Craig Newmark: Who Do You Trust?
Matthew Ingram recently chatted with Craig Newmark, the founder of Craigslist, and asked him what was the web’s next big problem. His answer?
The question of who to trust online, according to Newmark. To solve it, he believes that what the web needs is a “distributed trust network” that allows us to manage our online relationships and reputations.
[…]
Newmark called some form of distributed trust system “the killingest of killer apps” for the web over the next decade (he said he wasn’t sure that was the best way to describe it, but was trying out to see how it sounded). He talked about “reputation and trust ruling the web, just the way it does in real life,” and how he was looking to big players such as Google, Facebook and Amazon as the kinds of entities that would have the scale to handle such a distributed trust or reputation management network. And he said that despite some occasional missteps by both Google and Facebook when it came to privacy (Google Buzz and Facebook Beacon, respectively), he believed that both were acting in good faith and had a policy of “not being evil.”
[…]
Newmark said that as a society we needed to “get our act together and make this happen,” adding with a wink that the idea for the distributed trust network was all part of his “hidden agenda to move ahead on the web to try and save the world.”
The Web may be the only hope that the world has, and if some new, distributed model of trust is going to be developed it will certainly have to be coming from the web.
How might such a system work? What would it be like?
A few thoughts: the open follower model (a la Twitter) is likely to be the dominant social motif of most web apps, at least for the next decade. It is a loose and distribuable model, and the work on interoperable activity streams suggests that open APIs across many different social tools could lead to a distributed sort of trust.
Imagine how it might work. Someone, Carla Botts, writes a piece about her Malibu. Others who read it could use whatever tools provided by the technology in which the link or post is delivered, like Tumblr or Google Reader. Someone at Chevrolet wants to know if Botts is a trusted source of information, and runs some application that gathers information from various systems — Twitter, Google Reader, Facebook, Tumblr, and so on — and calculates some trust metric.
Here’s the hitch: while all the tools involved can run independently — although in an interoperable way — the algorithm for calculating trust should be completely open, or completely closed.
In the case of going completely open, the factors involved in trust metrics would be made public. This allows anyone anywhere to build trust measures into tools and services, but in such a way that the trust metric is dependable. Of course, people might try to ‘game’ an open trust system, just as people have gamed Digg and other ‘democratic’ voting based tools. Again, just as with Digg, a great deal of time and energy would have to be spent on discovering people fabricating trust. This approach is analogous to the Jabber XMPP protocol, which merely defines how cooperating servers talk to each other and authenticate identities.
The alternative is to create a closed trust metric, where the factors involved are concealed except in the most general sense. A server — or a network of trust servers — could constantly be updating trust metrics on individuals, in a way analogous to the way that search engines — like Google — are constantly recalculating page rank of indexed pages.
Both of these approaches have their merits, and their negatives. But I agree with Newmark that it would be useful — and potentially radical — to have a trusted trust framework that is not controlled by a market player — like Google or Yahoo — or any government. Newmark suggests there might be a role for government in such a system, but in a public-private setting, where checks and balances involved non-governmental groups.
[original comment(s) from /Message]
System won’t work if you spell “Carla Botts” one time and “Carla Boots” the second time. Unless you use social network analysis, and it reveals that Botts and Boots and are likely the same person!
March 19, 2010 | Valdis Krebs (valdis@orgnet.com)
Valdis, you’re right, which is why I don’t see a day far off when there is a browser plugin that recognizes names and suggests online profiles that match and enable linked references to identity (Flowtown does this already based on email addresses).
March 19, 2010 | Marc Vermut (marc@fine-points.com)
Fixed the mistyping, so that may make your comment a bit off, Valdis.
March 19, 2010 | Stowe Boyd
Interesting. One of the things we have to watch out for is not understand the psychology of things, and in this case that might apply. Trust is an emotional + rational process. It is not completely rational in the first place so regardless of the ‘information” computers make available to a person, the degree of trust is ideosyncratic. I could see something like you describe working for very unsophisticated people trying to simplify their lives and wanting to take the “say” of others so they don’t have to choose on their own (and I don’t mean that as a knock). I don’t see this as being attractive for others of a more critical bent. I don not “trust” anyone online until such point where they have demonstrated expertise show over time about what they are taking about.
10 million people might tell me to listen to that person. Popularity shouln’t determine trust. The other way around, yes.
Which brings up the second issue and one that is of great concern to me, and that is that social media pushes us to determine truth on the basis of popularity — most followers, friends, retweeted, likable, most positive and so on, none of which have any relationship to veracity.
From looking at people many others trust, particularly the social meda experts, I am often surprised at how often they are wrong or inaccurate in their interpretations of data, even in situations where it’s obvious.
…and those people are the influencers.
I know no more about online trust than most, so mostly I’m speculating here. Fascinating topic.
March 21, 2010 | Robert Bacal (ceo@work911.com)
A visualization for our #emergent era of #TRUST #networks,
courtesy of Gistics.com http://bit.ly/9imB5C
@VenessaMiemis posting reminded me.
