Bitcoin Forum
October 01, 2016, 03:29:44 PM *
News: Latest stable version of Bitcoin Core: 0.13.0 (New!) [Torrent]. Make sure you verify it.
 
   Home   Help Search Donate Login Register  
Pages: [1] 2 »  All
  Print  
Author Topic: Reputation and game theory  (Read 4495 times)
grondilu
Legendary
*
Offline Offline

Activity: 1134


View Profile
September 27, 2010, 03:20:15 PM
 #1

Now, I want to say first that Bitcoin is a great invention.

But obviously it's not the perfect tool for commerce.  I mean, it's an excellent electronic commodidty, perfectly suitable for use as a money.  But still, with a money you have to deal with the problem of the "exchange".

Bitcoin allows you to perform money transfer, in an irrevokable way.   Therefore it is perfect for donations, but unfortunately not for commercial transaction, for a commercial exchange is a bit more than just "two symetric transfers".

Say Alice want to buy Bob some service or good, let's say a nice cake for Xmas.

Alice sends her order to Bob and Bob accepts and tell her to send him the money.

Alice does it, and now she waits for her cake, that she is supposed to get delivered before Xmas.

At Xmas eve, if Alice did receive her cake, she's happy and can tell everyone that Bob is an honnest marchand.

If she didn't, she's unhappy and tell everyone that "Bob is a thief".

She might also say "Bob is a thief" even if she DID receive the cake.  In that case, Bob will say "Alice is a liar".

Now, in the market, reputation of people will be defined by three characteristics :  honnest, thief, and liar.   People will call each other by those names, and somehow one has to find a way to make sure everyone has been correctly qualified for what he is.

Such a situation reminds me of the famous game theory example of "doves" and "falcons".

Somehow I wonder if game theory could be of any help in order to design a rating system that would be efficient.
Bitcoin addresses contain a checksum, so it is very unlikely that mistyping an address will cause you to lose money.
Advertised sites are not endorsed by the Bitcoin Forum. They may be unsafe, untrustworthy, or illegal in your jurisdiction. Advertise here.
1475335784
Hero Member
*
Offline Offline

Posts: 1475335784

View Profile Personal Message (Offline)

Ignore
1475335784
Reply with quote  #2

1475335784
Report to moderator
TTBit
Legendary
*
Offline Offline

Activity: 1136


View Profile
September 27, 2010, 07:13:26 PM
 #2

Don't think its a big problem, people have a pretty good trust meter already.

When I go to buy gas, I give the clerk $50 and say "$50 on pump #6". No receipt. He could easily just pocket it, or only give me $30 worth. I wouldn't know until I'm out of the store and pumping my gas.

good judgment comes from experience, and experience comes from bad judgment
FreeMoney
Legendary
*
Offline Offline

Activity: 1246


Strength in numbers


View Profile WWW
September 27, 2010, 07:59:12 PM
 #3

Don't think its a big problem, people have a pretty good trust meter already.

When I go to buy gas, I give the clerk $50 and say "$50 on pump #6". No receipt. He could easily just pocket it, or only give me $30 worth. I wouldn't know until I'm out of the store and pumping my gas.

That guy is in one known physical location with lots of relationships at risk economic and otherwise. I'm not saying internet trust is imposable, only that it is a different situation.

Play Bitcoin Poker at sealswithclubs.eu. We're active and open to everyone.
Anonymous
Guest

September 28, 2010, 04:25:56 AM
 #4

Ebay has a pretty good system in place for reputation ratings. 
Timo Y
Legendary
*
Offline Offline

Activity: 938


bitcoin - the aerogel of money


View Profile
September 28, 2010, 02:38:53 PM
 #5

2. Prevent people from having multiple accounts. Alice says Bob is a good guy, but Alice is Bob.

This too can be handled by the repuation system. There doesn't need to be a top-down one-account rule. Alice can claim she is Alice all she likes, but if Jose, Sabrina, Anna, Colin, Peter, Jasmin, and Michael all claim that Alice is Bob, then nobody is likely to believe Alice/Bob.

It is easy to open a Facebook account claiming you are somebody that you are not. But it is almost impossible to make that account *well connected* and uphold the lie. The person you are impersonating is more likely to find out the more connections you have, and they will tell those connections.

Poorly connected identities should thus always be viewed with suspicion, and remember, connectedness is not only a function of the number of connections but of the topological distance from the center of mass of the network, and the latter is extremely hard to forge.


