The Future of Reputation: a view from Wikipedia

As Mr. Richardson points out, commenting on the first half of The Future of Reputation is difficult.  Solove lays out some examples, but has not yet presented his solutions. However, I would like to add to his anecdotes by examining one of the most prolific reputation and defamation engines on the internet.

Let's look at Wikipedia.

(Solove devotes only a few pages to Wikipedia in chapter 6, so I'm not stealing much thunder by introducing the topic here.)

Control of reputation versus the free flow of information

Solove suggests that people might have rights over their own reputation. Some balance should be struck between control and the free flow of information in order to preserve our freedom.

This is a normative belief for Solove, but it begs an empirical question. Why can't the free flow of information correct itself?  After all, there are social customs for reliable reporting and against perceived invasions of privacy.  Conceivably, ethical reporting could thrive in the marketplace of ideas. Mr. Richardson writes about the battle between slow-spreading good memes and fast-spreading bad memes.  Can conscientious internet users maintain an ethical balance between useful reputation data and sensationalized garbage?

To me, Wikipedia suggests that self-regulation fails. I agree with Solove's premise that the internet disproportionately promotes undesirable elements of gossip.

There’s an asymmetry to reputation, especially when opinions on real people can be delivered pseudonymously. Malicious rumor mongers have never had a better platform for defamation. At the same time, there’s never been a better time for sharing positive or benign information, but the incentives are mismatched. Defamers want to spread their gossip far and wide, while positive feedback only seeks to be logged (if that). Moreover, information consumers find negative material more valuable. Shaming naturally spreads better, and the internet obliterates one of the best safeguards against malice: the reputation of the commentator herself. Traditionally, one would be worried about behaving boorishly or being caught in a lie, but, as expressed in this webcomic, anonymity breeds degenerate behavior.  

Wikipedia is a microcosm of the internet’s reputation problems: the site has a large audience (and is often the number one result in Google), but pseudonymous authors are protected from the consequences of their work.

Wikipedia: where no one controls reputation

Wikipedia fashions itself as an encyclopedia.  Although it has a radically permissive editing policy (most articles can be edited by anyone, without requiring even an account set-up), it has a focused culture. Wikipedia is not a tabloid, political blog, or humor forum.  A community of editors strive to enforce its standards.

Compared to the internet at large, Wikipedia editors are unified in purpose. Less than 2% of editors are responsible for most of the site's edits, and these people—including over 1500 volunteer administrators—bring anonymously-contributed content into line with the site's norms, policies, and guidelines.  There's even a sort of supreme court of Wikipedia (although they object to the analogy).  The Arbitration Committee or ArbCom, accepts cases when four of its members agree to hear the controversy.  ArbCom clerks open up evidence pages, and various parties post evidence and opinion, which the arbitrators distill into a ruling which commands the respect of all administrators.  No cogs in this bureaucracy are paid, and very few content changes are implemented from above by the Wikimedia Foundation. Essentially, Wikipedia is an idyllic cross-section of the internet—a confederation of users unified by strong truth-telling encyclopedic aspirations.

If voluntary privacy controls can emerge anywhere, we should see them on Wikipedia, but the site utterly fails to guard reputation, and decidedly favors the free flow of information over control. 

Over 269,000 biographies of living people exist on Wikipedia. The vast majority of them can be edited by any user without even signing up for an account.

The potential for abuse became clear in the 2005 Siegenthaler incident, which is described in Chapter 6. A bored dock worker decided to edit the biography on John Siegenthaler, Jr., a journalist and one-time associate of Bobby Kennedy. The biography falsely claimed that Siegenthaler was investigated in connection to Kennedy’s assassination, then spent several years in the Soviet Union. Although the claims were wildly false, they persisted in Siegenthaler’s biography for four months.

Wikipedia addressed some of Siegenthaler's concerns by passing a "biographies of living people" policy (WP:BLP).  This rule promises higher standards for such biographies, including a vow to remove claims without reliable supporting sources, but the “BLP problem” has never been fixed. Many editors believe that open and anonymous editing is the project’s central strength, so refuse to compromise these aspects of the site.

