[bruhmanfiflo]

a few idea's for safeguards are first tying someone's account to their actual identity meaning one account per person. This could prevent anonymous ratings, harassment, and swarming behavior

Tying ratings to actual meetings or verified transactions preventing leaving multiple reviews from one verified account.

negative feedback can't be hidden or deleted, only improved in future transactions unless deemed harmful by moderators or community.

Anti swarming, "hey friends leave a bad review for this business I hate" wouldn't work as they would need a validated transaction, but this could also have an additional layer of " these socially, professionally, etc connected accounts have all rated this business badly in (some amount of timing factor) this might need to be flagged.

Context protocol. If someone has only ever left negative reviews their "credibility score" becomes lower which in turn lowers their trust score.

Business suddenly gets a flood of positive reviews, this could be flagged as bad behavior by the system for review.

Think less "Yelp-style drive-by ratings" and more "I trust/distrust this business/person and here's why"

[verdverm]

Are you familiar with "brushing"?

https://www.newsweek.com/seeds-mystery-packages-mail-china-b...

point is, these problems already exist and have lots of people working on them, yet they persist as problems. Many of the things you suggest are already part of this cat-n-mouse game

> negative feedback can't be hidden or deleted ... unless deemed harmful by moderators or community

So in fact anything can be modified in the end if the powers that be decide so?

---

edit: I should add a suggestion...

I'm interested in how we might build a more transparent review system on ATProto, where we can all have access to the raw review data, and we can leverage the plug-n-play algos, stacked moderation, and other features of the network. We could create our own lexicons for reviews and make them generic enough to apply to all other lexicon and records in the system. We could end up with a two level system (1) the review records themselves (2) the labellers and algos that build on, annotate, or attest to them.

A discussion for an "attestation" lexicon has already been started: https://github.com/orgs/lexicon-community/discussions/8

What this offers is a system where there is not a privileged group or a single community moderating, but rather an ecosystem where these entities compete and users can choose which one(s) they subscribe to (or trust).

[bruhmanfiflo]

No I wasn't familiar with brushing, I'll have to look into it. When you say powers that be, that's basically the community, I mentioned mods but I don't think it should be "policed" or moderated by some upper level overlords or even me if I were to build it, the power of the system lies in community connections and if the community is intent on seeing themselves better as a whole then I would hope they themselves would make good decisions here. Your suggestion points in the direction of my thinking.