User-Created Trust Networks in Second Life [UPDATED]
I’m a big fan of the idea of user-created trust and ratings networks, although most of them seem to not work very well. I recently came across two interesting examples in the virtual world of Second Life, though, which are worth pointing out here. The first is Dale Glass’s TrustNet, a fee-based subscription system with a slightly confusing Web site (here’s the basic product description). The second is known as BanLink, created by Travis Lambert of »The Shelter« and Mera Pixel. Both systems seem to have their good and bad points. I present them here merely as example of ways to address issues of trust, ratings, conflict resolution and land bans in a virtual world like Second Life, not necessarily as product endorsements. I came across BanLink in a blog post by Mera, and TrustNet through a BlogHUD post I’ve since lost the link for, apologies.
TrustNet is the ratings system; BanLink provides a ban list. Both systems work by applying the ratings or bans of your first-order associates, as well as your own. TrustNet also applies the ratings of your closest associates’ closest associates (i.e., those who are two links away from you in your network). More detailed descriptions below the fold:
TrustNet works like this:
• If I rate someone, I see them as +/-10
• If one of my closest associates (i.e., someone I’ve given a rating to) then rates Avatar X, I see Avatar X as +1. So in the illustration above, Alice sees Bob as +10 because she rated him positively, but she sees Carol as +1 because Bob rated her. Scores are added together, so that if Alice were to rate Carol, Carol would then be +11 for Alice.
• Negative ratings are not followed, so if Alice had rated Bob negatively, and Bob had rated Carol positively, Alice would see Bob as -10, but she wouldn’t have a score for Carol, since the branch would end at Bob.
• Positive branches are followed one step further, with a further 1/10th weighting. In the illlustration above, Alice sees Eve as -0.2. Alice has rated Bob as +10, so Bob’s ratings count as +/-1 for her. Bob has rated Carol and Dave as +10, so Alice sees them both as +1. Carol and Dave have both rated Eve as -10, so Alice sees her as -0.2. If they had both rated her positively, she would see her as +0.2. Bob sees her as -2 because he is one step closer to her in the tree.
This is an interesting system, but in practice these things usually depend on broad adoption for them to work. I’m not sure how many members TrustNet has at the moment. [UPDATE: Dale tells me he has “240 users, and growing at a pretty decent rate.”] Dale has to charge a subscription fee for the service to defray his server costs, which I imagine makes it harder for people to just pick up and use. But I do like how the system provides an immediate way to assess an unknown person, at least in terms of pos/neg ratings your friends have made — which, to be sure, is an imperfect basis at best.
One nice thing about TrustNet is that the ratings are not public and not universal, they change based on your point of view. This definitely seems better to me than just looking at the cumulative mass of anonymous ratings that can go into such scores — though that system works fine for eBay, for the most part. It’s in sharp contrast to a pos/neg rating system like the one found on Xbox Live, where you know only the percentage of people who have rated you positively or negatively, not the number of people. Thus a single negative rating means you have a 100 percent negative rating score. The problem here is compounded by the fact that the Xbox Live screen reports this as 100 percent of “people you’ve played with” rather than “people who have rated you,” which is actually the case.
BanLink is a different concept, though it also creates a network of trusted links that helps users assess people they don’t already know. BanLink is basically a collection of SL communities’ ban lists. It’s not a universal ban list, however. Instead, you as a community administrator can choose which other communities to trust; the people banned from those communities are then automatically banned from yours. If one of those people then clears their name with you, you can unban them from your community, even while they are still banned in one of the communities you trust. (This last feature isn’t explicitly spelled out on the BanLink site, but that seems to be the case. If it isn’t the case, it should be.)
That’s good as far as it goes, but one interesting feature of BanLink is that the communities’ ban lists are all made public. Those who’ve been banned can dispute the judgement, and when people are unbanned they are taken off the list. I like the transparency of this. (It also makes for great reading. One person was unbanned because they “dressed during ban process,” for instance.)
Judging from the available ban lists, BanLink is being used by some 40 communities in Second Life, including clubs, hangouts and retail locations. These include Luskwood, one of the largest communities in SL. What would be interesting would be to see the network of trust links among all those communities. This would perhaps give a better indication of whether the networked aspect of the service was getting a lot of use. It would be interesting to know how many banned people are on the lists as well.
