Bitcoin Forum

Distributed social networks combined with a reputation system would be huge. Like eBay, LinkedIn and Facebook combined on a distributed basis. You could take something like Diaspora (http://www.joindiaspora.com) and build on it a reputation system where you can rate your experiences with every person or brand on earth. The value of the reviews you give would depend on your own rating and your connectedness to the social network, so you couldn't cheat by creating new accounts and writing fake reviews. You could place more value on reviews by your friends or their friends, and see how you're connected to the merchant.

A system like this would be a very good incentive to be honest and not anti-social, both in business and personal life. You could instantly check the reputation of any person (maybe their credit rating hovering over their heads on your augmented vision glasses? ;D), and few people would like to trade or otherwise interact with people who have unsettled anti-social things on their record. Maybe some would require higher prices from those people because of the risk involved.

It would be way better in maintaining peace and justice than the state-monopolized courts we have: practically free, non-violent, instantaneous and effective.

What do you think about this?

Reputation rating is essential if we are to move to a society based on voluntary interactions.It definitely needs to be as distributed as possible to lessen the chance of it being gamed.

I'll describe the art judging system (not quite the same as a reputation system, but related) used in A Tale in the Desert. Despite intense pressure to game it, and a really smart userbase, it has stood up remarkably well, and I think it may be as game-proof as any system could be given the constraints. Maybe some concepts from this can be used in a reputation system.

Background: Players create location-based "real" art within the virtual world. Progress in the game based on how good their art is. "Good" is judged by fellow players (competitors in many cases.)

Ok, first the constraints: Anonymous judges, known artists, continuous judging period, costly (in terms of player time) to judge the art.

The big idea: "A good judge is one who tends to judge the way other good judges do."

Details: A judge's Influence is the produce of his Quality and Worldlyness. As a judge, your Worldlyness is a score 0-1 based on the number of other judges that have judged a piece of art that you have judged. If you have co-judged any piece of art with another judge, then you are considered Peers. As a judge, your quality is a measure of how closely your scores on specific pieces of art correlate with Worldly judges.

We limit the peer list on a given piece of art to a fixed number, and then randomly rotate judges out. This prevents a few frequently judged pieces from dominating the Worldlyness computations.

We give limited feedback to judges in the form of "You tend to score paintings higher than your peers." We don't provide feedback of the form "Your scores on paintings tend to be opposite your peers."

A reputation system to replace courts? I highly dislike this idea because the point of courts is to judge people based on evidence and objectivity and to specificly ignore subjective factors such as reputation.  Someone's reputation can be strongly affected by hearsay, gossip, superstition, moral panics, prejudice, hysterical media, and the like.  That would be a world in which, de facto, a popular kid would have more rights than an unpopular kid. The thought makes me shudder.

I don't think it's anyone's "right" to be an asshole and force others to still buy from him or be his friend. I think everyone should be free to choose who they associate or don't associate with, by whatever criteria they might have.

Reputation systems have always ruled, because they're very useful. Long ago hearsay and gossip were the only channels to spread reputation, but the Internet opens us whole new possibilities that give us much more accurate information about things. People already check the reputation of shops and their products, or the employees they hire, from the Internet. I'm just speaking for a more refined version of the currently used systems. Instead of a million product review sites there would be one decentralized and well-adjusted system. It would also make things better for the reviewee, who could respond to all criticism targeted at him.

Yes, but the point of courts is not to restrict freedom of association, the point is to protect freedoms, such as right to property. Unconditionally. Irrespective of reputation.

If you get rid of courts and replace them with a purely repuation-based system, those freedoms would cease to be universal.

It is one thing not to associate with someone because you think they are an asshole. It is another to beat them up without fear of reprecussions simply because they are disliked by a majority.

 

Ok. I'm not hoping for a world without justice and protection organizations. I'm just envisioning a system to compensate for the shortcomings of the state monopolized courts we have now. They don't really serve the ideals you mentioned. In fact, it's often quite the opposite.


