DefaultTrust changes

[KingZee]

@theymos No comment on this suggestion? I have a nice graph too

---

That's why you need to read the ratings and/or use custom trust lists.

If you don't like the trust system, set your own trust list. I still haven't done it, because I prefer to see users as most people see them. But feel free to start promoting custom trust lists, if enough people agree, DT will become less powerful.

It's a good idea but a shame that so many people promote custom trust lists to solve a problem that shouldn't exist in the first place.

If the majority agrees that DT is a select group of people that might not reflect the true list of actual most trusted people in the forum, then doesn't keeping DT as it is make it even more of a problem? Custom lists might look like they're solving the problem, but they're really not because the majority of the forum uses default trust. So in the end you're going to be living in your own echo chamber by removing default trust from your list, because you're completely oblivious to how everyone else actually sees you.

I'd say it depends on the situation. I've received some positive trust for being helpful. I don't think that's "worse" than someone who receives positive trust  after a few small trades with DT-members. In fact, it took me many years to get this (and I appreciate the appreciation), while it's quite easy to gain trust by doing a few trades.

This can easily be fixed if the risked amount is also taken into account (Why is it there in the first place if it's not?). In my opinion, someone with a few trades is more trusted than someone who posts all day on the forums, because at the end of the day, someone who had money risked through their hands means that they're not tempted to scam at least that much. Of course, account rank also comes into play, because a legendary hero wouldn't scam someone off a few dozen bucks simply because their account is worth more through sig campaigns.

Actually, using just 3 variables :
1. Account rank
2. Account current trust
3. Amount risked

I can come up with a system that won't depend on a centralized default trust, and at the same time gets updated in real time depending on member's trust over time :



If any of these 3 members get negative trust, then all their network is going to have less trust points because that person becomes shady.
If User 1 has biased vendetta against User 2, because there's a risked amount variable then baseless claims will have little effect.
This can't be spammed using multiple accounts because newbie ranks and risked amounts are too little to matter.

So in the end the only issue becomes actually verifying that the trust, risked amounts, and if the trades actually happened. Which wouldn't be a hard task because you'd only need to check people with a suspiciously high amount of trust.

This cliché of me pitching up ideas is getting rather old, is btctalk hiring?

[1713919612]

1713919612

[1713919612]

1713919612

[1713919612]

1713919612

[Lauda]

@theymos No comment on this suggestion? I have a nice graph too
-snip-

Using amount risked in any calculation does not work for several reasons:

1) It can't be verified, thus it can be easily abused.
2) Sometimes users want to keep amounts private.
3) Positive trust for reasons other than trading.

[Zin-Zang]

I would be willing to bet a couple BTC that Vod and actmyname are not the same person.

You called Hhampuz an ass-kisser for enjoying the change in Default Trust, which is baffling to me; everyone who approves of the change is an ass-kisser?
It's easy to make a monster out of the person calling you on your nonsense, but in reality (where the rest of us are) it's more likely than not something about your behavior and personality that is accumulating so much distaste. You're experiencing something along the lines of delusional referencing.

Similar to a young man that is unable to get laid, it's not everybody else, so let's stop pretending.

Well you lose that BTC, because actmyname=vod

Dude, you have no idea of my true mindset, so keep your bad dime store shrink analyses to yourself, because frankly you're not very good at it.



 

[Hhampuz]

I would be willing to bet a couple BTC that Vod and actmyname are not the same person.

You called Hhampuz an ass-kisser for enjoying the change in Default Trust, which is baffling to me; everyone who approves of the change is an ass-kisser?
It's easy to make a monster out of the person calling you on your nonsense, but in reality (where the rest of us are) it's more likely than not something about your behavior and personality that is accumulating so much distaste. You're experiencing something along the lines of delusional referencing.

Similar to a young man that is unable to get laid, it's not everybody else, so let's stop pretending.

Well you lose that BTC, because actmyname=vod

Dude, you have no idea of my true mindset, so keep your bad dime store shrink analyses to yourself, because frankly you're not very good at it.



 

Could you link your other accounts, please? I won't do anything, pinkyswear!

[Zin-Zang]

I would be willing to bet a couple BTC that Vod and actmyname are not the same person.