It’s thought that defamation should be corrected as it occurs, but distortion on Wikipedia is another case of mismatched incentives. The site’s foundation has weak incentives to prevent errors because section 230 immunity shields them from legal liability. Most articles are corrected by editors, who find some personal satisfaction in correcting errors, but this incentive only works for biographies that are watched by many impartial eyes. For example, the biography on George W. Bush is constantly assaulted, but errors don’t persist for more than a few minutes because of the attention it draws. A mistake about Judge Richard Posner might persist for five days (until Prof. Sunstein cleans it up). Siegenthaler, on the other hand, was defamed for months.

Solutions have been proposed, but no consensus appears likely.  Some editors suppose that “marginally notable” biography subjects should have a right to opt out of having a Wikipedia article.  A handful of articles have been deleted by sympathetic administrators who support their cause. An article on journalist Seth Finkelstein was deleted in this rogue fashion.  The deletion withstood the review process, and remains off the site. But subjects apparently have no right to opt out of their biography if they are deemed too notable. Hollywood producer Don Murphy is an example of this. Murphy stridently opposes the existence of any freely-editable biography on him, and has tried to get his article deleted for months. When an administrator deleted per his wishes in March, the deletion was overturned by overwhelming community opposition.  Murphy's demands to either delete the article or prohibit anonymous editing were deemed unreasonable.  One editor—a Yale graduate in his early 20s—commented on Murphy's supposed harassment.  Deleting the article, he said, would be "essentially endorsing terrorism."  Murphy in turn heavily linked an unfriendly donmurphy.net thread about the individual, propelling it to #2 in google.  Murphy still has an article on Wikipedia.

As a final insult, biography subjects are discouraged from editing their own articles because of the conflict of interest guideline (WP:COI). Subjects can be understandably outraged when told not to fix errors about them. See for example, Pamela Jones’ reaction. Jones is the author of the pro-Linux blog Groklaw, and took offense to a comment suggesting that she shouldn’t correct her own biography due to COI. Plenty of autobiographical corrections occur anyway—see e.g. Martha Nussbaum, who stepped in herself after a colleague was unable to affect similar changes.

Shaming and outing pseudonyms.

Some individuals have taken vigilante action against Wikipedia’s anonymity.  Daniel Brandt, a privacy advocate who became an important critic during the Siegenthaler controversy, began publicly identifying Wikipedia administrators on "Wikipedia's Hive Mind" in 2006.  Most administrators are not on the list—Brandt focuses on those who he thinks have pseudonymously attacked real people, or those who are particularly prominent in the community. Brandt sometimes uses the threat of outing in an attempt to prompt reform. 

For example, Brandt has recently discovered the identity of an influential ArbCom arbitrator called Newyorkbrad. This user is a Manhattan litigator in real life, and Newyorkbrad has momentarily abandoned Wikipedia in light of Brandt’s threat last week. Brandt has presented Newyorkbrad with several optional demands. If none of them are met, Brandt promises to post Newyorkbrad’s name and law firm on his site.  "It's time for Newyorkbrad to leave Wikipedia and save his career," Brandt wrote.

This dialog occurred on Wikipedia Review, a criticism forum, but some of WR's moderators took exception to Brandt's tactics. Newyorkbrad is generally regarded on WR as one of the “good guys” on Wikipedia.

Pseudonymous ax-grinding.

This brings me to my story.

The Siegenthaler incident was basically just simple vandalism—a user marred an article just for the fun of it. What happens when a pseudonymous editor has an ax to grind? Well, in the first place it’s difficult to detect. Problematic users are sometimes banned from Wikipedia, but because people can sign up for new accounts easily and can evade technological detection by using IP anonymity networks like Tor, no one can be decisively banned from Wikipedia. Nor can individuals be limited to a single account. I think this is a design failure of the site, but it’s one shared by nearly all websites on the internet—contributors cannot be verified, and no one can be sure that separate usernames represent separate people.  Apparently different accounts might be "sockpuppets" of each other, controlled by one individual. 