The repercussion for beating up a kid who stole candy would be bad reputation. Nobody wants bad reputation. On the other hand, somebody who "robs" a rapist who refused to compensate to his victim, might even get a reputation increase. That would be even further incentive to not be a criminal, in addition to getting a lower reputation rating.

Hm. Morality by majority vote? I don't think I like where this is going. What if you get a bad reputation just for being, say, an atheist? There's also the problem of extortion: "Pay up, or me and my 100 buddies will give you a really bad reputation." (This actually happened in a MMORPG, but I can't seem to find the news article right now).

Then the problem you have is not the reputation database but the community you live in. In a democracy the majority is supposed to be in control of the state guns anyway, so non-violent reputation databases won't be your first concern. The kind of people you want to deal with won't ostracize you because there's "he's an atheist!" from a random dude on your profile.


Yes, that's an issue that needs to be kept in mind when planning the system. Similarly the problem of "recommend me and I'll recommend you" schemes between strangers needs to be eliminated. Maybe by requiring good connectedness to the social network in some way.

Well, I'm not sure what the perfect system is but majority/democracy is certainly not it..  Look at what's happened in the United States, everything is illegal.  Smoking is being outlawed in phases because the majority doesn't smoke so nobody cares that it infringes on the rights of the smokers.  In California they make a law against anything that bothers someone.. imagine something comes to vote:  "Let's outlaw X because it's annoying" then the idiot majority goes "Hmm yea X is annoying and *I* don't do X anyway so whatever.."

Sooner or later something that you do comes around for vote and you can't do anything about it because you're not the majority.

Basically democratically deciding things is a sure fire way to infringe on absolutely everyone's rights because the majority is not one group but rather a complex overlapping structure.  You might be included in one majority but you're the minority in 100 other ones.

There has been quite a lot of scholarly research on reputation system in the last 10 or 15 years; see http://scholar.google.com/scholar?q=reputation+system (http://scholar.google.com/scholar?q=reputation+system).

This article looks particularly relevant for Bitcoin:

Reputation Systems for Anonymous Networks
Elli Androulaki, Seung Geol Choi, Steven M. Bellovin, and Tal Malkin Department of Computer Science, Columbia University
{elli,sgchoi,smb,tal}@cs.columbia.edu

That looks interesting indeed, gavin. Requires a trusted, central bank though, and I'm not sure yet if one can get rid of it in that scheme. Then again, Bitcoin itself is about getting rid of the central bank. ;D

I've looked at several distributed reputation schemes in the past, and the best one I found was Havelaar (pdf) (http://www.disco.ethz.ch/publications/netecon06.pdf). The big downside is that it assumes a lot of transactions per peer, plus the availability of a network store (i.e. DHT).  :-\ Edit: Misremembered that. The downsides were for a different aspect of the Kangaroo application mentioned in the paper. Havelaar only really requires somewhat persistent peer IDs that can be contacted within one round (week) because reputation is propagated to specific peers to make attacks more difficult. It's also a little slow to propagate the reputation.

 

4chan is the final boss of the internet  :D

Collected papers on reputation systems:

http://databases.si.umich.edu/reputations/bib/bib.html

Personally, EBay has taught me that 100 Gold Stars cannot overcome 1 well-considered Failing Mark.  Good commercial behaviour is a self-correcting system and will tend towards a minimum of fraud as it goes along.

As for this:
We're not building a justice system here.  That would be an outgrowth of the reputation system itself, a consequence of having one as it were.  Fraud, on both sides, would be dealt with via arbitration.  What that system will look like cannot and should not be planned.

I think, it will be enough, if I just have a way to define my trust relationship to my known peers,
a way to recommend something/someone positively or negatively, optionally with details describing my opinion,
and a way to query my peers about recommendations for something/someone, recursively, along with their trust on their source peers.

I'd like to have not only a recommendations exchange system, but also a link sharing, to supplement or maybe even subvert the internet search machines like google.
That kind of search machine will be human powered and will be less SEO-affected, so will contain more useful information and less noise and advertising.