You called Hhampuz an ass-kisser for enjoying the change in Default Trust, which is baffling to me; everyone who approves of the change is an ass-kisser?
It's easy to make a monster out of the person calling you on your nonsense, but in reality (where the rest of us are) it's more likely than not something about your behavior and personality that is accumulating so much distaste. You're experiencing something along the lines of delusional referencing.

Similar to a young man that is unable to get laid, it's not everybody else, so let's stop pretending.

Well you lose that BTC, because actmyname=vod

Dude, you have no idea of my true mindset, so keep your bad dime store shrink analyses to yourself, because frankly you're not very good at it.



 

Could you link your other accounts, please? I won't do anything, pinkyswear!

Sure

ZIN - ZANG

There you go dumbass.  

You think I would have been as peeved, if I was using multiple account like all of you lauda bitches.
Doubtful.
I would have just dropped this one and used the others.
But since I am not cheating and stealing from people like you guys, one account name to speak my mind was all I needed.

 

[Johny Depp]

This appears to be a fundamental change in DT. I'd love to see how it plays out against scammers in the long run.

[suchmoon]

I thought that I had a good idea for limiting each individual truster when handling the last two criteria: set it up as a circulation problem as below, and then find the maximum flow. The "user tX"s through whom flow passes would be the DT1s selected.

https://i.imgur.com/JFDH3Qg.png

(The orders of the users would be randomized on each run.)

There are efficient algorithms for maximizing the flow in problems similar to this, which is why thinking about it in this way occurred to me. However, it turns out that the "exactly 0 or exactly 10" requirement on the rightmost edges makes finding an exact solution too difficult.

I might try to write an algorithm for finding an approximate solution to this, but I probably won't get to it in the near future. Anyone else want to give it a try? You might also be able to structure it as a knapsack problem or something else, but I haven't gotten around to thinking about that yet.

Crap, I wish I had paid more attention in my CS classes. No recollection of max flow whatsoever. Google is extremely unhelpful. So what's the range for X and how does it depend on merit? Can it be negative? How big is N - that's only users who have custom trust lists, right? I feel like I'm missing something obvious here.

[KingZee]

@theymos No comment on this suggestion? I have a nice graph too
-snip-

Using amount risked in any calculation does not work for several reasons:

1) It can't be verified, thus it can be easily abused.
2) Sometimes users want to keep amounts private.

It CAN be verified manually. You speak as if in the current state, you can verify anything at all in trust ratings? It's impossible to fix that problem, people can just neg/positively trust each other and unless there's another person that verifies that trust behind the scenes it may as well go undetected. And I quote from that exact same post :

So in the end the only issue becomes actually verifying that the trust, risked amounts, and if the trades actually happened. Which wouldn't be a hard task because you'd only need to check people with a suspiciously high amount of trust.

3) Positive trust for reasons other than trading.

Is that really a bad thing? Name one reason where you would want to red tag someone without bias, that doesn't involve money.

It can easily be fixed by setting a default non-zero number for trust ratings that don't involve risked amounts, which is completely fine imo because any trust rating that isn't exactly this : "How much money I trust this user with", is honestly just a biased opinion that is the whole cause of why the first trust system is being binned.

EDIT : Regardless of how important I think risked amounts are, if @theymos would rather start with a less complex system first then you might as well just remove risked amounts, my system is still solid. The concept is similar to what he wants to do but won't require an exponential amount of time to calculate compared to the intractability of the current way on how he wants to calculate trust. If the list starts to have numbers in the triple digits, Trust will actually stop being updated instantly just because of the time needed to calculate propagation through the tree.

[suchmoon]

It CAN be verified manually.

How?

You speak as if in the current state, you can verify anything at all in trust ratings?

You can check the reference link and other proof presented in the feedback and everyone can decide on their own what level of verification is needed/acceptable.

[theymos]

@theymos No comment on this suggestion? I have a nice graph too

It's an interesting idea, but I think that trust ratings and trust lists are fundamentally different concepts which shouldn't be mixed. Just because you had a good trade with someone doesn't mean that you trust their judgement generally. For example, your system would tend to strongly amplify long cons like pirateat40, I think.

Also, we're not going to moderate things like "did a trade actually occur, and with x value?".

