Guys, I fixed global https://v.nostr.build/ZlKP.mov Of course, I also made it impossible for new people to find followers for people who use this feature. What should the default be for new users? I'm thinking 1 — enough to keep people safe, but a low enough bar to allow for discoverability.
Can you share what the text says when you click “WOT Score”? I’m curious how you define that? Is it literally just 1=friends, 2=friends of friends, etc
"WoT (Web Of Trust) is one of the best ways to determine how trustworthy someone might be, based on your social graph. In Coracle, this number is calculated by how many people you follow also follow that person. This allows you to see at a glance if someone is already in your network. This helps reduce spam, impostors, and objectionable content. You can set a minimum web of trust score on your content settings page, which will automatically mute anyone with a lower score than your threshold."
Is that score calculated by an algorithm that we can see somewhere? I feel like the “degree of separation” is more simple but maybe it’s not robust enough?
No, it's very simple, literally just "how many people you follow also follow that person". I'll change the wording to make it less ambiguous.
Would it make sense to have the Web of Trust number scale as the user's follow count increases?
One of the things that makes WoT so challenging is that there is no end to the tweaks that could be implemented; but what may be great for one group of users may not work at all for another. In the short run, simple algos like this one can be very powerful and have wide appeal, so I’m glad to see this WoT implementation in coracle! But I imagine it could be tempting for a dev to try to add too many improvements. Where to draw the line is a judgement call. In the long run, we want our WoT to help us decide which filtering algorithms to use, including how to tweak them and in what context. All decisions, the simple ones and the complex, the little ones and the big ones, ultimately will be farmed out to your WoT (at your discretion, of course). How to build this? Not an easy problem! Most importantly (to me), the fact that YOUR WoT is anchored to YOU (it’s clearly YOUR WoT, not “the” WoT) is of tremendous importance. We want the user to get used to the fact that ultimately, the heavy lifting for all data curation - whether it’s for “safety” or otherwise - must be handled by your WoT, not by the state or The Coracle Corporation or whomever.
Our DBA is actually King Coracle, actually
ackkckcktually 🤦♂️
traditional moderation is already proven, dont reinvent the wheel, but you can improve it, distributed moderation works better than centralized moderation, hardly any nostr app has touched the idea
👀
have you not used iris.to?
I was just talking yesterday about wanting a PoW system to overcome this. I don't think BIP-13 is the correct solution to this though. I think it should be a PoW computed over just the pubkey and a nonce, so it can be attached to any event (in a top level attribute not covered by the signature, though that might require relay cooperation...)
Awesome feature "safe" is the wrong word tho, that's the language censors use. maybe "spam-free" would be better. Or "1 - enough to ensure posts have *some* low level of quality, but low enough to allow for discoverability"
"Safe" is an ok word in this case, since your point is exactly what this feature is trying to accomplish. It's web of trust, so that qualifier is user-defined. But yeah, good instinct for sure.
Disagree and I'll explain why: Safe implies notes outside the wot are dangerous, when really they're just spam. Ideally all the "dangerous" notes (in the sense of challenging your current paradigm, or empowering you to resist authoritarian control) would be in your wot. Censors would like you to think that "dangerous" means it may hurt someone's feelings, and that may be somewhat true, but it will be used as an excuse to also censor things "dangerous" in the good sense. Just my 2 sats, words matter to me a lot, others may disagree l.
Words matter to me as well, I would normally be arguing your position as well. For example, I hate the term "hate speech". I'm just using safe idiomatically here to mean "secure from spam/porn/challenging viewpoints", in a context where "safe" normally means "Trust and Safety" which is equally dystopian.
safer would be a better word, you are trusting what your followers follow is ok anyway distributed moderation is a better system from my view already using this wot on iris.to a lot
I get what you're saying, however it's more than spam though. It's unwanted speech, though that's not the correct word in these examples. Safe is a better description here.
its not unwanted speech as this hides wanted speech, its a trust based filter
unwanted speech? so anything outside of your wot is "unwanted" idk about that. maybe "untrusted"?
Yes. I don't use global now. I haven't for at least 6 montha or more it's a shit show. Implementing a WoT would allow people that want to turn this feature on be able to turn it on and use global once again. Don't like it because you want a full blown fire hose of everything in global? Then don't turn it on. This allows you to get what you want and allows me to get what I want. Win/win.
"trusted" ? "approved" ? If we're going to be pedantic about what this score means, it's really just "popularity" heh.
It's more a friend of a friend of a friend of a friend. And so forth. A network of friends.
Right, but I can't think of a better term than popularity to sum it up. adjacency? heh I think nostr:nprofile1qqsf03c2gsmx5ef4c9zmxvlew04gdh7u94afnknp33qvv3c94kvwxgspz3mhxue69uhhyetvv9ujuerpd46hxtnfduq3xamnwvaz7tmjv4kxz7tpvfkx2tn0wfnszxmhwden5te0vd58y6tnw3cxjmrv9ehx7um5wgcjucm0d5w28rcn implied recursion isn't even in play here so it's nothing beyond friend-of-a-friend?