Does something like that already running somewhere? Who can tell?

I think people in here knows that democracy can be pretty dangerous.

Google is quite good. The only thing I am worried about is the data falling in the wrong hand.

Google is good, but his adversary in his fight with SEOs is numerous and well funded.
Well, okay, SEOs of the wolrd are not well coordinated, but everyone is trying to fool Google at his best
and periodically manages to inject his noise into the list of high rated links.

By the way, have you asked yourself,
why can't you change the order of results, that Google provides to you? What if it's rating method
does not correlate with your's? Why you are not allowed to propose your own ordering?

And after all, I don't believe, that a machine will have a chance in separating informational noise from precious bits.
Well, I believe, that only a human may do that the best, but I can only trust results, that came from a trusted
human or trusted community.

I hope, I will find that I search  ::).

Google have more resources than black SEOs will ever have. Google have thousand of engineers, testers, etc, at its disposal, and then they will just use YOU.  Good SEO people will work with Google, not trying to game Google.



Everytime you search, they use that data to make better ordering over time.


Human intelligence has given us Google.


Look, if somebody thought they have something that could kill google, they would have won by now.

Id use FOAF and web of trust ...

^^^

Is foaf added to the metadata of your site?

I like the web of trust idea because it provides a swarm backup for a review/rating site.Say you had a site like yelp where companies game the system by paying for positive reviews.If the reviews on the site were different to the reviews in the cloud it would raise red flags.It would be almost impossible to game both the swarm and the site if there was a big enough swarm.The company would have to hire more reviewers than the swarm contained to affect anything.Much like bitcoins....

If you had access to the data from the web of trust you could display it on your site as well which would make it open for anyone to compare the two immediately.

Nope. Google is ONE, SEOs are huge horde. The more reputation Google collects,
the more effect is produced from fooling it, the more funding SEOs get.
Google have thousand of engineers, perhaps, no more than a several hundreeds work on improving the search engine.
SEOs are millions. Well, professional SEOs are no more than hundreed of thousands, perhaps.

So? You are not allowed to change ordering to, say, date of last modification ordering.
To size of page ordering.
To total number of links in a page ordering.
To number of links, matching regexp ordering.
To number of links, pointing to this page ordering.
Etc, etc, etc...
You cannot enter your formula for rating the pages and sort them accordingly.
Just not allowed and that's it.
And it is no accident. It is policy. You get free search, but we rate and filter results at our will.
If every user had his own rating formula, SEOs will starve and die, since they are not targeting Google,
they are targeting it's formula, which is global among users yet.
Russian SEOs are forced to target both Google's and Yandex's rating algorithms, that is more complicated ;D.
So, internet search is not FREE, it is only unpaid by users.

I don't believe, that NOW any machine or algorithm may become more intelligent, than a group of trused friends
in separating informational junk from meaningful information.
The problem is that the definition of sense is still discussed as a philosophical term only, there is no
successful mathematical models for sense, that work. It is only known, that sense is impossible without context.
Information search is a process of determining whether something makes sense about a subject of the search or not.
Machine can only determine, whether something is maybe about a subject of the search or not.

It does not care, whether that is a domain parking, that contains no sensiful information at all.
Well, It may be engineered to detect patterns of domain parking, that are in common now, but
the next step of SEOs will be to make their noise look ever more closer to useful pages.
But they will not add any sense to their parkings.
This struggle have no end, until machine start to understand the sense, which I believe is not yet possible in near future.

And google is easily told by the govs or by corporate to rate down some links. It is not trusted in any sense. It is single.

Do you work with Google, seriously?
If you trust it's search engine, then I may try to trust you and you may make a proxy from me to google search machine.
We can try that for some time and then, if something bad (like unwanted noisy advertising) happens, either you will distrust google search or I will distrust you for search results.
Sooner or later the network will either break to distrusting parts and die from isolation OR stabilize and become very low noise search engine.
That is the scheme I'm looking for.

