Nostr Web Client

This is a proposed roadmap to replace proxy trust data with explicit contextual trust with the goal of better content curation and discovery. The roadmap is divided into baby steps to make the transition as simple and easy as possible for devs and users alike.

Step 1: Calculate WoT scores using proxy trust data (follows, mutes, likes, etc). This is already being done by several clients.

Step 2: Use the WoT scores to filter content. This is also already being done by several clients, such as [Wikifreedia](https://wikifreedia.xyz), where articles can be hidden if their WoT score is below a cutoff.

Step 3: Use the WoT scores to **stratify** content. I am not aware of anyone doing this currently. An obvious place to start: to stratify wiki articles, when you have multiple versions by multiple authors of the same topic.

Step 4: Enable users to issue explicit contextual trust attestations using [Lez](nostr:npub1elta7cneng3w8p9y4dw633qzdjr4kyvaparuyuttyrx6e8xp7xnq32cume)'s proposed [NIP-77](https://github.com/lez/nips/blob/master/77.md). This has been submitted for pull request, the discussion of which is [here](https://github.com/nostr-protocol/nips/pull/1208).

Step 5: Use a combination of proxy trust data AND NIP-77 data to calculate WoT scores.

Step 6: Phase out reliance on proxy trust gradually, as NIP-77 and other contextual trust data becomes available.

Longer term:

This roadmap will be updated as I receive feedback.

Comments?

Lez 1y ago

For step 3-4:

I'm thinking about a site where you can have discussions on each context category. A discussion can be initiated by posting a articles with some personal comments, or just asking a question. So this becomes a reddit clone based on web of trust.

The web of trust calculation can start as simple as taking follows and mutes into count.

Next to the discussion comments there is a BUTTON where you can express your trust in the author's expertise in the category under which it was posted. It creates a new trust event (NIP-77), that is processed the same way as a follow, only difference is that it has much much more score for the exact category it was issued for, somewhat more score for parent categories, and waaaay less score for all other categories.

With this app there is 1 to 1 link between category and trust attestations, so it's probably good way to bootstrap contextual trust.

Problem with this approach is that there is no initial data to start with, so it might be hard to bootstrap the idea.

Reply to this note

Please Login to reply.

Discussion

david 1y ago

I think this is a great idea. And I have a solution to the bootstrap problem.

The solution is: interpretation of proxy data “as if” it were NIP-77 formatted data. For example, if you “like” a wikifreedia article filed under Category X, my software interprets that “as if” you issued a NIP-77 attestation, to that author, in the context of X. But since like != trust (well, maybe it does and maybe it doesn’t, I have no way of knowing), I “interpret” a really low confidence, like maybe 1% or 5%. That way, we start out with a sea of NIP-77-formatted trust data. And you’ll know that if you REALLY trust someone in some context, like if I REALLY trust nostr:npub1a2cww4kn9wqte4ry70vyfwqyqvpswksna27rtxd8vty6c74era8sdcw83a to write wikifreedia articles on economics, and if I want my app to give an added boost to her other content and I’m OK with saying this publicly, then I’m gonna have to issue an actual NIP-77 attestation and set the confidence to something more meaningful than 1% or 5%. In this way, the pool of high quality contextual trust attestations can accumulate gradually.

david 1y ago

If Alice “follows” Bob in nostr, we interpret that “as if” she had issued a trust attestation of Bob, score: 100, confidence: 5%, context: to curate content on nostr.

Our next step is to replace WoT scores with “influence” scores. Two salient features of influence score for this discussion: they are contextual, and they have to take the confidence variable into account.

Lez 1y ago

Yes, I can bootstrap trust data from zero by using follows, but the more difficult part is to bootstrap the content itself, e.g. make people to post lots of interesting articles and discussion starters and come back regularly. Like it happened in reddit. I still like the simplicity of the idea.

david 1y ago

Is there anyone doing anything like what you envision?

Lez 1y ago

Seems like no. There were uncommitted trials doing that.

david 1y ago

What about curating content that already exists? So that you would not need to build a pile of content from scratch, ie. no need to build Reddit from scratch. (Not that it isn’t a cool idea in and of itself.)

There are lots of different types of content on nostr. So one question to consider: which category of content would be the lowest hanging fruit for decentralized curation? Right now I’m working on wiki articles, but there are so many other juicy targets.

Some may be more amenable than others. Stratification of a nostr feed is typically chronological, and it’s not immediately obvious how to combine age + WoT score to reorder the nostr feed. For a wiki topic, I think it’s more obvious: just order them best to worst.

Certain lists like products for sale, movies, music, etc certainly would benefit directly from curation by WoT.