Voting has to be one of the least consequential things we do—at least from the standpoint of affecting an election. My vote has never changed an election, at least directly. And the closest I’ve ever come to influencing even indirectly an election was to canvas a neighborhood trying to drum up support for a bond issue to build a new high school where I live in Whitewater, Wisconsin. I knocked on some doors, encouraged other citizens that seemed positively disposed to the cause, and saw the bond pass later on that month by 27 votes.
Yet the concept of voting resonates in America. With every change of administration in Washington D.C. and in our statehouses, we brag about the peaceful transfer of power, almost as if the actual policies and conduct of the new power wielders is incidental. However, we are now being told in the aftermath of the 2010 Republican sweep that “elections have consequences,” the nature of which we can examine by looking at the function of voting in Facebook.
One of the most significant of Facebook’s product innovations, according to David Kirkpatrick, author of The Facebook Effect, has been open culture platform translation driven by voting informed choices (Kindle ed. 4792). Translations are collected and then arranged in a list to be voted on by a translation team consisting of Facebook users who are supposed to be native speakers of the target language. Translation thus speeds along. Kirkpatrick points out that the French version was completed by 4,000 users in less than two days (Kindle ed. 4802).
Voting is an ongoing process and will continue indefinitely as long as the dynamic environment of Facebook continues to change. According to Michal Burger, an engineer at Facebook headquarters in Palo Alto and the acting administrator for the translation project for Slovak, when someone adds new text to the Facebook site, expressions appear in the translation application for normal users to translate, usually along with some context, but often without:
Even with provided context, strings posted for translation can be quite challenging (‘Babby is formed” Is it a test phrase inadvertantly sent to the translation community or an urban dialect riff on the notion that the user has committed and error?). Some rudimentary grammatical structures are available to mediate between languages, so that, for example, distinctions of grammatical gender in inflectionally rich Slovak and the three way case sensitivity to quantification of the target language can be reflected in the Slovak translations:
In this manner, translations are collected and then presented to users of the translation application for a vote.
Users may vote for or against each alternative listed. According to Burger, “translations are automatically chosen after they get ‘enough’ positive votes, there’s actually a very complicated formula that determines this. It takes into account both positive and negative votes, votes of other translations, and the ‘trustworthiness’ of people who voted on them (personal communication, 3/9/2011).
Though the translation is described as “automatic,” some of the decision making clearly comes from the side of the site administrators, both in the design and implementation of the delivery system for translation and even on a manual, case-by-case basis. Burger again: “Translation admins (such as me) can override this decision [=voting outcome ML] and unlock a given phrase for translation again, or choose a different translation” (personal communication, 3/9/2011).
Kirkpatrick writes, “Rather than ask its own employees or contractors to spend precious years translating the site’s three hundred thousand words and phrases into numerous other languages, Facebook turned the task over to the crowd and found an enormous amount of wisdom there”(Kindle ed. 4794). However, internal control and the ability to override voting outcomes show that the translations are not quite as grass roots as Kirkpatrick leads us to believe. The 20 or so most widely used languages have professional translators employed by Facebook to administer and assert quality control over the translations (Burger 2/26/2011). Other languages supported on Facebook, like Slovak, have Facebook employees administering translation as an informal add-on to their normal, unrelated workload. Because Facebook wants to get such things as the Help Center, Statements of Policy, and the Developer Sites right at a higher level of accuracy, Burger indicates that languages without support of professional translators may remain forever untranslated in this regard (see my blog entitled “Limits of Translation”).
If voting on translations of Facebook shows at least some user control of site content, it is and open question whether voting on other aspects of Facebook is similarly consequential. That doesn’t mean there aren’t other interesting, perhaps more important consequences, at least from the perspective of site administrators. Kirkpatrick describes how Mark Zuckerberg, Facebook CEO, has used voting as a strategy for redirecting user blowback on a range of site innovations. In February 2009, Zuckerberg was hit with an angry insurrection over change in Facebook’s Terms of Service, which expanded Facebook’s rights to its users’ content into perpetuity (TFE 5334). Threatened with lawsuits from various consumer protection organizations and buffeted by criticism from the blogosphere and mainstream press, as well as a 100,000 member user group “People Against the New Terms of Service,” Zuckerberg responded by conceding some infelicities in the style of presentation, invited users to offer suggestions, and announced revised documents in the subsequent week, along with an invitation to vote against them prior to their becoming official.
Zuckerberg proposed that a negative voting outcome in an election in which at least 30% of Facebook users participate would be binding and force further revision. As Facebook had, at the time, more that 200 million users, this meant that it would take a vote involving 60 million users in order to defeat any measure . . . a quantity vastly beyond the numbers of people that had on various occasions weighed in on any controversial development (and 600 times the number of complaints directed at the original change in Terms of Service).
The subsequent vote (in favor of the new changes but involving only 660,000 votes) defused the controversy and mollified detractors, who were brought into the process of dialogue about changes. Zuckerberg resolved to handle criticism with similar such votes in the future, launching a “Facebook Site Governance Group” as a resource to disseminate votes and to invite user collaboration (TFE 5362).
Kirkpatrick quotes Zuckerberg: “History tells us that systems are most fairly governed when there is an open and transparent dialogue between the people who make decisions and the people effected by them” (Zuckerberg, TFE 5353).
Voting as a form of dialogue in order to justify the agenda of decision makers and to allay users’ attempts to undercut those decisions—this is revealed as an additional, perhaps even more consequential function of “voting.” It’s more important for the election to happen, regardless of outcome. In this way, decision makers can declare buy-in, while asserting their will with a more accommodating “electorate” to deal with the outcome, regardless of their self-interest in the consequences.