Sovereign Webs of Trust

Sovereign WoT on Nostr begins with a single assertion : _”Users should be able to control which accounts are trusted and how content is fed to them … across any client.”_ The rest is up to us…

This is a living document being refined from feedback by many people interested in better “webs of trust” for nostr.

Implementation standards are being designed by Sovereign WoT WG. This “working group” is open for all. We will host regular chatroom sessions for informal brainstorming on nostr nests and corny chat. The results of which will be reflected in this document as it matures.

Join the discussion on Nostr #wot #wotnip #sovwot —

WoT#

Webs of Trust is how content “moderation” works on Nostr. Trusted connections between real people is what keeps this open network (owned by nobody) from being overrun by bots and bad actors.

Because Nostr doesn’t have centralized moderation or content control, the surfacing of quality content is becoming a growing concern for users and developers alike. Decentralized “webs of trust” (friends trusting friends for content and account recommendations) is our best protection, but the tools that form these webs (content filters and trust rankings) need to be in the hands of end users. The ad-hoc manner in which webs of trust is implemented on Nostr today (based on follows and different for each client) is both inaccessible and unscalable.

Sovereign WoT for Nostr#

Users should be able to control which accounts are trusted and how content is fed to them … across any client.

Sovereign WoT is a Nostr Implementation Possibility (NIP) being designed to put the end user in charge of their own feeds and recommendations. It proposes two standards. Private Trust Lists allows for end users to mark individual accounts as “trusted” and Subscribable Filters allows for developers to publish filters that can suggest “trustworthy” content and other stuff for their subscribed users.

Private Trust Lists#

Because “followed ≠ trusted” (on social media), users should have the ability to explicitly and privately mark any account as “trusted”. This will provide a canonical reference for algos suggesting “trustworthy” content and other stuff.

15 years of “follow” implemented by walled garden social media, where trust is taken for granted, has guaranteed that “follow ≠ trust”. Any meaningful WoT implementation must have a “more explicit” marker for trust.
“Muted” or “blocked” is a blacklist approach. “Trusted” is a whitelist approach. Both are needed. Different people will use differently.
“Is trusted” should have an underlying scalar value, but read as boolean to satisfy most use cases. (It is either 1 or less than 1) This allows filters to be developed that “nudge” trust scores toward or away from “is trusted”.
“Is trusted” is NOT the WoT score. A users subscribed WoT filters will provide a “trustworthy” ranking (based on “is trusted” across their network) for all accounts, content, and other stuff that they interact with.

Subscribable Filters#

Because there is no “best” algo, Nostr should have a “free market” of filters to which end users can subscribe and share. These WoT filters may be published by clients, or relays, or any vendor, and may use any Nostr event data to surface “trustworthy” content and other stuff.

Some WoT filter possibilities#

Maybe WoT filters for content feeds?

all my folllows content?
all my follows trusted by my WoT?
only from my trusted follows?
only content reacted to by a trusted follow?

Maybe WoT filters for recommending new follows?

accounts who post reactions or comments on a “marked” post?
accounts followed by my trusted accounts?
accounts trusted by my trusted accounts?
accounts with content “reposted” by a trusted account?
WARN of imposter (having similar profile data as trusted or followed)?
better WARN of imposter (also checks for similar pfp)?

Maybe WoT filters for suggesting wot filters?

used by my trusted friends?
used by anybody in my WoT?

Maybe WoT filters to “automagically” managed trusted follows?

all my follows are trusted?
all my follows whom I’ve “interacted with” in 30 days are trusted?
TrustNet style “let me assign labels and numbers to qualify trust level.”
NIVS style “let me verify profile metadata on a case by case basis for each user”