Brainstorming Trust System Ideas
I'm sure this thread exists somewhere already but in my searching I couldn't find it. My hope is that we can brainstorm some good ideas in addition to what deimos has already proposed.
The recent discussion about lobste.rs reminded me of the fact that they use an invite system to limit the effectiveness of spammers and manage an invite tree to see the relationship of user accounts. A modified invite tree system would go a long way to help the trust system be more effective while not being stifling.
While it (unmodified strict invite system) would definitely make it more difficult for people to join the site the benefits of such a system cannot be overlook. Even a naive system where we would limit new accounts to require current users to "vouch" for new ones would have noticable benefits later on down the line. The main benefit of this is that users who are a source of consistently inviting troublesome accounts could be held responsible. The added benefit is that you would be a little more conscientious of the people you invite as it would affect your reputation as well.
The naive approach to this system would be a strict invite tree-based site. The only way you could get an new account on tildes is to get an invite from a current user of tildes. Like lobste.rs this woud be a very good deterrent but also (possibly unnecessarily) stifle growth. Such a system is a good starting point but I believe since a lot of effort will be put into making the trust system effective, we can do better if we put in some creative brainpower and work.
Possible ideas/tweaks for an invite tree-based system :
Parent account = person who invited user
Child account = person who was invited
- Users could create new accounts without an invite but would be severly limited in the rate they could post content (1 post a day), make comments (5 comments a day), etc. At any point they could be "vouched" for by a parent account with reputation (become child account) and those limitations would be lifted as though they had an invite from the beginning.
- User reputation would be impacted by child accounts (invite more, more likely to be punished as a result of children's bad behavior) to discourage individuals from becoming invite mills without repercussions.
- If a parent account is banned all child accounts would be demoted to "uninvited" status, child accounts would be encouraged to build up reputation to lose child status and become an independent (reputation-wise) parent account. Accounts would still be historically linked to parent accounts to pinpoint bad behavior down the line if it occurs.
Obviously a lot of effort has already been made by people a lot smarter than me to make such systems work and I think we can draw on that for tildes as well. I tried to look for some peer-reviewed papers and graduate level work about creating such a system but didn't come up with anything useful, perhaps you'll have better luck.
I like the idea of anyone being able to make an account, but only being able to get full privileges from another user who has them. Sounds like this could be our version of Reddit Gold! If you see someone consistently contributing to discussion, you could give them a Wave (pending title) and they'd get a "full" account. Instead of having parents and children, this could potentially turn into a web of users!
I'm probably gonna read this in the morning and see all the flaws in it then haha, but in my current state this sounds amazing.
I think this has merit, I was starting to move towards this idea in my previous comment. Perhaps a web (or graph) where ones ability to invite is only unlocked after a certain no. of edges connect to you (sorry for jumping straight into graph jargon).
My immediate thought is of GPG's web-of-trust that has a similar model. Unfortunately the requirements they recommend for 'vouching' for someone are quite stringent and would be entirely impractical for tildes. I am currently on mobile so can't fully flesh out what I am thinking, but some form of 'kudos' system where one must initially have enough edges before one may start forming edges to other people, and then another barrier before one can start vouching for new , untouched for users. My rough model is still susceptible to botnets exploiting this by having a group of trusted accounts and creating numerous other accounts from this core. I am thinking along the lines of restricting forging new relationships with second-degree edges (or. Two hops away).
Disclaimer: this has definitely not been well thought through :)
I enjoyed reading through this and pretending to understand what it says.
Jokes aside, I get what you're going for, this could definitely be fleshed out, but in my eyes user understandability should be a focus for this sort of system, and this isn't exactly easy to visualize. Would love to see a visual representation if you have the time later!
Hi, so still on mobile, but I did try and draw what I was thinking. It's a little simplified and my handwriting is pretty poor. Feel free to question and comment!
Title here... https://imgur.com/a/nVRJwZH
That helps a lot. To oversimplify even more, the more vouches you get, the more "complete" your account becomes, and your ability to vouch for others increases as well. This could be a great solution.
Exactly!
Thanks for the write-up, I think you bring an interesting refinement of a trust-system to the table.
I would like to state that having some sort of unvouched status is a good idea, even being 'adopted' is a nice way to integrate people into the trust-system.
I realise that this hasn't been backed up by research and you aren't claiming these as the one and only solution, so I hope I don't come off as aggressive or dismissive.
I worry that if /when a user is banned who may have a respectable no. of children , do the children of their children also get orphaned? Or is it only direct descendants?
If it's the former, that is going to be a huge mess of people trying to find another parent.
If the latter, then the new parent takes the risk of not only the new child, but any children the new child may have.
I haven't thought through all the edge cases so you bring up a very valid concern. I tried to addres this by making it so child accounts could become reputationally independent from parent accounts once they gained enough trust on their own. This would be the work around to large numbers of users getting orphaned and lose certain benefits of being "trusted"
Such a system would ideally be tested on live data but without actual reprocussions to gauge how effective it would be at rooting out bad behavior. Like you said, it's based on theory and theory doesn't always translate to practice.
Children accounts could get flagged if they drop below the threshold, for manual review