Is there such a thing as "Web of Distrust" where we look at reputable keys who have shared/common muted users?
Essentially asking my followers who I should mute
nostr:npub176p7sup477k5738qhxx0hk2n0cty2k5je5uvalzvkvwmw4tltmeqw7vgup
#asknostr
Discussion
yes, but I don't think it would be as useful as the default contact lists graph since people mute other users for a bunch of reasons. whereas users generally follow other users only because they want to see their content
Right, I think a pubkey would need to be muted by multiple reputable keys in order to be labeled a "bad actor"
Because you're right that people get in arguments and rage mute users that may otherwise be interesting to follow
This is the way to go about it. The data will always be messy but you can take multiple signals from multiple sources to turn weak signals into strong signals. Don’t take **just** the mute lists either, but it can be a good start.
by the time you get to a number that wouldn't result in most users having me muted, you also wouldn't have actual bad npubs muted, because I get muted way more than npubs that are actually bad
It might be useful if it only recommended muting npubs that had been muted by a threshold of, say, 5+ of those you follow. That would weed out most random reasons folks might mute someone and leave only those who have been consistently muted. Could even have a user customizable threshold, if they feel 5+ is too aggressive.
5+ is way too aggressive and by the time you get to a number that wouldn't result in most users having me muted, you also wouldn't have actual bad npubs muted, because I get muted way more than npubs that are actually bad
I don't think it makes sense to recomending muting npubs, since your mute list isn't something you really want to keep growing. unlike your follow list
That said I'm not sure what else a graph if muted users could be used for... 🤔
This is true, but I would still like to leverage my follows' judgement in this case. Without it being used to create my own list that lives on a relay, it would have to be something handled client-side that filters out posts from users who are often muted and/or reported by my follows.
I feel like how nostr:npub1jlrs53pkdfjnts29kveljul2sm0actt6n8dxrrzqcersttvcuv3qdjynqn has implemented web-of-trust filters that takes muting and reporting from my follows into account is a decent approach. I can then make it as aggressive or as lenient as I want it to be for filtering posts, and yet it isn't requiring me to add anyone to my mute list to achieve this.
Amethyst can do something like this, as I understand it.
If you turn on the option to "warn when posts have reports from your follows," amethyst will display a warning in place of a post when a threshold of those you follow have reported the author. You are free to dismiss the warning and view the post anyway.
Wasn't aware of this Amethyst feature, but when I checked it, I found out that it was enabled and had a number of accounts on the blocklist already.
Explains why I see so little spam and annoying stuff :-)
It's an extremely helpful feature that nostr:npub1gcxzte5zlkncx26j68ez60fzkvtkm9e0vrwdcvsjakxf9mu9qewqlfnj5z has unfortunately received a lot of flack for, because some have considered it "censorship." It's not perfect, by any means, but not having to manually kick every troll out of your feed because you can offload some of that to those you follow who have already muted and/or reported them is a major help in keeping your feed and your replies filled with things you actually want to see.
I am not keen on preemptively muting people, no matter what the 'source of truth' is. If an npub becomes annoying it is easy enough to mute them there and then. The ones that would be nice to mute in advance e.g. the reply bots cannot be muted like this cause they just spawn a new key every time.
Would be an interesting opt in!
modelling distrust is challenging, as it has different properties from positive relationships (e.g. follows).
A follow of a follow is not a stranger
yet, a mute of a mute is unclear.
It's also not obvious how to add mutes to random walk based algorithms, because probabilities are always positive, and can't be negative. Anyway, I have a plan on how to do it.
Not a good idea. The npubs people think of as reputable aren't actually trustworthy, a huge number of them have me muted