[eddie13]

I am still unsure of the workings of exclusions, or want to clarify my understanding..

From what I think I understand exclusions do absolutely nothing for anything other than your own personal view unless you are on DT1?
An exclusion does not have effect on anyone's trust view other than your own?

Example: Suchmoon adds me to his list and I add member1. Member1 has added member2 but I think member2 has bad ratings so I exclude member2 from my DT1 view..

Member1 is my DT0 and is Suchmoon's DT1
Member2 is my DT1 and is Suchmoon's DT2  Excluded from my view but not from Suchmoon's view?

I exclude member2 to take his bad ratings off of my DT1 view but my exclusion of member2 does not exclude member2 from Suchmoon's DT2 view?


I think this is how this is working but I think it would work better if my exclusion from my view worked up the chain and also excluded member2 from Suchmoon's view..

If not, my work weeding out the bad ratings that slip through only fix my view, but Suchmoon's view of DT2 is still going to look like crap from my "subordinates" that I have actually excluded but still look included to Suchmoon, and therefore my trust list to him still looks like crap.. Right?

[Lauda]

From what I think I understand exclusions do absolutely nothing for anything other than your own personal view unless you are on DT1?
An exclusion does not have effect on anyone's trust view other than your own?

Partially correct. It also affects the people that have added you to their list.

Example: Suchmoon adds me to his list and I add member1. Member1 has added member2 but I think member2 has bad ratings so I exclude member2 from my DT1 view..

Member1 is my DT0 and is Suchmoon's DT1
Member2 is my DT1 and is Suchmoon's DT2  Excluded from my view but not from Suchmoon's view?

If you exclude him, he'd be excluded from suchmoon's view (assuming nobody else has inclusions).

I exclude member2 to take his bad ratings off of my DT1 view but my exclusion of member2 does not exclude member2 from Suchmoon's DT2 view?

It does.

[theymos]

Crap, I wish I had paid more attention in my CS classes. No recollection of max flow whatsoever. Google is extremely unhelpful. So what's the range for X and how does it depend on merit? Can it be negative? How big is N - that's only users who have custom trust lists, right? I feel like I'm missing something obvious here.

https://en.wikipedia.org/wiki/Max_flow
https://en.wikipedia.org/wiki/Edmonds%E2%80%93Karp_algorithm

Note that you can require a "minimum flow" per edge and still be the same problem, but that refers to requiring that much flow, where 0 is not allowed.

My intuition is that with my "0 or 10" requirement, it becomes the (NP-hard) knapsack problem. There are good approximations for that, though.


I was thinking that X = earned_merit intdiv (10 or 250), but I'm not sure.

N = all users who match the current truster criteria, either 10 or 250 earned merit. The "excluding merit sent by the trustee" thing couldn't be added here AFAICT, and would have to just limit users allowed into this step. M = the number of distinct users trusted by the users on the left side.

[suchmoon]

I am still unsure of the workings of exclusions, or want to clarify my understanding..

From what I think I understand exclusions do absolutely nothing for anything other than your own personal view unless you are on DT1?
An exclusion does not have effect on anyone's trust view other than your own?

Example: Suchmoon adds me to his list and I add member1. Member1 has added member2 but I think member2 has bad ratings so I exclude member2 from my DT1 view..

Member1 is my DT0 and is Suchmoon's DT1
Member2 is my DT1 and is Suchmoon's DT2  Excluded from my view but not from Suchmoon's view?

I exclude member2 to take his bad ratings off of my DT1 view but my exclusion of member2 does not exclude member2 from Suchmoon's DT2 view?


I think this is how this is working but I think it would work better if my exclusion from my view worked up the chain and also excluded member2 from Suchmoon's view..

If not, my work weeding out the bad ratings that slip through only fix my view, but Suchmoon's view of DT2 is still going to look like crap from my "subordinates" that I have actually excluded but still look included to Suchmoon, and therefore my trust list to him still looks like crap.. Right?

My understanding is that it does work up the chain as you're expecting. I'll see member2 striked out unless someone else includes them, and even in this latter case I can still see that YOU excluded member2, which is all that really matters as far as crappiness (or not) of your list.

Besides my depth setting is 1 so it ends at your direct includes and I don't plan to care any deeper.