GPG ID: FA868D77   bitcoin-otc:forever-d
Babylon
Sr. Member
****
Offline Offline

Activity: 308



View Profile
September 28, 2010, 10:59:41 PM
 #6

I haven't seen it yet but I'd really like to see a reputation site where people's opinions can be ranked.  So, for instance, Alice can say that Bob is a good guy, and that she value's Bob's judgement highly.  Giving him a 10 out of 10 on both.  Bob feels that Carl is a good guy, giving him a 10 as well.  if Bob is the only opinion that Alice has placed any trust in then she'll also give  Carl a 10.  Meanwhile if she has ranked Donna as a 5 on her judgement scale, and Donna gives Carl a 0 (calling him a thief) then Alice's ranking of Carl is going to be 6.6 (since she values Donna's judgement half as much as she does Bob's) This would weed out people ranking themselves, since none of their many accounts would be highly valued by anyone else. 

kiba
Legendary
*
Offline Offline

Activity: 980


View Profile
September 28, 2010, 11:48:51 PM
 #7

Combine with public key encryption to verify that you are who you are across bitcoin sites would be great?

FreeMoney
Legendary
*
Offline Offline

Activity: 1246


Strength in numbers


View Profile WWW
September 29, 2010, 03:42:23 AM
 #8

I haven't seen it yet but I'd really like to see a reputation site where people's opinions can be ranked.  So, for instance, Alice can say that Bob is a good guy, and that she value's Bob's judgement highly.  Giving him a 10 out of 10 on both.  Bob feels that Carl is a good guy, giving him a 10 as well.  if Bob is the only opinion that Alice has placed any trust in then she'll also give  Carl a 10.  Meanwhile if she has ranked Donna as a 5 on her judgement scale, and Donna gives Carl a 0 (calling him a thief) then Alice's ranking of Carl is going to be 6.6 (since she values Donna's judgement half as much as she does Bob's) This would weed out people ranking themselves, since none of their many accounts would be highly valued by anyone else. 

Maybe it could be even more complicated  Tongue

Since Bob having a positive encounter or several with Carl only means that Bob doesn't know he's a thief, not that he knows that he isn't, maybe positive and negative ratings shouldn't be simply averaged. You could rank each of your trusted friends ranking abilities according to how likely they are to give false negatives (think they were stolen from, but weren't) and false positives (actually believing that the laptop got water damage in the mail). This way if a very trusting/optimistic friend made an accusation it could be given a lot of weight.

Play Bitcoin Poker at sealswithclubs.eu. We're active and open to everyone.
db
Sr. Member
****
Offline Offline

Activity: 279



View Profile
October 01, 2010, 09:41:03 AM
 #9

Don't think its a big problem, people have a pretty good trust meter already.

When I go to buy gas, I give the clerk $50 and say "$50 on pump #6". No receipt. He could easily just pocket it, or only give me $30 worth. I wouldn't know until I'm out of the store and pumping my gas.

That guy is in one known physical location with lots of relationships at risk economic and otherwise. I'm not saying internet trust is imposable, only that it is a different situation.

It works the other way too. When I buy gas I fill up first and go inside and pay afterwards. I could just drive off, but I don't. It warms my heart.
Babylon
Sr. Member
****
Offline Offline

Activity: 308



View Profile
October 02, 2010, 04:30:45 AM
 #10

I haven't seen it yet but I'd really like to see a reputation site where people's opinions can be ranked.  So, for instance, Alice can say that Bob is a good guy, and that she value's Bob's judgement highly.  Giving him a 10 out of 10 on both.  Bob feels that Carl is a good guy, giving him a 10 as well.  if Bob is the only opinion that Alice has placed any trust in then she'll also give  Carl a 10.  Meanwhile if she has ranked Donna as a 5 on her judgement scale, and Donna gives Carl a 0 (calling him a thief) then Alice's ranking of Carl is going to be 6.6 (since she values Donna's judgement half as much as she does Bob's) This would weed out people ranking themselves, since none of their many accounts would be highly valued by anyone else. 

Maybe it could be even more complicated  Tongue

Since Bob having a positive encounter or several with Carl only means that Bob doesn't know he's a thief, not that he knows that he isn't, maybe positive and negative ratings shouldn't be simply averaged. You could rank each of your trusted friends ranking abilities according to how likely they are to give false negatives (think they were stolen from, but weren't) and false positives (actually believing that the laptop got water damage in the mail). This way if a very trusting/optimistic friend made an accusation it could be given a lot of weight.

Definitely could be a useful twist.

grondilu
Legendary
*
Offline Offline

Activity: 1134


View Profile
October 05, 2010, 03:46:24 AM
 #11

Ebay has a pretty good system in place for reputation ratings. 

Sure, but it's a centralized system.  So it's easy to make a rating system.

