The problem with web of trust, is that it assumes I trust the people I follow. Maybe I'm just the edge case where this is not true.
Login to reply
Replies (27)
I’d suggest that we apply a naive model of WoT when we look only at follows and followers. @david is doing much deeper work to build a more meaningful web of trust.
I think as long as it's opt-in, I don't see any issues. But if I'm forced to see what my "friends" are following, no bueno.
you're not alone
Trust is context dependent. One trust does not fit all.
There's a use case for actively following people you DON'T trust. WoT by default probably messes with that.
💯
This is true.
Need a system where we can vouche for who we trust. As discussed with @franzap in Madeira
Web of Trust should not be assumed to be web of follow. I agree. Also we trust different people for different things.
Yes, absolutely. I follow two people who I know are complete nutcases. I definitely don't trust them but I enjoy it from a psychological standpoint.
I'm working on something related to this
@david also working on this with the Tapestry protocol
if you follow that user it is assumed that you want to see their notes
seeing their notes is not trusting them, is it?
You’re not alone. Just like my IRL friends, the people I follow I generally only trust in a few categories and may explicitly distrust in others 😂
It would be so nice to have some granularity. Like I want to see their notes about A but not B.
As it stands now, with following and muting being the only tools, some very inelegant choices get made.
I trust you to post interesting content. That's all. Anything more requires applying weighting to categories of trust.
we should moderate that, it's useful but boring
This is not a problem with WoT. This is a problem with 15 years of “follow” implemented by “walled garden” socials where trust is taken for granted.
The problem is that “follow != trust” and this matters even more on Nostr.
Sovereign Webs of Trust…
Curious?
*Read this article for my short take.*
https://yakihonne.com/article/naddr1qqgx2e3cxqenqef3x56kvd3nxymxzq3qmanlnflyzyjhgh970t8mmngrdytcp3jrmaa66u846ggg7t20cgqqxpqqqp65w7jk7e3
*Comment bellow to offer your own.*
1. What does Sovereign WoT mean to you?
2. Does it require “is trusted” (for a followed account) to be explicitly set by the user, or could “is trusted” be ONLY calculated by algos and still count as sovereign?
3. If explicit is required, can “is trusted” simply be a “follows” (kind3) list, or does it actually need a dedicated ( private event kind) “trusted” list?
4. Would (your) Sovereign WoT require that content filters follow users across clients, and be sharable by them?
5. Should these filters (for end users) be publishable ONLY by clients or ONLY by relays or ONLY by DVMs or ANY and ALL of these?
View quoted note →
🎯
We should wean ourselves slowly off of trust proxies like follows and mute lists and replace them with explicit, contextual trust attestations. Something like this:
GitHub
tapestry-protocol/guides/grapevineIncorporation/steps.md at main · wds4/tapestry-protocol
Contribute to wds4/tapestry-protocol development by creating an account on GitHub.
Yes … tapestry is nice … you have obviously put lots of time and effort … BUT it does carry a lot of assumptions about “trust” that might be better left to the free market.
IOW… actual WoT NIP for nostr could be even simpler… describing a standard for subscribable filters by which end users could choose the “trust” assumptions that work for them.
Scalable WoT solution will be simple and not cumbersome for developers or end users… with minimal assumptions about HOW trust should be determined.
Scalable WoT definitely needs to be intuitive and easy for the users. Design is going to be of utmost importance.
As for the developers, the question is how to make it as easy as possible and still get the job done. There is a reason that WoT has failed to live up to expectations for the > 3 decades we have been trying to make it work, ever since PGP was introduced in 1991. In my mind, WoT is like an airplane, in the sense that if we want it to fly, there are certain elements that must be in place. Omit just one essential element and it will never get off the ground. For too long we have been too lazy to accept that fact and figure out what those elements are. And that’s what the tapestry protocol is about.
Spot on.
IMHO … the essential elements are:
- “is trusted” should be explicitly determined by end users.
- “filters” that operate from “is trusted” should be subscribable and sharable by end users across all apps.
Using this simple architecture, “filters” may “ingest” any data or metadata from the network to satisfy for ALL of the diversity of “WoT algorithms” that smart people like yourself have designed.
- use filters for content feeds
- use filters for recommendations
- use filters to manage the “is trusted” list itself, so a user doesn’t have to. 🤯
What are your essential elements?
Agree with what you said.
This is not exhaustive, but a few essential elements:
- explicit trust attestations need to be contextual
- trust (or lack thereof) in a broad context automatically implies trust (or lack thereof) in all subcontexts
- the list of contexts and their relationships must ultimately be curated by one’s web of trust (but ok for devs to manage these in early product iterations)
We don’t need to roll out all of the above in one fell swoop. We can and should roll them out in baby steps. Builders who know DESIGN and who know PRODUCT will be CRUCIAL in figuring out how to roll things out, one step at a time, in a manner that will be accepted by users!
Imagine having a list of filters and tools that allow your WoT to recommend filter A (or some set of filters) to be good ones to use for some given purpose. e.g. my Grapevine recommends Filter A to show me a list of financial armageddon movies from an orange pill perspective
Agreed. “Contexts of trust” does not need to be in the standard. A WoT NIP should allow for users and developers to assign context as needed.
Sovereign WoT is a simple and flexible architecture that allows for context to emerge in a free market of subscribable filters.
Subscribable filters will allow you (for example) to have a custom feed of content containing “ai generated list of all the posts on TOPIC from trusted users in your WoT, weighted by how often you ‘like’ content from each user.”
Filters can be as smart as developers wanna make them, allowing context for trust to emerge from the market of real world use cases.
Yes. We are on the same page here.
Another essential element: we are going to have to track the degree of confidence in influence scores and trust attestations. I might say I think Bob is an expert in some topic, and I’m 99% sure bc I’ve known Bob for years; or I might say I think he’s an expert, but I’m only 5% sure bc it’s based on a single brief interaction. And the Grapevine might say Bob is an expert in something with 99% certainty bc lots of highly trusted individuals independently arrived at the same conclusion; or only 5% certainty bc it’s based on a single attestation by only one user who is multiple degrees of separation away from me.
One of the many approaches to this problem: 
GitHub
nips/77.md at master · lez/nips
Nostr Implementation Possibilities. Contribute to lez/nips development by creating an account on GitHub.