Distributed Expertise

Last post I talked about how the WikiTrust project provides an option to colour-code every word in Wikipedia by reputation. [For a relatively no-BS assessment of the multitude of misleading reports on the subject, see here.]

The WikiTrust labs use edit longevity as the principle measure of reliability. Now let’s be clear: longevity here refers to the lifespan of an edit in the abstract sense: the WikiTrust system rates reliability based on the number of times the contribution is left untouched or reverted-to by subsequent editors. A better word might be survivability, which should theoretically be a great measure of merit in the cut-and-thrust intellectual arena of Wikipedia.

If a contribution survives another editor’s red pen then it accrues reputation; even more so if the editor had a high reputation themselves. The algorithm even factors the “distance” in words from subsequent edits: if an editor alters the preceding sentence and the following sentence and leaves the middle one untouched, it implies more approval of the middle sentence than if she had altered something elsewhere in the article.

Implicit approval is the crux of the issue here. WikiTrust, like most systems which don’t rely on active flagging, is a negative means of determining reputation. Kind of like Edwardian London; a page’s virtue could be assumed to be intact if you hadn’t heard otherwise. By contrast, as thousands of students and dozens of otherwise reputable journalists have found to their cost, regular Wikipedia is more like Elizabethan London: it’s best to assume a page has no virtue at all unless you hear it from elsewhere. [citation needed]

Except in this case, the ‘virtue’ accrues to an editor. WikiTrust stores a reputation rating for your profile. In the past, whoever last edited a Wikipedia page, no matter how small their contribution, could fairly be considered responsible for the content of the whole page (which is one of the reasons suing for libel on Wikipedia is such a tricky business). With WikiTrust, because authors continue to accrue reputation as long as the word is in use, the original authorship of a word or section is preserved by the system so long as the word is not substantively moved or altered.

But what happens if someone rewrites a section, but doesn’t add any new information? Who gets the reputation then? And what about the fact that editors will now have a rating, always in flux, based on others’ opinion of their contributions? Wikipedia editors have always been able to wake up, log on, track who has changed their articles and, if they want, take it all very personally. But they’ve never before had a tangible indication of their supposed value to the community, entirely mediated by others.

So here’s another way in which an experimental Wikimedia extension could have enormous implications for the Social Web. Social currency is at the centre of so many conversations about the web right now: whether it’s the need for some form of Social Proof in online sales to something tangible like Facebook Credits, probably the closest thing to a credible potential global internet currency.

Once you start awarding points, then Wikipedia becomes a game. And once it’s a game, you will get people trying to game it. More on that in my last post.