In short, I’ve no idea how well these services are working at the moment, but I do think they’re a hopeful indication of ways that Second Life residents might cobble together tools for self-governance. There are, of course, many similar things around on the Web; these are not ideas original to SL or unique to virtual worlds. I’ll be interested to see how these things grow in Second Life, whether there are similar things out there, and whether one or more can become truly useful, widespread tools.
[Oops, I’d meant to include in the original post a link to this extensive blog post on ratings systems, which Raph Koster linked to yesterday. Complete with algorithms, as Raph points out.]
Feeds
I’ve used both systems. TrustNet is very cool, but I agree, it’s going to be tough without charging. The way I offset my server costs with SLTrivia is by getting content creators to sponsor, and making it free. A system like TrustNet is hardly gameable except at very local levels. Dale Glass did a very good job on thinking out a social trust system that could only be abused at an absolutely minimal level.
I’m a big fan of BanLink; I trust Luskwood and the Shelter, and run Indigo’s own list as well. The upshot of this is, if we ban a troublemaker at the GNUbie Store in Indigo, then the same griefer can not move on to The Shelter or Luskwood, and vice versa. All three locations have been frequent targets of annoying griefers for some time and the hard work Travis put into building this system has already saved me quite a bit of time.
I oppose these fascistic little systems. They go against the idea of an open, free, democrat society by imposing the whims of tiny cliques through self-replicating memes.
I also oppose Travis’ BanLink system. The idea that there is only trust as a basis, and only intelligent perusal of criteria and lists to see if they are a fit is wrong. In fact, other forces simply scoop up everything in BanLink and use it to make mega-ban lists.
I’ve written extensively on cases of how this system continues to discriminate across SL. Example: somebody who is 0 days old appears at the Shelter with a gun. They don’t obey the first order to put it away, and then are banned permanently and put in Ban-Link. They are merely guilty of reading the Lindens’ web page where they advertise cool avatars in armour and imply SL is a haven for shooters. They may go on to do other things — like rent and build in my rentals. But because they’re on the master shit list, they cannot change, reform, or have any other freedom because of the discretionary use of this list.
Or take FlipperPA. He has banned me from Indigo near the Gnubie Store “just because” he doesn’t like my writings or ideas or my criticism of him. I’m no land griefer or shooter or grid-crasher, but I’m banned from a place that sometimes is used for public meetings. So, for example, if Flipper decides to bring in a public figure like Richard Bartle, his event isn’t really a public event because he’s using his paroachial, subjective ideas to set up a ban list and ban me from a pubilc event, not for any justifiable cause based on actual griefing criteria, but “just because”.
Or the example I’ve discussed at length on my blog. The W-Hat makes an object out of something of mine — a griefing ship called USS Prokofy Neva. Somebody sees this object and decides I’m part of W-Hat and bans me from all the ageplay furry areas just because W-Hat routinely attacks them and have spoofed my name. I’ve also discovered how W-Hat will show me as creator or even owner of self-replicating objects (you can confirm this with Lindens if you find this hard to believe, yes, it is indeed possible and done not only to me) that crash the grid, prompting landowners all over to ban me or send me hate mail when they find the grief balls on their property (and that was their idea, of course, it’s an uber-form of griefing to make others think someone is a griefer, and not you).
Furthermore, without any public discussion, merely with an SL Views where the Lindens and Travis and others got to hash this out, the Lindens are very likely to GOM BanLink, and are relying heavily on the Ban-Link concept to make their “residence governance tools”. No one else got to submit any body of evidence or conceptualization to them about how this will be misused.
The other day a tenant of mine logged on to find someone in her yard. She felt this was the same person who had bothered her on a previous log in and ejected her and asked her to stop coming there. The person ejected took a fit and became shouting across the sim that she would put this tenant on a master banlist all over SL (and is in fact likely in a position to do so). She said she was TP’d to the area and just landed accidently. The incident might well be a misunderstanding on both sides, but it has been resolved on both sides through the use of weapons — bans are weapons, as they don’t rely on any kind of judgement or weighing of facts or rational use, they are just on/off switches. This sort of incident though could well cost a new resident entrance to hordes of sims where the idea replicates that it’s ok to keep rolling master banlists.