Enter Overstock.com.  Overstock is an internet retailer that claims its stock price is hurt by naked short selling—that is, short sells that don’t borrow the underlying security. Manipulative naked short selling is theoretically illegal, but most commentators assume that Overstock’s CEO, Patrick Byrne, uses it as a diversion from the company’s reliable record of annual losses. Byrne is a colorful character, and is best remembered for the conference call where he referred to a shadowy "Sith Lord" determined to drive down the company's market value.  A plaintiff shop has taken Overstock's case on contingency, suing several analysts and a hedge fund for their alleged criminal conspiracy.

For a time Patrick Byrne employed a publicist called Judd Bagley, known and frequently reviled on Wikipedia as WordBomb. Bagley came to the conclusion that an unkind editor on Wikipedia, Mantanmoreland, was operated by critical Forbes columnist Gary Weiss. Bagley created a site documenting his theories, Antisocialmedia.net, where he claimed that Weiss had started a second account, Samiharris.

Being from the Beehive State myself, I was aware of Overstock’s reputation for nuttiness, and paid no mind to the dispute until I saw an article in The Register, the British tech tabloid critical of Wikipedia. I decided to look at a dispute on the article Gary Weiss. Several editors (including Wikipedia's founder Jimmy Wales) wanted to include a single sentence stating that Weiss was critical of Overstock.com. Their sources included a New York Times column, and everyone thought it was a fairly neutral statement—a quick look at Gary Weiss’ blog confirms that Overstock.com is one of his main interests. User:Samiharris, however, was violently opposed to any mention of Overstock.com, and disliked the Times article because it quoted Bagley and might therefore spread "Bagley memes."  I concluded that Samiharris was not an neutral party—editors don't normally oppose citing the New York Times for fear of spreading certain memes.

More than a month later, WordBomb shows up with a disposable account and requests a “check user” of Samiharris and Mantanmoreland. Normally, logged-in users do not reveal their IP addresses, and no user can check them unless there is prima facie evidence of abusive editing. Surprisingly, the check users decide to honor the request, even though they know WordBomb made it.  They discover that the Samiharris account always edited from an anonymous proxy, so that it is not technologically possible to determine whether the two accounts edit from the same computer. This seems highly suspicious, and one administrator launches a sockpuppet investigation aimed to discover similarities between the accounts. 

I was honestly annoyed by the accounts’ responses to the investigation. Mantanmoreland argued that the entire inquiry was tainted by the initial request of WordBomb, and that no real controversy existed. On Wikipedia, procedural arguments are denigrated as “Wikilawyering” and viewed dimly by the community. When the sockpuppet investigation advanced to a request for comments, I resolved to present a knock-down case of all the evidence linking the two accounts. Thirty-four users signed my summary of the evidence, and only Mantanmoreland and four other accounts signed opinions dissenting on procedural grounds. The case was accepted by ArbCom.

Nicholas Carr argues that pseudonymous data cannot be secreted away forever, and that's probably the case on Wikipedia. After making over 5000 edits spanning two years, it was noticed that Mantanmoreland had once copyedited the article for a small coastal city in India, Varkala. Four months later, Gary Weiss announced on his blog that he was on vacation in Varkala. One user suggested that we check Mantanmoreland’s editing during the period for evidence of decline. The results seemed surprisingly clear to me.

At that moment, the nature of the case shifted from an inquiry about sockpuppets to arbitrator concerns about slandering a real individual—Gary Weiss. Overstock CEO Patrick Byrne wrote a blog post about Gary Weiss, claiming that “All but Gary Weiss’s most dogmatic defenders were silenced, however, when a law student from Chicago published a graph…”    (I would like to state that I'm not suggesting that these accounts were certainly operated by anyone.  It's possible that they were constructed to discredit Gary Weiss.)  ArbCom refused to reach the issue, but meekly concluded that the accounts' editing patters were "suggestive of . . . a relationship between the two accounts."  Hours after the case closed, Samiharris was permanently banned by administrators in the community, by what some called a "lynch mob."