The question is :  how would work a rating system in a decentralized network ?
Is it even possible ?
azazar
Newbie
*
Offline Offline

Activity: 6


View Profile
October 07, 2010, 08:56:57 AM
 #12

Ebay has a pretty good system in place for reputation ratings.  

Sure, but it's a centralized system.  So it's easy to make a rating system.

The question is :  how would work a rating system in a decentralized network ?
Is it even possible ?


This can be implemented as a separate project. It can be a kind of social network, maybe even similar to facebook, but decentralized. Having a trust system is not the only thing it could provide...
grondilu
Legendary
*
Offline Offline

Activity: 1134


View Profile
October 07, 2010, 09:06:49 AM
 #13

Ebay has a pretty good system in place for reputation ratings. 

Sure, but it's a centralized system.  So it's easy to make a rating system.

The question is :  how would work a rating system in a decentralized network ?
Is it even possible ?


This can be implemented as a separate project. It can be a kind of social network, maybe even similar to facebook, but decentralized. Having a trust system is not the only thing it could provide...

Well, this would be Diaspora.

But before it is implemented as a software, I'd like to see some theoretical, almost mathematical/game-theory-oriented approach to this, in order to determin whether or not it is possible, and how efficient it would be.
azazar
Newbie
*
Offline Offline

Activity: 6


View Profile
October 07, 2010, 09:42:14 AM
 #14

This can be implemented as a separate project. It can be a kind of social network, maybe even similar to facebook, but decentralized. Having a trust system is not the only thing it could provide...

Well, this would be Diaspora.

But before it is implemented as a software, I'd like to see some theoretical, almost mathematical/game-theory-oriented approach to this, in order to determin whether or not it is possible, and how efficient it would be.


I think, that every person in that network should have it's own small trust list, which is filled manually. And a large trust list, which is loaded from other people, that are present in that small list. This can be done recursively to make a full list.

For example:
A trusts both B and C
B and C trusts D
So A should see that, there two of his trusted parties, that trust D

I'm not not familiar to game theory, but I think it should work.
mpkomara
Full Member
***
Offline Offline

Activity: 180



View Profile
October 07, 2010, 01:43:35 PM
 #15

Trust is never binary.

A trusts B 99% of the time
A trusts C 99% of the time
B trusts D 99% of the time
C trusts D 99% of the time
Do we know if B trusts C and vice versa?
If not then

A trusts D = min(AB*BD,AC*CD) = 0.99 * 0.99 = 0.9801

azazar
Newbie
*
Offline Offline

Activity: 6


View Profile
October 07, 2010, 02:01:05 PM
 #16

I simplified it a bit. Trust can also be negative for those, who is not trusted. A trusts D = average(AB*BD,AC*CD)
chickenado
Full Member
***
Offline Offline

Activity: 161


View Profile
October 07, 2010, 02:18:58 PM
 #17

There are at least two independent metrics

1) how much A trusts B's honesty
2) how much A trusts B's trust in C's honesty
azazar
Newbie
*
Offline Offline

Activity: 6


View Profile
October 07, 2010, 05:04:09 PM
 #18

There are at least two independent metrics

1) how much A trusts B's honesty
2) how much A trusts B's trust in C's honesty


There are two options:
1. Use same values for trust, and trust in honesty(and absence of stupidity in choice of trusted parties).
2. Use separate values.

I think, that it should be optional to end-user.
Babylon
Sr. Member
****
Offline Offline

Activity: 308



View Profile
October 08, 2010, 06:45:48 PM
 #19

This can be implemented as a separate project. It can be a kind of social network, maybe even similar to facebook, but decentralized. Having a trust system is not the only thing it could provide...

Well, this would be Diaspora.

But before it is implemented as a software, I'd like to see some theoretical, almost mathematical/game-theory-oriented approach to this, in order to determin whether or not it is possible, and how efficient it would be.


I think, that every person in that network should have it's own small trust list, which is filled manually. And a large trust list, which is loaded from other people, that are present in that small list. This can be done recursively to make a full list.

For example:
A trusts both B and C
B and C trusts D
So A should see that, there two of his trusted parties, that trust D

I'm not not familiar to game theory, but I think it should work.

Basically the idea I outlined.  Except that in mine it was not a binary rating system but a scaled one.

kiba
Legendary
*
Offline Offline

Activity: 980


View Profile
October 09, 2010, 02:20:55 AM
 #20

Perhaps now is the time to discuss actual implementation of a reputation system. As it stand, a German scammer probably got away with some bitcoins.

Pages: [1] 2 »  All
  Print  
 
Jump to:  

Sponsored by , a Bitcoin-accepting VPN.
Powered by MySQL Powered by PHP Powered by SMF 1.1.19 | SMF © 2006-2009, Simple Machines Valid XHTML 1.0! Valid CSS!