Replying to Avatar Niel Liesmons

Indeed! Don't expect anyone to do more than that online.

Giving other users a trust score on every category of trust isn't doable nor useful in the first place.

1. Humans suck at giving numerological ratings (because we don't know what it means, at aaaall). Best we can do is ordinally rank people which is an even more obvious UX nightmare.

2. Categories are not clear-cut, and therefor cannot be clear.

3. The cost is on the ones providing the information, instead of the ones who want to use it. You cannot gamify your way out of that.

4. How would you even interpret those ratings without algorithms, that could have solved the whole problem without even bothering the user?

And those are just the issues when these WoT ratings would be static, which they can't be. Maintaining them dynamically over time, lol, good luck.

In this case Big Tech might actually be inspiring. They don't bother you with asking Bob how much you he would score Alice on Trust in a relationship setting, they know he's in love with her before he does.

How? Just look at users actions.

Literally every event = an expression of Trust.

Let algorithms compete on recognizing the patterns in those actions and in being useful for specific applications. Some will be crude and basically free when that's good enough, others will be very advanced and pricey.

The data is public. This will happen anyway. Users can willingly share more data when and to whom they want.

How are you planning to compete with opt-in free market solutions that are built on the actions users are already doing?

yes i've been saying this...

you can count replies as a value in the formula, for example...

you can rank the amount of content in the replies as a value in the formula...

you can categorise negative and positive terms in association with nouns and pronouns to establish some sort of a metric, this one is a bit more fuzzy and would need an AI to recognise the patterns better

ah yes and you can also catch themes as well, from infrequent keywords that will give some measure of the ways in which you trust, ie, this person is a good philosopher, this person is a great coach and pep talker, this person is comic relief, etc etc etc

i think there's a lot you can do before you have to bring ML/AI into the scheme as well, and that is good because it means you can keep that processing on the client and not move it out to the network, and it means that privileged data can contribute (eg DMs)

Reply to this note

Please Login to reply.

Discussion

Yup! Most of the common WoT-needs of today don't even need ML.

What I'm curious about is how to let clients give users access those various simple algo's and then run them locally in the client, because DVM's cannot do that as far as I understand.

well, that gets my hamster wheels turning also... it could be a relay specific feature that generates this data as an indexing scheme and when search results come up it filters them through it and if a high reject confidence comes up it passes over sending it to you at all

i recently fully grokked how the filter search works in nostr and i've been thinking about full text search and this would be an interesting additional feature... it would just need an extra field in event envelopes to carry this information to the client so the user would be able to set the client to drop messages if that value is in the range the user doesn't want

oh yeah, and the whole thing of p2p relays... then the relay is in your hands and you don't have to think about whether you trust it when you set its parameters... i think there will turn out to be a lot of interesting things in this coming down the pipe