Mantanmoreland has not edited Overstock.com topics since.  Articles formerly maintained by Samiharris and Mantanmoreland have been reclaimed by the others, and some of the changes have been dramatic.

Whoever was behind these accounts, they succeeded in maintaining an anti-Overstock slant in a constellation of articles for many months.  They were only rooted out because of the persistent harping of an organized corporate adversary.  Anyone on Wikipedia could have checked their work, but because they were more motivated than impartial editors, their point of view prevailed.

I believe this is characteristic of the internet.  High profile individuals might have effective error correction online, but a dispersed preference for accuracy does not trump the focused motives of a defamer.  Solove's central observations ring true for me.

Comments

[Dan Jones]

I don't buy the argument. You obviously are much more involved with Wikipedia than I am, but Wikipedia seems like a model going forward in balancing the pseudonymity of the internet with editorial correction. Your examples only press the point. Most telling is the idea that four months in the Siegenthaler case is a long time, as opposed to the five days of Judge Posner or the five minutes of George Bush. With print biographies, errors go uncorrected for years (or forever), until a new version comes out. Of course, the Siegenthalers of the world probably wouldn't warrant a print biography in the first place. Which brings me to my second point: what was the harm there? You mention that the higher profile the page, the more rapid the correction. But that seems good, not bad, in the sense that higher profile figures are also generally more invested in their reputation and their Wikipedia entries are going to get more page views, so there is a greater chance of harm. So Wikipedia's editors have priorities in the aggregate that roughly match the preferences of its users. It would be strange to have the reverse, and I wouldn't discount "a dispersed preference for accuracy."

Of course Wikipedia has problems and its editors fight an uphill battle against determined troublemakers. But the conclusion "Wikipedia suggests that self-regulation fails. I’ve come to believe that the internet disproportionately promotes undesirable elements of gossip" seems like a stretch. Disproportionate to what exactly? I think the ratio of accurate content to rumormongering on Wikipedia is fairly high.

[Frank_B]

I should clarify. By "disproportionate" I mean that the negative aspects of gossip are enhanced online more effectively than the positive aspects.

You've expressed a view about Wikipedia known as "eventualism"--that more eyeballs will somehow inevitably fix the site. The problem is that Wikipedia is now over seven years old, and this aspect of it never improves. I've been editing on it for more than four years. The opportunity for abuse is greater now than it ever has been. With over 2 million articles there are just not enough eyeballs to go around. And again, most people don't really care about people they've never heard about. Lots of folks are looking out for George W. Bush, and a few might check on Judge Posner, but most of the hundreds of thousands of people on the site do not have any kind of safety net at all. People invested in low-profile biographies are likely to be:

1. The subject himself.
2. The subject's ax-grinding enemy.

Ideally, we wouldn't want either party editing an article, but this is precisely what happened in the Mantanmoreland case, and it's very common. WordBomb was banned for his conflict of interest, so Mantanmoreland was free to shape the articles as he saw fit, and most people don't care much one way or another about Overstock.com or naked short selling. Reputation cannot be reliable in such an environment, and this is the internet.

[Dan Jones]

Forget eventualism: I don't think Wikipedia is broken now, so I don't understand the phrase "eventually fix the site." In terms of social welfare, the benefits of Wikipedia seem to far outweigh the harms of abuse of low-profile entries. If, as you say, "most people don't care much one way or another about Overstock.com or naked short selling," then who cares whether the content is accurate or biased? If I had an entry on Wikipedia, it would be extremely low-profile. Let's say only myself and my archenemy (call him Vikas) care about the content. If we go back and forth with biased edits, what is the harm? It's just an irrelevant little spat. True, if some third party eventually comes along to read the entry, they will get inaccurate information, but that is a very minor cost compared to the general utility of Wikipedia.