Right, but I don't think I would call it popularity, since that's independent of your standpoint. It's maybe closer to popularity within friend group, but a better word might be "alignment"
I mean every word so far has been an imperfect descriptor, I wouldn't throw out "popularity" just because it's an imperfect match. "alignment" is probably closer to correct, but it's not very clear to a new user imho. An imperfect understanding of the word "popular" will still lead someone to basically correctly understand the WoT score. "apparent popularity" actually is a near perfect match imho, but it's a bit wordy. "in groupiness" was my first thought after popularity, but I think it's also a bit unclear to your average new user, and it sounds goofy lol
Happy to help you bikeshed the term for the rest of the day instead of doing more important things 😉 😂
Let's go with in-groupiness haha
Last comment on this I swear! What about just calling it a "Score" or "WoT Score" and if the user cares enough to investigate what that is, give them a precise description in a tooltip or something. Otherwise, all the user REALLY needs to know is it's some kind of score, higher is better.
People are so used to hearing these terms used in a dystopian way it's hard to even talk about Nostr's paradigm. "Safe" is a subjective adjective, "unwanted" is a passive verb. The agent isn't specified, and around here it's not who it usually is.
I guess it's somewhat hard to realize that not everyone wants to control you and some are legitimately trying to give control back to you.
safe is the woke misused word
timid is a more accurate word
Maybe I should bump my WoT threshold to 7 to avoid reply guys 🤔 https://i.nostr.build/85jg.png
haha
this filters more than spam and is not optimal
OMG! 👀👀👀⚡⚡⚡ nostr:nevent1qqszjqannxfj5yp2y775jd29jvtq5q7h3xyxwnvetgkdt2mr4jex2dcpz4mhxue69uhhyetvv9ujumn0wd68ytnzvuhsygyhcu9ygdn2v56uz3dnx0uh865xmlwz675emfsccsxxguz6mx8rygpsgqqqqqqsnwta9t
This looks nice. I'm missing global these days!
works fine on amethyst and better than this type of filtering
Care to share what you mean?
you have a list of people whose reports filter out spam etc, currently that list must be your follow list but that will change to a separate list one day you need quite a few people in the list, many hands make light work, amethyst also has a word filter
First real solution for Global? nostr:nevent1qqszjqannxfj5yp2y775jd29jvtq5q7h3xyxwnvetgkdt2mr4jex2dcpremhxue69uhkummnw3ez6ur4vgh8wetvd3hhyer9wghxuet59upzp978pfzrv6n9xhq5tvenl9e74pklmskh4xw6vxxyp3j8qkke3cezqvzqqqqqqyh367pa
How are you labeling the score? Is it a Nostr event as well?
If it's a Nostr event, other clients could implement this 👀👀👀
NIP-32 could work for this.
@hodlbod just implemented a user configurable WoT for global 👀 nostr:nevent1qqszjqannxfj5yp2y775jd29jvtq5q7h3xyxwnvetgkdt2mr4jex2dcpz4mhxue69uhhyetvv9ujuerpd46hxtnfduhsygyhcu9ygdn2v56uz3dnx0uh865xmlwz675emfsccsxxguz6mx8rygpsgqqqqqqsvfdw2p
Sweet! I can't find this in my content settings in coracle.social though, did you deploy this yet?
nostr:nevent1qqs8g5nhgprpx2xtdtg9aygplcrkpfe87axga5nxpp4skwq48g0rjccpr9mhxue69uhhyetvv9ujumn0wd68yam0d3nzucm0d5pzputlnahl3ca9a72mh5sftpnckqscmujay5arfp0h2mv82hnzpr4dqvzqqqqqqyhup4xn
Ok, so maybe I oversold this, apparently nostr:nprofile1qqsy2ga7trfetvd3j65m3jptqw9k39wtq2mg85xz2w542p5dhg06e5qpz4mhxue69uhk2er9dchxummnw3ezumrpdejqz9mhwden5te0dehhxarj9enx6apwwa5h5tnzd9aqz9rhwden5te0wfjkccte9ejxzmt4wvhxjmc70ca3s has had this in Iris since forever. Nevertheless, it feels like a big improvement to Coracle. nostr:nevent1qqszjqannxfj5yp2y775jd29jvtq5q7h3xyxwnvetgkdt2mr4jex2dcpz3mhxue69uhhyetvv9ujuerpd46hxtnfdu2ymx9s
Maybe it could make sense to scale the default based on the number of follows a user has? Seems like there is some trade-off there
note1k2598qfwpxyurvpjj92pfrgawknzypgeskwnduz2dhmd6c0p5fhqpaafk5
Makes sense. I’m guessing this is really a logarithmic effect too. Adjustments by more than 1-2 follows is likely more a question of how strict you want global than it is dependent on your follow count
Logarithms reflect reality way better than linear curves, I wish they were more intuitive to people
Hmmm. x10 photons -> x2 brightness x10 amplitude -> x2 loudness x10 following -> x2 WoT threshold Perfectly natural 😂
How big will the filter be when you follow thousands of people?
To implement this, all you need to do is download follow lists for each person you follow. It's computationally intensive, but easily cached, so not really a performance problem. As far as dialing in the correct threshold, it should be fairly intuitive for users, although there's lots that could be done to make this tunable/more complex.
Call it Tribes. Let people select interests and then followers.
Coracle keeps booting me off relays. I add a bunch, a day later it reverts to 2 or 3. Any reason? Am I doing something wrong? Can't confirm any trigger as of yet (IE if it's after logout, etc).