It's not hard to only count likes/zaps/comments from users in a web of trust (ie only count likes within 2 degrees on my follows)
Missing out on fire notes because we can only order by date is ngmi
Discussion
I agree with you.
Right. It's not that all algorithms are bad. It's the lack of transparency and choice that's bad. We can solve this.
Yes. And. Followed != trusted.
So WoT on Nostr will need better measure of “trust” to THEN provide the backbone for transparency and choice in the algos applied. Working on this also… for social onboarding.
nostr:note1ta9gzmtewn7jnslhjd8ew9322qp3rnyc56e9a0rj9f0x2ftndxhqcy7dwt
In more-speech I differentiate between follow and trust. I trust my contacts. I follow people and threads by collecting the events in filtered tabs.
From: manime<-mazin at 05/08 12:27
> Yes. And. Followed != trusted.
>
> So WoT on Nostr will need better measure of “trust” to THEN provide the backbone for transparency and choice in the algos applied. Working on this also… for social onboarding.
> nostr:note1ta9gzmtewn7jnslhjd8ew9322qp3rnyc56e9a0rj9f0x2ftndxhqcy7dwt
CC: #[4]
CC: #[5]
CC: #[6]
i don't think it should be a public announcement who i trust, what do you think?
What’s more-speech?
https://github.com/unclebob/more-speech
From: manime<-mazin at 05/08 14:00
> What’s more-speech?
CC: #[4]
CC: #[5]
CC: #[6]
CC: #[7]
CC: #[8]
https://www.youtube.com/watch?v=q3gQ42aUhls&list=PLB2PP9uiaifzD0ULVx03AkFK8S-2P-q4y&index=2
From: manime<-mazin at 05/08 14:00
> What’s more-speech?
CC: #[4]
CC: #[5]
CC: #[6]
CC: #[7]
CC: #[8]
The theory of operation of more-speech is:
1. Relays are categorized as trust, web-of-trust, or all.
* trust: Only events from those who are trusted (kind 3) are subscribed.
* web-of-trust: Only events from those who are trusted, and those whom they trust are subscribed.
* All: Every event is subscribed.
Users make this choice based upon the volume of the relay. A very high volume relay will have to be constrained to just 'trust'. Moderate volume relays can be relaxed to 'web-of-trust'. Lower volume relays can be set to 'all'.
2. Every event is subject to a set of filters associated with tabs on the display. The filters within a tab can select or block based upon the ids of users, events, or the content of events. So, for example, I can set up a tab that will select for all events mentioning GAZA while also blocking certain nasty people based on their ids. I can also set up a tab that selects for any event sent as a reply to a particular event id -- thus capturing the thread of an event.
From: manime<-mazin at 05/08 14:00
> What’s more-speech?
CC: #[4]
CC: #[5]
CC: #[6]
CC: #[7]
CC: #[8]
I'm late to reply here but kind3 of a trusted npub is good enough to assume that an npub isn't a spammer which is all I care about at the relay level. Sort of a bare minimum trust implemention that can easily be improved but already does a lot for us when it comes to stopping spam
This is a hot topic. Every time I post about it people have interesting takes. You are not alone in thinking “follow is good enough”. And still others have much more elaborate ideas for measuring trust. I should host some more panel discussions …
algorithms can be good for you