The cyberpolice are going to backtrace you an the consequences will never be the same  :P

Lots of ways to add the meta data, in html tags, in meta tags, or in a separate document linked in (like a stylesheet).  I use foaf in my webpages, and also link to my bitcoin address.  e.g. the semantics would be me hasAddress "1DeNqFwDSooBxSDYHSDA3CvnhjA84wfdFz"

You can then say things about yourself, your accounts, your public key and your relationships.

It's quite simplistic really, just a data typed version of the hyperlink, which links documents, instead here you are linking people, accounts and other info.  so a basic web of trust is built in, but we can do smarter things ... with time ...

You sending a Robocop to target me?

lol

irc.freenode.net #foaf


Otherly (http://www.urbandictionary.com/define.php?term=otherly), I contemplated a kind of idea for providing a measure of trust, reliability using p2p.  Here is an example written illustration:

entity A is me
entity B is bitcoinmonopoly.com
entity C is trustworthybitcoinllc.com
entity D is Satoshi
entity E is theymos
entity F is bitcoinscam.com

entity A rates entity B 0.25 (e.g. 25%)
entity A rates entity C 1.0 (100%)
entity A rates entity D 1.0
entity A rates entity E 1.0
entity A rates entity E 1.0
entity A rates entity F 0.0
entity B doesn't need to rate anything because it is not a user
entity C same o/
entity D rates entity B 1.0
entity D rates entity C 0.0
entity D rates entity F 1.0 (for example, they are associated with individuals responsible for scam site)
entity E rates entity A 1.0
entity E rates entity D 0.1

With this data:
Entity A, since it has specifically rated entity F as 0.0, it will evaluate rating as 0.0.
Entity D, since it has specifically rated entity F as 1.0, it will evaluate rating as 1.0.
entity E, since it has not rated entity F, it will calculate such a rating through its peers.
-- Entity E peer of entity A rated entity F as 0.0 and entity E rated entity A as 1.0.  So 0.0 * 1.0 == 0.0
-- Entity E peer of entity D rated entity F as 1.0 and entity E rated entity D as 0.1.  So 1.0 * 0.1 == 0.1
-- Therefore entity E will rate entity F as ((0.0 + 0.1) / 2) == 0.05

Does this make sense?  Is there somethnig that exists that can offer this type of rating mechanism and perhaps for more than one type of rating instead of just one used in this example?

I think, that trust rating is the statistical probability of taking someone into account in the process of making the decision.
And since there are a lot of different decision types, there should be different ratings.

At least, one trust rating value is for trusting <someone> to do <something> (perhaps that is a function of <something>),
and the other rating value is for trusting his own trust on the others.

And you should operate on that values properly, since they are probabilities.
That also means there cannot be negative ratings, noone can downrate you below zero.

Wow, throughput, did you ask google why?

Because it looks like they did it.

http://www.google.com/search?q=google+search+info&hl=en&prmdo=1&tbs=rltm:1&source=lnt&sa=X&ei=8wNxTKaTIYL6swP9pbmPCw&ved=0CAcQpwU#hl=en&prmdo=1&tbs=rltm%3A1&q=bitcoin&aq=f&aqi=g10&aql=&oq=&gs_rfai=&pbx=1&fp=a602d8652b4f7ca (http://www.google.com/search?q=google+search+info&hl=en&prmdo=1&tbs=rltm:1&source=lnt&sa=X&ei=8wNxTKaTIYL6swP9pbmPCw&ved=0CAcQpwU#hl=en&prmdo=1&tbs=rltm%3A1&q=bitcoin&aq=f&aqi=g10&aql=&oq=&gs_rfai=&pbx=1&fp=a602d8652b4f7ca)

This distributed social networking may be build comlpletely decentrailized in a p2p model ,using distributed storage of user data at different peers  such that the  data may be accessed by only those peers who are authorized by the p2p network/creators of the data set .This may be developed as separate application like Ares Galaxy instead of a website which can also have bitcoin as a payment system within the trusted network .