[Frank_B]

You seem to be saying that it shouldn't really matter if people are smeared because the site overall is good (never mind if it could be better by dispensing with anonymity or by locking down biographies). I find this position morally questionable. Why should anyone have to babysit their own entry for factual errors? Ruben's comic puts the futility of online error correction very well. What about people who are hurt by an entry they're totally unaware of?

A student looking up the article on Martha Nussbaum last summer doesn't know and probably hardly cares whether or not she perjured herself in the Romer case. But Martha Nussbaum sure as hell cares: she likely cares as a potential future expert witness, as a teacher, and as a human being. I consider that a failure. If policy had been strictly applied to her, the changes would have been reverted, and she might have been banned by an established editor. Most edit wars eventually have a winner, and it isn't always the side of truth and justice.

Wikipedia has basically built a massive library of potential defamation billboards for any passing person with a grievance. If you write something that seems plausible to the uncaring masses, it will probably stick. Unlike many of Solove's examples, what you write doesn't have to be true or even a very good meme. But you can shape an individual's top ten Google hit for any purpose whatsoever.

If Wikipedia were bound by defamation law, or even ethics, it would remedy this situation, but no individual or group of individuals has a strong incentive to do anything at all. As long as the site is seen as generally reliable (as you see it), there's no cause to act. Editors and the foundation cannot internalize the cost of libelous externalities.

This problem is not unique to Wikipedia, but few sites can claim Wikipedia's Google rankings.

People who "think the ratio of accurate content to rumormongering on Wikipedia is fairly high," actually make defamers more powerful than they ought to be. If people didn't make such an assumption, then defamation would matter much less--Wikipedia would be treated more like anonymous blog comments.

[Sarah Cottrell]

I think the point that Frank makes about the generally perceived quality of Wikipedia making it more dangerous is a good one. Personally, I'm not willing to say that someone else's concerns about inaccuracies on their own Wikipedia entry should be outweighed by the generic utility of Wikipedia. Some of the posts have suggested that if someone is only a little defamed, or if not too many people read the slanderous statements, then it's really not that bad. I think this seriously underestimates the potential damage that can be caused by slanderous statements, and also is a view that is likely only held by those who have not been the subject of internet slander. While none of us has likely been the subject of internet defamation, I think it's important to realize that part of the potential of the internet is that any of us could. What view would we have then?

[Vikas]

I disagree with the notion the because Wikipedia generally gets it right, we shouldn't worry about these minor problems within unseen articles. Two points. First, Wikipedia is helpful because it is comprehensive. I don't necessarily go there to learn about who George Bush is. Most of us know who he is and have available so many other resources to learn abou thim. I go to quickly discover information about slightly more obscure figures. Therefore, to the extent that these less popular articles are more likely to have errors, it might be disproportionately harmful to readers. Second is the problem of inability to distinguish. Let's say 5% of Wikipedia is riddled with errors and the rest of it is correct and useful. How do I tell which 5% is the false part? I think that's the reason most scholars still refuse to cite to Wikipedia. The general utility of Wikipedia decreases because of the errors in a small fraction of the substance.

[Graham]

The example that you site does not strengthen your case. The Patrick M. Byrne article in Wikipedia now reads like an extended press release.

You say that Byrne "is best remembered for the conference call where he referred to a shadowy 'Sith Lord' determined to drive down the company's market value," but the article now makes no reference to the Sith Lord embarrassment. Instead it contains prominent reference to his being listed on a "top innovators" list in some magazine five years ago, and getting a regional award by his accounting firm.

