Topic: DefaultTrust changes - page 51. (Read 85467 times)

It's not flawless though, there are for instance still 4 banned users on DT2:

    5. Legendary DiamondCardz (Trust: +9 / =0 / -0) (81 Merit earned (history)) (BPIP)

I would have to go look through their feedback and what not, but it might still be valuable enough to deserve to be included within someone's trust list.

It's not flawless though, there are for instance still 4 banned users on DT2:

This seems to me the system is working as intended and spreading out the Default trust as opposed to having it be a core group.

I didn't say that those who have negative trust should not be on DT. Although this may be the case some of the time.

It seems there might be a mismatch between who is on DT and who should be on DT.

Just because someone has received negative "feedback" from someone deemed trustworthy by some users doesn't mean they can't also be trustworthy to other users. This seems to me the system is working as intended and spreading out the Default trust as opposed to having it be a core group.

It should really only be used by newbies anyways as more experienced users should create their own custom trust lists.

There are a lot of people who are on DT who have negative trust. It seems there might be a mismatch between who is on DT and who should be on DT.

~~~~~~

Guys I am again on this.
I thought about this problem quite extensively during this weekend on the italian riviera.

I think Quickseller formula is not correct as it does allow repetition, while we must find a solution without repetition (if a user is selected, it cannot be taken out again on the same round).
I think Theymos was then on the right path using binomial coefficients, so I am going to use the same technique.

The right probability of any of the 4 excluded in the first round to be excluded in the second round is equal to 1 - the probability of everyone of such 4 to be selected:

1-C(100,4)/C(106,4)=0.210654248

another less intuitive method give the  same exact result:

1-C(102,6)/C(106,6)=0.210654248

(probability of being amongst the 6 excluded from the second extraction chosen by the 100 selected from the first extraction + the 2 new addition).

I am almost sure about this, but please double check me again.
Theymos said he simulated and got a result very similar to Quickseller, this scares me, also because I saw LoyceV and other heavyweight meriting previous solution... shall I go back to school?

EDIT: Forgot to mention, but clearly an hypothesis here: every candidate on the first round is a candidate also for the second round. This simplifies calculations, when we agree on the solution, we'll be able to remove this hypothesis.

For years I've been unhappy with how DefaultTrust ended up as a centralized and largely-untouchable authority, but I was reluctant to change it because the alternatives seemed too messy. However, I've finally decided to try some changes, and we'll see how it works.

#1
As a special exception to the normal algorithm for determining a user's trust network, if you are on the default trust list ("DT1") but more other DT1 members distrust you than explicitly trust you, then it is as if you are distrusted by the default trust list for all purposes except for this very DT1-composition determination.

So if someone on DT1 is doing something stupid, you can ask other DT1 members to distrust them.

See here for live info on this "DT voting".

#2
You can view any page as if you were using the default trust settings by putting ;dt at the end of the URL. Eg. https://bitcointalk.org/index.php?action=trust;u=35;dt

#3
I will periodically (maybe every month) be reconstructing the default trust list to include everyone who matches these criteria:
 - If rank was determined solely using earned merit, then you must be of at least Member rank.
 - You must have been online sometime within the last 3 days.
 - Your trust list must include at least 10 users, not including ~distrust entries.
 - You must not be banned or manually blacklisted from selection.
 - You must have posted sometime within the last 30 days.
 - You must have at least 10 people directly trusting you each with an earned merit of at least 10, not including merit you yourself sent. These "votes" are limited.
 - You must have at least 2 people directly trusting you with an earned merit of at least 250, not including merit you yourself sent. These "votes" are limited.

Unlike the previous policy, I will not generally be trying to cultivate a good list; that will be left to the DT1 members themselves. However, I reserve the right to remove you and blacklist you from future selection if you engage in egregious and obvious abuse, or if multiple known alt accounts could be selected.

Currently not that many users are eligible. If hundreds of users would be selected in the future, I plan to instead choose a random subset of about 100 eligible users each time. This DT1 reconstruction may even automatically happen on a schedule in the future, but it doesn't currently.

---

A major goal of this is to allow retaliatory distrusts and ratings to actually have some chance of mattering so that contentious ratings have an actual cost. If someone is obviously scamming, then any retaliatory rating should not last long due to the DT1 "voting", but if you negative-rate someone for generally disliking them, then their retaliation against you may stick. In borderline cases, it should result in something of a political battle.

This is inspired partly by something that David Friedman said once (though I can't find the quote), that one of the requirements for a peaceful society is the credible threat of retaliation in case you are harmed. As DT was organized previously, one or both sides of a dispute was usually unable to effectively retaliate to a rating, at least via the trust system itself. Now your ability to effectively retaliate will tend to increase as you become more established in the community, which should discourage abuse generally. (Or that's the idea, at least.)

All that being said, I still discourage retaliatory ratings, and with these changes I encourage people to try to "bury the hatchet" and de-escalate rather than trying to use any increased retaliatory power you now have. Also, it's best to make your own custom list, and you must do this if you want to be on DT1.

I am never completely tied to anything, but let's try this for at least a few months and see how it works.

@malevolent @Vispilio @gospodin @fillippone @TheNewAnon135246 @TalkStar @1miau :
See Beginners guide to correct use of the Trust system

It's a first version, there are probably more things I should add, so please post if I forgot stuff. But I also try to keep it "short", and it's already much longer than most users will ever read.

How long did it take you to write it? 

@malevolent @Vispilio @gospodin @fillippone @TheNewAnon135246 @TalkStar @1miau :
See Beginners guide to correct use of the Trust system

Good wall of text.

commenting there my doubt.

@malevolent @Vispilio @gospodin @fillippone @TheNewAnon135246 @TalkStar @1miau :
See Beginners guide to correct use of the Trust system

It's a first version, there are probably more things I should add, so please post if I forgot stuff. But I also try to keep it "short", and it's already much longer than most users will ever read.

I'm convinced, I've started working on my Beginners guide to correct use of the Trust system.

I would recommend also an advanced version.
Apparently it's a mystery to many users who cannot be classified as beginners.

Edit: I reserve the Italian translation.

I can help translate it and post it in my Local section.