March 21, 2010 | Cocreatr
The idea of a distributed web of trust goes back over 20 years. The PGP encryption key management system, and its many derivatives, created a distributed set of key servers, with users “vouching” for the authenticity of user keys (identity claims.) It worked technically, but did not take hold universally after many tries. (Some info here: http://en.wikipedia.org/wiki/Web_of_trust)
One of the issues with a universal web of trust is that people do not trust people unilaterally - this trust is established topically, temporally and for specific scenarios. Someone you trust to mow your lawn may not be someone you would trust to take care of a toddler, or to run your business. Programmatically assessing and categorizing the situational aspect of trust, and the difficulty in defining the realm of situation for a particular trust question, appears to be well past our ability to manage in software form (so far) and to extend well in to the reaches of artificial intelligence that are not even ready for consideration.
Attempts to generate programmatic trust determination already exist, in the form of Google’s search rankings. These are a sophisticated set of algorithms that attempt to categorize authority and trustability levels of items on the web based on references and other characteristics. Do you reflexively trust the top returns of a web search?
March 24, 2010 | Joel Foner
Joel -
I am not so sure of the direct relevance of the PGP system in this context. The Google page rank algorithm is potentially a path to go down, as are karma systems like slashdot. It’s a question of machines harvesting social gestures like page rank’s counting the links we make as votes, or analyzing email communication patterns to infer connections and trust.
I am not so certain as I once was that trust isn’t fungible. I used to believe that someone I trusted for advice on what are the coolest cars wouldn’t be a person to ask about where to eat or to watch my kids. But some recent evidence (which I will have to track down) suggests that when people trust others is is very diffuse, and not so clear cut as our intuitive categorization around topics might lead us to believe.
March 24, 2010 | Stowe Boyd
I tried to reason about this in some papers in the past, for example in “Trust metrics on controversial users: balancing between tyranny of the majority and echo chambers”
http://www.gnuband.org/papers/trust_metrics_on_controversial_users_balancing_between_tyranny_of_the_majority_and_echo_chambers-2/
I tried to back all my claims with empirical analysis on real trust networks (mainly derived from Epinions.com)
I’m totally for what I call local trust metrics: the code is open source, you run it locally on your computer, so this means that you can set your own parameters (trust horizon, propagation, algorithm, timeout, …). Of course most people will use the detault values of the most known open source trust metrics available but it is at least possible to not been told by someone else (google, the government, the matrix, whatever) “who you should trust”. What if the global trust metric tells you you should not trust your mother? ;)
March 25, 2010 | paolo (massa@fbk.eu)
Stowe,
The real problem is that just as somebody’s “terrorist” often turns out to be a freedom fighter, “truth fighters” online are often labeled as trolls. I think the worry is misplaced. People should be preoccupied with getting as many useful comments as possible, rather than making sure there is no barking going on… At best, comments are pointing out fundamental flows and truly helping out the author (if the pursuit of truth is really the ultimate objective) and at at worst they are just barks that are bound to die down relatively quickly.
Delia
P.S. I seriously doubt a technological way of figuring out trustworthiness would really work in practice. I think it’s telling that somebody like Craig Newmark has no problem complaining about trust issues and being taken seriously, especially in the wake of the ebay v. craigslist suit. It’s hard to see how he did not abuse craigslisters’ trust by having them believe he did not financially benefit from the ebay deal (to the tune of about $10 million, as it turns out) or when he vowed ebay didn’t get more than 25% of craigslist.
http://craigslistcriticism.blogspot.com/2009/12/ebay-v-craigslist-newmark-and.html
April 14, 2010 | Delia
Delia - I don’t know that trolls should be tolerated when they cross the line between striving to make people understand and personal attacks. And, by the way, your Argument Ad Hominem (argument of the man) regarding Newmark’s comments about trust should be devalued or completely ignored because of the ebay v craigslist issue is an example of the things trolls do. People’s ideas speak for themselves, and your comments rapidly veered from a discussion of trust to sharpening your ax on Newmark’s alledged actions in that case.
April 14, 2010 | Stowe Boyd
Stowe,
It was a short comment — that’s all I had time for. I really thought the first paragraph made my point (the P.S. was just meant to give a case very in point of just how doubtful it is that a system of figuring out trustworthiness would succeed in practice). And I didn’t say true trolls should be tolerated, just that the *focus* should be on getting useful comments. As far as I’m concerned somebody who has serious trust issues of his own has no standing preaching to people about trust. And I don’t see how pointing out serious documented concerns (I gave the direct links so people can see what I’m talking about) could constitute personal attack. You should always be able to point-out the truth — how could anybody fill attacked by it, unless they are hiding from it…
Delia
P.S. I gotta go… will check tonight (Eastern time) if you’d like to continue this — take care! D.
April 14, 2010 | Delia
Social anthropologist, clairvoyant, postfuturist.
