FYI, I added the top 15 people in your list to my trust settings, and accounts that were proven scammers were showing up in my trusted feedback list.
That's because not enough people use trust exclusions. Under the new system, people will need to get into the habit of excluding people who leave bad feedback.
If someone has a trust network like this:
ExamplePerson
Idiot
Scammer
Guy
Dude
Then either ExamplePerson or Guy and Dude together can cause Scammer to be excluded. So if a decent number of people are excluding people who give bad ratings (or who trust such people), then bad ratings are likely to be excluded.
A second concern is that I think this system is going to be slow to be able to react to someone who was previously honest and later turns into a scammer.
If someone directly trusts a a scammer, then they are indeed in a bad situation, and they'll need to remove the person manually. I might add a warning to trust pages for people who directly trust a scammer if this ever happens.
It's not a big deal at lower depths due to exclusions.
A last concern is one that was touched on before, but not heavily discussed. This system would not be difficult to manipulate, but it would be much more difficult to detect manipulation. One could quietly buy up a lot of accounts then buy a 2nd set of accounts they want to be trusted. The first set of accounts could all have the 2nd set of accounts added to their trust list which would result in them being often suggested for newer users to add to their trust list. More experienced users may not even notice when this is happening because they are not being asked to add new users to their trust list.
You'd need a lot of accounts for that. 20 full members to make the list, ~100 to get reasonably high in it (currently -- the requirements will probably become higher if this system is adopted). And I'd stop this from happening once I'd notice it, so people buying these accounts would be spending a lot of money on only a very short-term advantage.
has anyone put together a concise pros and cons between current and proposed systems?
New system pros:
- More people in the typical trust network, so more default-visible ratings, more accurate scores, etc.
- Newbies will be more aware of how the trust system works, so they'll be more likely to use it properly.
- Everything will be less complicated for everyone involved. You won't be
able to know what everyone else sees, so all you'll be able to do is maintain your ratings and trust list according to your own feelings. This is how I intended the trust system to work.
- There won't be people who are clearly "at the top" of the trust system. Furthermore, I will no longer need to carefully ensure that the default trust network is OK for everyone.
New system cons:
- More people in the typical trust network, so inaccurate ratings might happen more often, though they should hopefully be balanced by an increase in accurate ratings.
- People who
in practice tend to be at the top of the trust system might feel less accountability/responsibility for maintaining their trust lists than they would if they were listed in DefaultTrust.
- It will be difficult to get a picture of how well someone is trusted for the typical forum user.
- People will need to interact with the trust list system at least a little bit, and not just leave it to DefaultTrust.
I like the suggestion of removing default trust lists completely.
Then newbies will be getting scammed left and right. Newbies need some sort of guidance.
The trust lists of everyone at depth 1 are public, which has historically kept Default Trust mostly comprised of reputable members. This is not the case with the "suggested trust" list under the new system.
All trust lists are public. You can see anyone's trust list by trusting them and then looking at the hierarchical view of the trust network on the Trust Settings page.
My point here is that Default Trust gives a new user a good starting point about who to trust and who not to trust, while this new system asks them to pick their own "Default Trust" pretty much at random, since they will probably have little reason to pick one name over another.
I think that it's OK to trust users at random as long as:
- None of them are creating new accounts to inflate their own trust. The trust score algorithm relies on everyone in a trust network being a separate person. People at depth 0 can trust additional users without consequence if the user trusting them isn't paying attention.
- More of the randomly-trusted users have actively good trust lists than actively bad ones.
- At least a few randomly-trusted users have actively good trust lists.
If these conditions are met, then any incorrect ratings will be excluded by the people with good trust lists.
I think that the suggestion system is likely to result in lists meeting the above criteria. And if a highly-trusted user who was previously suggested starts creating and trusting fake accounts, I'll do something to stop him, or at least warn users.
A quick glance shows that I'm not one of the 50 most trusted members here?
You don't meet the criteria for suggested users. You only have two users on your trust list. You would have 154 points otherwise.
I also see that bobsag3 who openly scammed me and had his original username scammer tagged is on the list, adding positive trust for companies like Black Arrow who have stolen hundreds of thousands if not millions of dollars from this community.
He's still quite widely trusted. Users will be able to avoid selecting him on the suggestion page.
I could manually exclude people like him, but doing that would likely be controversial in itself, and I'd prefer to keep this as automated as possible.
Will the new system have the negative trust indexable via the search engines, for as it stands now nary a negative comment has ever been indexed, only found on this forum via jumping through hoops depending on what settings are ticked?
Trust pages are entirely customized per user, so they can't be viewed by non-users such as search engines.
Show Posts
