Issue link: https://cp.revolio.com/i/457819
THE DATA CENTER JOURNAL | 23 www.datacenterjournal.com 90% correct for each given user trait. Aer running the process on just six or seven traits (say, products the user is likely to buy), there is about a one in two chance that at least one of the results is wrong. In business, that uncertainty might only mean some lost advertising revenue. In govern- ment, however, it could mean the differ- ence between concluding that someone is a terrorist building bombs and concluding that he or she is just a chemistry hobbyist. In the realm of social media, any number of factors can contribute to biases in data-mining results. For instance, Facebook has a "like" button but no "dis- like" button. Likes may provide a certain amount of data, but what does the absence of a user like mean? Did the individual dislike something? Was that individual ambivalent? Or did he simply not see it? In addition, users may oen create a post but refrain from posting it. Was that post an honest reflection of the user's think- ing, or did she change her mind? Or did she simply forget to post it? And since social media may have effects on users that mimic the presence of a camera, each post is tainted with uncertainty as to whether it is genuine or for show. Naturally, some as- sumptions—perhaps even good ones—can be made, but uncertainty is unavoidable. sociaL meDia as petri Dish Governments and businesses aren't the only ones interested in social media. ese organizations tend to mainly ob- serve, looking for information that allows them to achieve their own goals—whatever those might be. But the natural extension is to search for information by imposing conditions on the network and seeing how it responds, also known as an experiment. Many Facebook users were up in arms over an experiment in which the company manipulated the news feeds of nearly 700,000 users to determine how the emotional content of those feeds af- fected user activity. In a sense, the reaction was entirely uncalled for, but not overly surprising: the indignation had no teeth thanks to Facebook's terms of service, which essentially absolve the company from responsibility for anything whatso- ever. And in the end, aer the usual corpo- rate mea culpas, no mass exodus occurred. e draw of the service remained more valuable than the cost in privacy and even dignity, despite the outrage. In addition, stories have swirled for years about government agencies attempt- ing to manipulate social networks for various purposes. Although some of these stories may be overblown, some are likely true, and the threat that government poses is a serious matter. Big Brother Social networks are an irresistible target for government spy agencies, as the Edward Snowden revelations indicated. Beyond the understandable perusal of social media by the IRS in search of overly boastful tax cheats, agencies like the NSA view these services as a convenient data- base that they could never have compiled on their own. Instead of needing to force through a data-collection program, risking the ire of citizens, these agencies needed only wait for users to offer far more data willingly to Facebook. Instead of twisting hundreds of millions of arms, the govern- ment needed only twist Mark Zuckerberg's arm to get access to all that data. e power of the government multi- plies greatly the dangers not only of inac- curacies resulting from social-media data mining, but also from simple overreach. Given the virtually innumerable laws and regulations at the federal, state and local levels, ignorance of the law may be no excuse as far as iron-fisted judges are con- cerned, but it sure is a darn good reason. A lack of protection from government sur- veillance thus opens virtually every citizen to prosecution—oen on felony charges, even if no real harm is done to anyone (see ree Felonies a Day: How the Feds Target the Innocent by Harvey Silverglate or Go Directly to Jail: e Criminalization of Almost Everything by Gene Healy). As in Orwell's nightmare world of 1984, however, the state doesn't pursue each and every crime (it entirely lacks the resources), but access to social-media data gives it a tremendous leg up if someone is targeted for prosecution, whether for real crimes or political ones. e danger even applies to politicians, whose leadership can easily be swayed by improprieties revealed through social networks. costs anD Benefits: anaLyze this What's conceivable isn't always prac- tical. We can easily imagine all the reams of social-media data collected day by day yielding an extraordinarily detailed picture of users—how they might be targeted for advertisements, experiments or worse. But with more data comes more storage and processing requirements to deal with it all. Although data analytics (colloquially termed big data) may be a high on the list of hot IT topics, its future is uncertain. As Moore's Law slows, for instance, will com- pute power come fast enough and cheap enough to provide the level of detailed results that organizations expect from analysis? Furthermore, once they have distilled the ostensible nuggets of valuable information from all this effort, will the value provide a good enough return on the investment? Even though some might find it hard to imagine a world without Facebook, Twitter and all the other social networks, these services haven't been around for all that long. Expectations of tremendous value falling out of number crunching on the mess of social-media data may well be akin to the perennial expectation of flying cars: always near, never here. Clearly, analysis of personal data can paint a detailed and reasonably accurate picture of an individual, but the errors and faux pas that can crop up reduce the value—even if it's just by way of annoying people. e future of data mining in social media is clouded by a number of uncer- tainties. First, for all its apparent value, the business model of social networking is dubious at best. Twitter, for instance, has yet to turn a profit and appears to be facing accelerating losses. Facebook is profitable, but its growth may be stagnating, and the value of its advertising model is coming under greater scrutiny. Second, should user sentiment begin turning back toward privacy as being more valuable than free services, social networking could find itself in a crisis almost immediately. Such an oc- currence seems unlikely, but it is possible. ird, and finally, unless data min- ing proves to be truly valuable—not just overhyped sci-fi—the costs of securing the resources to conduct the analysis will hamper its pursuit. Governments that don't care about debts may continue building data centers to process your photos from the local restaurant in hopes of finding a terrorist hiding in your fettuccini alfredo, but companies that need to make a buck will look elsewhere to drive their business decisions. n