It seems to me, based on this evidence, that Overstock is a manipulator of Wikipedia and not the innocent victim that is portrayed in your item. Also I note that you are exceedingly modest about discussing your personal involvement in editing the article, such as here:
http://en.wikipedia.org/wiki/Talk:Patrick_M._Byrne/Archive1#WEIGHT_and_Nocera

[Ruben Rodrigues]

I'm sure many of you have seen this (somewhat old) study before, which finds that among scientific articles Britannica has 3 errors per entry compared to Wikipedia's 4 errors per entry: http://blogs.nature.com/wp/nascent/2005/12/comparing_wikipedia_and_britan_1.html

Of course, no one can simply go ahead and defame someone in the pages of the Encyclopedia Britannica, but I can't help but feel like one of the flaws in Frank's argument is the reliance on "what ifs." Frank points out a lot of bad stuff that can happen, but these scenarios have tended to be the exception and not the rule. I fear that imposing some of the restrictions Frank would like to see on anonymous postings would hamper some of the benefits wikipedia provides. I'm pretty convinced that CDA liability would simply lead to shutting down the site (though since Wikipedia attempts to impose some editorial control in some areas, its not clear they'd be totally free of liability under CDA 230).

Also, Frank focuses mostly on the defamatory power of Wikipedia biographies, I feel like people naturally discount information on Wikipedia when it comes to bios over other more "factual" entries. Your average Wikipedia user realizes that anyone can change an entry, and the motivations for doing so on a bio may be more malicious than for an entry on bituminous coal (Frank, don't go editing bituminous coal on me now just to prove a point ;-).

[Frank_B]

Wikipedia attempts very little editorial control. The entire structure is run by volunteers from ArbCom down. Things like copyright takedown notices are taken seriously, but the foundation does surprisingly little to exert editorial control. It was thought that the arbitrators did not want to rule on the "Mantanmoreland" issue because they would be individually liable. The foundation does not officially grant them power, but the community heeds them.

Also, after vandalizing the entry on wheat, I'm done with vandalizing Wikipedia to make a point. There's a policy against that. WP:POINT. I agree that non-biographies are much less problematic.

I disagree "Graham," of New York City: partisanship is actually my whole point. I would not argue that Overstock.com is an innocent victim, that misses the mark. Overstock.com probably *does* manipulate its entry, just like virtually every other self-aware corporation.

The point is that the ordeal shifted control away from one set of interested editors to another; neutral parties (myself included) have very little incentive to puppy guard an article, but the partisans can and do continue to fight--even taking their dispute on to unrelated blogs. I think this is why so many critics compare the site to a MMORPG. I hoped this would balance the playing field, and I'm a little bit surprised the "Sith Lord" press conference was taken out entirely. Should probably add that back in.

[Graham P.]

If the balance of power shifted from one set of interested editors to another, than I am at a loss to understand your point. Clearly the article itself is worse off, and omits, as you concede, significant information.

[Frank_B]

My point is that anonymity undermines Wikipedia's reliability on certain topics. This isn't a point specific to Overstock.com at all.

I had hoped that both partisans could be removed from the article. I now believe I overestimated the site.

[Barry Kort]

It occurs to me that the only realistic way of organizing community sites like Wikipedia (or even Wikipedia Review) is for such sites to establish their own individual social contracts, through which the participants in a community set forth their mutually agreeable terms of engagement.

On Wikipedia, Doc Glasgow sought to introduce a Wikipedia Responsible Editing Pledge, WP:REEP, which might have been part of a Wikipedia Social Contract if the site had been initially organized around the Social Contract Model.

In the absence of any such Responsible Participation Pledge, neither WP (nor WR) can anticipate a normative standard of behavior to which all participants voluntarily adhere.

On Wikipedia Review, Kato has long criticized Wikipedia as a site given to irresponsible policies and practices.

It's almost surely way too late for Wikipedia to convert to the Social Contract Model. As for commentary sites like Wikipedia Review, I don't know if it's possible at this late stage of the game, but I'd be comforted if WR's normative policies and practices developed a reliable reputation for responsible commentary that would attract and retain participants who subscribed to the highest ethical and journalistic standards in online media analysis, commentary, and criticism, including those Wikipedians who shared that vision and ethic.