Web of Trust + Account Feedback

[cass]

We are preparing to add some new features to help reduce fraud / errors from accounts that look like bter. 

1) You can now "burn" shares for or against an account while posting an optionally signed message on the "wall" of the account. 
2) You can now link two accounts with arbitrary meta data, ie:  the account "bytemaster" can link to "bter-com" and define various properties such as trust-level.

So now users have a way of flagging accounts for both good/bad behavior, leaving anonymous or signed messages on accounts, and build an arbitrary web-of-trust among the accounts.

These features also provide the basis of future features including:
1) Allowing Overstock to issue shares
2) Paying dividends on User Issued Assets
3) Identity authorities for voting

The purpose of this thread is to discuss these features and their implications.

This could be a critical component of the delegate selection mechanism, both for individual delegates and slates.

Also escrow and commerce per Xeroc.

Is there any way to integrate this into the avatars? To be simplistic, BM gets five gold stars, throwaway123 gets none...

Using a combination of standardized colours/shapes to signify various level of trust/reputation.

Perhaps using the Bitshares logo; accounts with no rep have an empty logo, logo slowly fills in section by section as the account builds more trust/rep.

BM gets a full logo, I get one wedge, etc.

Getting one's 'diamond' (very top of the logo) could be a significant event.

Yes, holy shit. The I3 team continues to knock it out of the park. First yield, now rep. 

Off to buy more BTSX...

many good ideas as i find! Will think more deeper on this for the GUI 2.0 --- thx for your input ...

[bitbro]

Reputations should be more based on approval voting with heavy costs for negative voting.  Reputations could also expire and could cost more to expire over longer amounts of time, and could last longer with more  congruent votes.  Needs a google esque algorithm

Mutual connections should strengthen the reputation portrait - less connections should mark the reputation as more uncertain.. Should be very difficult to obtain reputation without mutual connections except if there are lots of votes

Negative votes should have a negative impact on ones own reputation, overall reputation, length of time holding funds should weight ones votes slightly more.

Peers should validate negative votes, should require proof, then should either confirm, abstain or deny.  This should impact the strength of the negative vote.  The more peer votes, the more potential weight the negative vote could have


Sent from my iPhone using Tapatalk

[oldman]

Optional is a very good idea. Different accounts for different purposes.

I'm not sure Neg > Pos is a good idea; possibly create an inherent bias.

Could prevent bad actors from getting their due, particularly on low-level scams where folks might not want to pay a lot to provide negative feedback on a small loss.

Open Bazaar has quite a bit of technical documentation on the burn-for-rep mechanism. Seems to be well-grounded.

[Akado]

For a DAC like a forum, market place (eBay or Craigslist style), or other situations where you have direct p2p interaction with an individual, and you need some level of trust, this makes sense.

This is not a one-size-fits-all implementation.

Do you think if a Market DAC was to be done, it could compete with OpenBazaar? I mean, I dont know the project that good and it's still very new and I dont even know if its running (i think now?) but since it's the first and people's expectations are great, I really dunno what we could do to diferentiate us from them

[gnarl]

I am really not a fan.  I don't want my transaction volume, or trading habits visible to anyone that has my account name.  I also am concerned with the governing body that would be required to officiate such a system.  Even if it is just based on forum posts or something like that.

Requiring users to burn a certain amount of BTSX with the strength of their vote being dependent upon how much BTSX they burn is a step in the right direction.  I would still rather see the bter account problem solved by those that are directly involved though.  Why does Bter.com not have a link you can click to fill out your wallet?  Why aren't users copy pasting the address?

I agree, exchange accounts should not be turned into social media accounts with public stats...

I thought Keyhotee was supposed to be the Identity\Reputation management platform.
Aren't there already plans to integrate Keyhotee that would cover these types of features?

[Riverhead]

The idea of it being optional, or enabled by DAC, makes more sense. Like Puppies said earlier I rely on the blockchain to shield me from bad actors on an exchange. I put up asks, bids, shorts, etc. and they either execute or they don't. They may be partially or fully participated in by many people. I don't care who they are and they probably do not care who I am except in the case of trading certain assets (Overstock.com for example).

For assets that have KYC requirements  you'd opt in to that for the account that wants to trade those assets. Like other exchanges (CampBX for example) you can buy/sell BTC on there all day long but as soon as you want to transfer out to fiat Know Your Customer comes into play.

For a DAC like a forum, market place (eBay or Craigslist style), or other situations where you have direct p2p interaction with an individual, and you need some level of trust, this makes sense.

This is not a one-size-fits-all implementation.

[luckybit]

I think we share much of the same concerns puppies. I can appreciate that you don't want to expose your individual trade history. I wouldn't want that directly exposed for my transactions either. But I also don't want a rep system that can be easily manipulated or bribed like kokojie and skyscraperfarms described.

I hugely appreciate you making this point. However, to a large degree I believe trust in the virtuous characteristics of an account cannot be isolated apart from transaction history. Until you made this post I was focused on the basis for a reputation (transaction history) and lost sight of the impact  such tracking can have on privacy and anonymity.

The two values, 1) reputation and 2) privacy are at odds with one another. To accurately "calculate" an objective reputation value requires tapping into the details of transaction history, but that would have a severe impact on privacy.

I'm going to review this thread with this in mind and I may reverse my position unless I can think of a way to obfuscate the transaction data without disposing of it's value. It may be the best we can do is an aggregated rep value that canNOT be tied to the specific  elements that would compromise privacy.

These are good points. If you're anonymous then no one will trust you. Pseudo-anonymous and you can have reputation.

