what we're trying to propose in the original article :)

In terms of UI, visualising the filters could be as simple as "98% of the rooms in this list have been hidden by your #nsfw filter", letting you peek behind the filter to see what you're missing, etc.

[zeusflight]

Great! Does that mean that I can design and plug in a custom feed filter logic?

[Arathorn]

Sure! you could get experimenting on the current basics via https://github.com/matrix-org/mjolnir today.

[zeusflight]

(Answering to the second part of you reply) Either I read the article wrong or it wasn't obvious. I really hope that matrix designs the controls to put the decision fully in the hands of the users.

> I don't really have a strong understanding of Matrix (is it a protocol, a company?)

Matrix is definitely a protocol. Element is the company driving its development.

(PS: I wonder why you are downvoted for a meaningful reply)