It’s funny how you assume that the use by 40 clubs represents some well-parsed, elaborate, rich and deed “trust network”. It represents, if anything, a “distrust network” fueled by hysterical suspicions.
Often I see a club owner get on SLAM, the existing group for sharing griefing information, and complain that someone has come to a club and IM’d his patrons advertising another club and its offers to a visitor. All that’s happened is that commercial entities are vying for customers. But the list is then used to spread the word that somebody has a commercial rival who should be banned, in the view of the list member.
Same for Dreamland — existing store owners use the list to complain about competitors they see coming in to rent near them — they spread the word on the list telling everybody not to sell or rent to their competitors. In the fierce, tribalistic world of Second Life, the word spreads, and soon everyone is hysterically talking about “the griefer” who in fact is merely a competitor in the economy.
That’s why the whole system of the “Better Business Bureau” that the FIC and their cronies keep dreaming up, or later iterations of would-be rulers of SL keep insisting on, is so suspect. It’s always used to create whitelists.
Honestly, Walker, with all your values and the values you come from, I can’t understand why you’d get ecstatic about these distrust networks. It substitutes the rule of law by some sort of open, public criteria, with offenses where people can learn the charges against them and appeal them in courts, with a tribal, subjective, fascistic system that enables each tyrant or warlord with a personal fiefdom to spread his enemies’ lists without recourse.
You may think it’s fine if little powerful and influential cliques ban me and spread me on a banlist, because you think it’s a marvelous social tool and people should have absolutely freedom to do WTF they want on their land.
But why would these powerful few — or their random and clueless followers in a “trust network” — get to set up a system where I am banned — or Pixeleen is banned as is also the case — for specious, political, drama reasons and not for actually having griefed somewhere.
So a negative rating can go on my profile, and then potential customers coming to see me see some whopping huge distrust-network black eye arranged by FlipperPA Peregrine, and I am denied customers. They flee thinking that this black mark I have isn’t because I question and challenge FlipperPA’s leadership of the SLCC, it’s because, oh, I must scam newbies out of first land or arbitrarily refund rentals or something.
Currently, we have a genuinely free economy where no Scarlet Letters and yellow stars and black marks drag around on a person denying them freedom of movement and access to the economy and the marketplace.
When the Lindens and their friends get done with this, we’ll have a highly sanitized guild marketplace where the masters get to ethnically-cleanse out people by identifiers that they use subjectively.
Currently, none of these systems have recourse or appeals of any sort, and the Lindens, while acknowledging that appeals might be needed, don’t have any system or thinking in place to handle appeals and don’t want to be dealing with appeals.
When I first criticized Ban-Link it was for lack of transparency and then later they publicized their list on the Internet. That made it easy for people simply to grab it all as a ready made list because they figured the cute dog couldn’t be wrong, or because they just wanted anyone else to do their homework for them.
The first problem you see with any ban list is that it rapidly fills up with dead accounts because people abandon them and make alts or the Lindens remove them. Currently, I don’t know of any way to put a script on scraping avatar data in the people list to see if someone is now gone from the list (but no doubt there are many who’d like this function and it’s a matter of time before it is created).
In my view, there’s a completely different way to be dealing with rampant griefing, and that is a public police blotter that that has the names of the offenders, the prosecuting Lindens, and those who abuse report.
I probably deal with even more annoying griefers than the Shelter, because I’m dealing with communities that aren’t just for newbies, such as the Shelter. Banning makes sense at the granular level of a tenant’s land where he makes the decision, and it makes sense when you are the owner of a sim. It doesn’t make sense to spread this discretionary and arbitrary judgement through all of SL.
Hello, some corrections :-)
First, a score of +11 should never happen, as an user’s own opinions are considered absolute. If you rate somebody personally, you see will +10 or -10 and that’s where it stops. This is because I think that you should be able to override your friends’ opinions. Since TrustNet is intended to support land access control as well, it would be very bad for it to go against your own opinion by doing something like kicking out from your land people who you like, but happen to be unpopular.