For DT purposes, member2 would be in DT4, very unlikely to matter to anyone. Even if you get into DT1, member2 would be DT3, still not particularly important.

[KingZee]

It CAN be verified manually.

How?

You speak as if in the current state, you can verify anything at all in trust ratings?

You can check the reference link and other proof presented in the feedback and everyone can decide on their own what level of verification is needed/acceptable.

It's still a manual verification. Just like you can red-tag someone caught posting a fake trust rating, you can red-tag someone who posted a trade with an amount that never happened. I really don't see how such a detail fundamentally changes anything. Every single part of information written in a trust rating cannot be automatically verified.

@theymos No comment on this suggestion? I have a nice graph too

It's an interesting idea, but I think that trust ratings and trust lists are fundamentally different concepts which shouldn't be mixed. Just because you had a good trade with someone doesn't mean that you trust their judgement generally. For example, your system would tend to strongly amplify long cons like pirateat40, I think.

Also, we're not going to moderate things like "did a trade actually occur, and with x value?".

Like I posted in that post, custom lists arent a solution. You can't just set a custom trust list and live in your own world, so that's why my suggestion forces a decentralized common trust list for everyone. Custom trust lists already mean nothing because only the person using them sees them.

Moderating the trust ratings also becomes the burden of the network itself. Newbies will have little to no weight compared to legendary members, and legendary members also have only a share of contribution compared to trusted legendary members. I honestly think it's possible for the network to calibrate itself. If someone is abusing trust, he's going to be red tagged. If someone is legitimate, he's going to be more and more influencive on the rest of the network and so-on. No centralized trust list needed. If someone's position changes, and he becomes trusted/untrusted after being otherwise, all the ratings specific to this account are going to propagate depending on the change.

Still, I can see that it might be a drastic change and not easy to implement compared to the current solution of updating and reworking the default trust list. Would be fair enough to wait for its results before judging.

[eddie13]

It does.

Ok good..  

I was confused because I was still seeing someone that was excluded under me coming in and out of my DT view a couple times.
I guess some person(s) higher than the exclusion must have included them and has since excluded them again, maybe a couple times...

It gave me the impression that my sub's exclusions weren't working for my view and must have missed what changes were going on above me at the time.

[suchmoon]

~

Thanks! I'm slowly starting to grasp this. Not that I'm likely to contribute anything useful, BFS/DFS is about as far as I ever got in search algorithms.

It's still a manual verification. Just like you can red-tag someone caught posting a fake trust rating, you can red-tag someone who posted a trade with an amount that never happened. I really don't see how such a detail fundamentally changes anything. Every single part of information written in a trust rating cannot be automatically verified.

Right, but now any DT rating has the same weight in the score. You're suggesting to amplify weight based on a potentially fake number (risked amount). I think the ability to game your system is higher, even if temporarily until being caught.

[EcuaMobi]

I've just seen this and I'm liking both the idea and the effects to the DT1 list.

Just one thought: since DT1 can change every month so can the DT2 list, so now it makes more sense for several DT1/DT2 members to tag scammers, just in case any tagger stops being DT2 for a few months (I know the idea is for everyone to customize their trust list, but most newbies will still rely on default trust).

And I want to insist on my suggestion regarding trust: guest should see some trust. Default trust would make sense but any trust. Non registered users are still being scammed by known scammers because they don't see the tags.

[Vod]

And I want to insist on my suggestion regarding trust: guest should see some trust. Default trust would make sense but any trust.

That is an excellent idea - I can add it to the BPIP extension.

[LoyceV]

And I want to insist on my suggestion regarding trust: guest should see some trust. Default trust would make sense but any trust. Non registered users are still being scammed by known scammers because they don't see the tags.

I'd also like to see trust on all boards that show signatures. I'm mainly thinking about user Velkro, who advertises his scam site in his signature and mainly posts on boards that don't show trust ratings.
I'll add another tag, but that doesn't mean anything as long as it's not clearly visible wherever his signature is shown.

Between this post and my tag, he got tagged twice already My Reference link points at a scam accusation by minerjones. He hasn't tagged this user yet.

Beginners & Help should show trust ratings, it's terrible idea not to warn Beginners!