Building a decentralized reputation system with prediction markets

Previous topic - Next topic

Tal

I've recently been interested in creating a decentralized reputation system, to make it easier for people to transact pseudonymously over the Internet.

One path I was considering is: could some system be created where I bet on whether your future transaction partners will be satisfied? Obviously if there was a working decentralized prediction market, I could create some event relating to whether my transaction with you would be successful and then encourage you to get people who trust you to bet yes, but this is a pretty heavy weight solution. So I'm imagining something inspired by prediction markets but where the predictions are more implicit/easier to make. Has anyone thought much about this? Links to any previous writing on the topic would be appreciated.

psztorc

Quote from: Tal on March 19, 2015, 08:22:20 AM
One path I was considering is: could some system be created where I bet on whether your future transaction partners will be satisfied? Obviously if there was a working decentralized prediction market, I could create some event relating to whether my transaction with you would be successful and then encourage you to get people who trust you to bet yes, but this is a pretty heavy weight solution. So I'm imagining something inspired by prediction markets but where the predictions are more implicit/easier to make. Has anyone thought much about this? Links to any previous writing on the topic would be appreciated.
I think that it would be possible, but largely impractical.

I'm imagining the "insurance" case of prediction markets -- but for this one must always have an amount corresponding to the worst case scenario deposited into the prediction market for the duration of the "insurance coverage" (which largely defeats the purpose of insurance).

Quote from: Tal on March 19, 2015, 08:22:20 AM
I've recently been interested in creating a decentralized reputation system, to make it easier for people to transact pseudonymously over the Internet.
Such a system would certainly be invaluable. I wonder if you'll find these rambling thoughts helpful:

I theorized the concept of purse.io / brawker back in 2012, but in a decentralized way. I had imagined that (a protocol-constrained number of) third party DACs would compete to resolve disputes via SVD. These DACs collected a % of transaction revenue, and - and this is my point- the protocol kept track of a running sum of the total transaction revenue paid. *This* number "was" the reputation (for the buyers/sellers)...it was costly to attain, you could multiply it by star ratings (5 stars, 4 stars, etc. to get 3015 BTC-stars, or whatever) or other vectors to compute all kinds of metrics. If you wanted to "import" reputation, you could trade with yourself...it would still cost a %to the DAC, so this "hack" was actually a feature.

( I didn't fully pursue the idea because I thought a firm ( ie legal purse.io organization ) could out-compete a blockchain on the inevitable credit card fraud ("asymmetric reversibility") problem. ...one more thing for my private "I was right all along" list. )

I think my point is that most people don't seem to really nail down exactly what they mean when they say "reputation". In the real world, it has a clear meaning....but that meaning is completely personal and subjective. I can "trust Person X", and I can tell you that "Person X is trustworthy", but I've never seen anything in WoT or cryptography or computer science that actually represents "the Baysian computation I do when Person Z advises me to trust Person Y" with tractable numbers. I have friends that I would trust with my life, but not with correctly filing my tax return. I also know that some people care about me so much, that they would make choices on my behalf that are actually *bad* for me! People often side with their strongest former enemies, because a new stronger enemy has shown up...so what does it mean to have a "bad reputation"?

This is why, in Truthcoin, reputation has this extremely specific "shareholder-equity modified-by-report-consistency" definition, and there are huge limits on what you can even do with "reputation" once you get it.
Nullius In Verba

lola

I actually think that could work.

You could use that for freelance work, loans or hostel booking.

And psztorc, here reputation is not link to a person but to an action done by a person. This is more like evaluating a risk that's why I think prediction market can work to determine this risk. With this information added you can take your decision of hiring this person for a particular job, lending money to this person or booking a room in this hostel. The difficulty would be in writting the contract that determine if it is a success or a fail for some cases.




zack

Using the blockchain to hire software developers will be easy, just use a truthcoin dominant assurance contract. No reputation or trust required.

Using the blockchain to loan money seems impossible to me.