So you could have privacy by opting-out of the reputation system but if you opt-in then you'll become pseudo-anonymous. The reputation system can be used for these forums for example where reputation might get you some added benefits. People with certain badges could access certain threads which people without those badges cannot access so you can use badges as access control.

[Thom]

I think we share much of the same concerns puppies. I can appreciate that you don't want to expose your individual trade history. I wouldn't want that directly exposed for my transactions either. But I also don't want a rep system that can be easily manipulated or bribed like kokojie and skyscraperfarms described.

I hugely appreciate you making this point. However, to a large degree I believe trust in the virtuous characteristics of an account cannot be isolated apart from transaction history. Until you made this post I was focused on the basis for a reputation (transaction history) and lost sight of the impact  such tracking can have on privacy and anonymity.

The two values, 1) reputation and 2) privacy are at odds with one another. To accurately "calculate" an objective reputation value requires tapping into the details of transaction history, but that would have a severe impact on privacy.

I'm going to review this thread with this in mind and I may reverse my position unless I can think of a way to obfuscate the transaction data without disposing of it's value. It may be the best we can do is an aggregated rep value that canNOT be tied to the specific  elements that would compromise privacy.

[puppies]

I don't understand all of your post, but what I got in essence is you're in favor of buying reputations.

I can't say I like that idea. If that happens I won't rely on it whatsoever. If it turns out to be a small contribution to an account's reputation I might  stomach that, as long as the other contributing factors are reasonable IMO.

Contributing factors as defined by who?  I take issue relying on anything that is not derived from the blockchain.  The whole point of this technology in my opinion is that is allows me to do business with people I don't trust, because I know I can always trust the blockchain.  While I don't believe that on blockchain reputation systems are needed, and that there are better solutions to the problem of people sending to the wrong account, at least on blockchain burning of BTSX is something I can verify. 

I know that you were mainly talking about on blockchain things like transactions and trading volumes, but I don't want to share that information.  I don't think I should need to in order to engage in commerce. 

[Ggozzo]

Can you explain a bit how do you plan to avoid a user registering for example, 100 accounts, and then use them leave good feedback for his main account?

This was my first thought too.  Someone could easily piss away 1000 BTSX in hopes of scamming or misleading someone in order to gain 10,000 BTSX or more.

[Thom]

I don't understand all of your post, but what I got in essence is you're in favor of buying reputations.

I can't say I like that idea. If that happens I won't rely on it whatsoever. If it turns out to be a small contribution to an account's reputation I might  stomach that, as long as the other contributing factors are reasonable IMO.

[puppies]

I am really not a fan.  I don't want my transaction volume, or trading habits visible to anyone that has my account name.  I also am concerned with the governing body that would be required to officiate such a system.  Even if it is just based on forum posts or something like that.

Requiring users to burn a certain amount of BTSX with the strength of their vote being dependent upon how much BTSX they burn is a step in the right direction.  I would still rather see the bter account problem solved by those that are directly involved though.  Why does Bter.com not have a link you can click to fill out your wallet?  Why aren't users copy pasting the address?

[Thom]

I would expect very little personal interaction generally. I believe this is about reputation, trust and transaction confidence. I'll step back and let bytemaster elaborate on what he had in mind.

[Riverhead]

How much personal interaction are we expecting on an exchange? Or is this more about the tool kit and other DACs?

[Thom]

You could take the amount of posts a person has made on Bitsharestalk and use that as criteria for earning a badge.

That's some eagle-eyed criteria. Might need an additional factor or two.

Exaclty. Luckybit is right, that particular example is objective and easy to establish as a basis, just like the "newbie", jr member", ... "hero" designations of this forum. the SMF allows admins to set the number of posts for each objective label. This is precisely the type of criteria I am in favor of. The "rules" for such assignments can't be anymore objective than numbers.

However, as donkeypong says, that's only one factor, and it may not be that important compared with others. My point is the more factors you can gleen universally and objectively from transaction stats, like the number of transactions, the easier it will be for everyone to make normalized judgements.

The aggregation of the factors into a single label, badge or color becomes a problem for how individuals utilize the label in their decisions. The aggregation into a single badge or label might be very useful for some and be terrible for others, depending on how each factor is weighted.

As a compromise, I could see the value of a badge or label as long as the criteria used to determine it is clearly defined. It would be useful first to take a poll of what factors people here use to evaluate transactions and trust levels. If each of those factors can be measured by the code and surfaced as a number, individuals can apply whatever weighting they want to each to come up with their own label or badge.

For example, and these are purely for the sake of making my point:

number of forum posts (0 - 100 = 1, 100 - 500 = 2, 500 - 1000 - 3, .......)
amount of time online (% of 24 hr day)
amount of time away (% of 24 hr day)
how many tips held (< 50 = 0, 50 - 100 = 1, 200 - 500 = 2...)
how many successful transactions
avg amount ($ value) of transactions
most traded asset (BTSX = 10, BTC = 5, BitUSD = 4, BitGLD = 3 etc...)

Sum all of the weighted criteria above to come up with a number, that number determines the badge image/color/text.

The weighting (to right in parenthesis) could even be set by preferences in the account. Which factors are used may need to be limited, or not I don't know. Which factors are available in the client vs. on some "account rating" website I don't know.

I believe if badges can be issued based on criteria that can be influenced by an organized  group of account holders it will lead to problems. In the example above the "badge" is very subjective based on the weighting criteria each user defines, but the data that supports that is highly objective, obtained from the blockchain or other public info.

I could imagine some sort of "universal weighting profile" that gave each factor equal influence over the final number, and that weighting critera could be considered a normalized reputation value. But individuals could set their own personalized weighting and their badge could be different. A public and private badge if you will.