Longer term, I plan to allow the user to choose the evaluation function (from a list provided by me), if they don’t like the default. The data would remain the same, just the way of processing it would change. So if somebody wants it to work the way you said, it could be done.
While it does cost money to subscribe, the TrustNet system can be tested for free. You can get the HUD on SL Exchange, SL Boutique, or from an in-world user (it’s no trans but there’s a “give copy” function) and try it for free for two weeks. After that you have to renew the subscription, which can be renewed for any amount you want. For example, it costs L$50/month, but you can pay L$25 and get two weeks.
Also, ratings aren’t completely public, yet not completely private either. When Alice sees Carol, she can ask the server where that score came from, and see it’s due to Bob’s rating. However, it’s not possible to download the whole database, or to see all the ratings given by an arbitrary person.
By the way, TrustNet is capable of doing what BanLink does, in pretty much the same way. At the time it’s somewhat less tuned for that specific function, but it’s perfectly doable. The object required to do that (a security orb) will become available soon.
People who make ratings should take public accountability for their ratings. NO one should be subjected to negrates they can’t see the origin of — under the old SL system, as a recipient at least you could see who made them as you received a notice. That was fine by me. Only the hysterical screeching by a few concerned by the few negrates blemishing their perfect record since beta enabled the downfall of the old ratings system. It wasn’t a bad system. It was gamed because they paid out money for it. It should be free or cost much less than $25.
I don’t see why you should absolve people of negrating and deprive them of public feedback to mitigate their subjective whims. But that’s what you do if you have a system that doesn’t show all ratings or who made them as an aggregated public figure.
I thought Will Wrights system in the Sims was promising as well. The green balloons (positive) or red balloons (negative, from “gloving”) were a way that you could see not only ratings, but also comments from people and the provenance of each comment, and how that person in turn related to a network.
This enabled you to tell at a glance, for example, if you saw a red balloon come in for no good reason, that in fact it was created by someone with a network of say, parcel griefers or some rival club to your own. Anyone else could also contrast and compare your reds and greens. One of the favourite passtimes of TSO was to wander through balloons finding connections to people.
It is possible to see how the scores are calculated. For instance, take Alice who meets Eve and without having met her before sees -0.2. She can use the “explain” button in the HUD, which would return the following:
Alice
Bob
Dave
Eve -0.1
Carol
Eve -0.1
It’s also possible to attach comments to ratings, and to set an expiration date for them. Comments will appear in the output of the explain command (they don’t yet, but that will be fixed soon).
There are a few reasons for the system working it does at the time.
1. The plan is that the ratings are your own personal opinion. It’s not supposed to be a popularity contest. IMO you should be able to think whatever you want, and it’s other people who decide whether your opinion is worth anything or not.
2. What you say is hard to represent in SL, and earlier there was no website so there was no way at all to represent it in a good way. For instance, a few weeks ago I made a graph of all the people I can get a score for. The list was of about 900 people, and it took 9 1024×1024 textures to make the names readable.
Of course, now my own situation changed a bit, so what you want should be possible. I will certainly consider it.
BTW, a good example of how these systems are misused is FlipperPA Peregrin’s placement of me on Ban-Link merely because I criticize him on blogs. Though I’ve never committed any griefing inworld, due to my writings and my criticism of his behaviour, I’ve been placed on a mass-ban list which makes others ban me because the “trust-circle” idea is really thrown out the window on this issue — people either says, “Oh, I think Flipper will have a good ban list” and accept his word, or don’t even know who he is or care, they just figure anybody’s ban list is worth picking up.
Well, I sure noticed this late.
Prokofy, while such a thing is possible in TrustNet as well, in my system, a person’s own ratings override their friends’ ones. So if for example I decided to trust FlipperPA and it resulted in you being banned from my land, I could give you a rating myself, which would override anything anybody else thinks.
Also, if using it for banning, it’s possible to choose a score threshold that requires ratings from multiple people. For example, a ban on a score
Erk, apparently this blog doesn’t deal correctly with angle brackets.
I was saying that a ban on people with a score equal or less than -2 would require negative ratings from at least two people who you rated positively.