You might want to think that your online activity on Facebook is somewhat private. If you buy into the kind of Web 2.0 rhetoric that says that by participating on social networking sites you have given up on privacy, then maybe you would say that your online activity is public.
Both of these characterizations are wrong. Your online activity is hyperpublic, it’s entirely novel, and right now you are the only one who can manage it effectively. But that will involve rethinking privacy and the nature of online spaces.
A recent New York Times article outlines a few of the ways that our privacy is being eroded online. The gist of the article is that even when you try to post anonymously, or try to remain silent about certain aspects of your life, say on Facebook or Twitter, you might be “outed” anyway. That’s because clever computer scientists have developed “data mining” techniques that use the tiny tidbits of information in your tweets, or on your friends’ blogs or Facebook pages, to predict things about you. And they’re pretty good at it.
According to the article, researchers were able to predict with 78% accuracy that an individual was gay, even if he never revealed the information himself. In another demonstration researchers were able to ascertain the social security numbers of 8.5% of all Americans born between 1989 and 2003, based on profile information and other “publicly” available information online. We should expect those numbers to improve as the wealth of online information surrounding any given individual increases.
The article concludes with a quote from a prof at Cornell, “When you’re doing stuff online, you should behave as if you’re doing it in public — because increasingly, it is.”
The idea that a person’s online activity is public carries a certain credibility. You post something on Facebook or tweet about something and many people see it. At first it might seem like shouting something out in a room filled with friends, or like standing on a soapbox on the street corner, shouting about the state of the world.
At the same time you might think that your actions, despite being accessible to many people, are essentially private. You have selected your friends, and so have chosen to give that information only to particular individuals. Random computer scientists pointing their algorithms at your profile are violating your privacy, you might say.
Both of these interpretations of privacy are common. The first treats information as essentially private only if you haven’t given it out in a “public” space. And it considers Facebook a public space. It is probably the most common interpretation of privacy. It’s also quite flawed.
Treating Facebook like a public space doesn’t quite match all of our privacy intuitions. Helen Nissenbaum has written about the contextual nature of privacy, which seems a lot more like the privacy I hear people describe in their day-to-day situations.
Contextual accounts of privacy recognize the importance of context on determining whether information is private or public. For example, you can consider a piece of information private if you tell it only to a particular group of friends. Even if you say it in a public space. It is the social context, determined in part by the social norms surrounding the activity, in this case the group of friends who were the intended recipients of the information, that determine the privacy status of the information. Similarly, the fact that you frequent a gay bar in a particular part of town would be private with respect to your heterosexual colleagues who don’t know your sexual orientation (you might have reasons that you don’t want them to know it, and have kept it from them intentionally). If one of them happens to see you out at the bar and tells all of your other colleagues about it, he would have violated your privacy, despite having seen you out in “public”. He is not a part of the social context in which your sexual orientation is known, and therefore ought to recognize that and respect it (there’s the normative aspect of contextual accounts of privacy).
The contextual account therefore captures how a data mining algorithm can violate your privacy–it is not a part of the social context (Facebook friends) that are meant (according to ordinary social norms) to have access to your posts. I have written about this kind of privacy violation. Data mining is especially problematic because, as the NYT article points out, it is intended to discover facts about you that you never divulged online in the first place, and is intended to intentionally violate social contexts that determine the privacy status of that information. Worst of all it succeeds, as we see in the case of discovering sexual orientation and social security numbers!!
Let’s forget for a moment about the normative aspects of online privacy, that is, let’s forget about the fact that computer scientists ought not to be mining our Facebook posts and trying to predict new tidbits of information about us. If we try to understand, like the NYT article does, what the effect of predictive data mining is on online privacy we cannot reasonably conclude that it makes online spaces more like public spaces.
There are many aspects of real and online spaces that distinguish them from one another.
If I yell something out in a crowded public space, say a street corner, I can expect certain things to be true in an offline context. I can expect that few people will be listening. I can expect that many of those who do listen will consider my words uninteresting. I can expect that they will not know or care who I am. I can also expect that the ones who listen carefully, with interest, will only remember certain of my phrases, but not the whole thing. And even most of those individuals, if not all of them, will have forgotten most of what I have said in the not-too-distant future.
Online, especially an online environment populated by data mining “bots”, all of those expectations would be false. I should expect that no one needs to be listening in order for the information to be “heard” (it can be accessed later as people browse through my posts). I should expect that at any given point in time someone (or something like a bot) would be interested in my information and might access it. I should expect that at the very least they will be able to figure out my name and other information and will want to do so, and will succeed in doing so. I should expect that every character of text is analyzed against every other character of text in my and my friends’ posts. Finally, I should expect that the data will never be forgotten, because currently it can’t be. I should expect all of this, because that is precisely what predictive data mining is designed to do.
Currently the online world is becoming increasingly like this, as the NYT article points out, but it is not really more like a public space. It’s hyperpublic. It is a space where memory is a concept without meaning. It is a space where social contexts are real-world constructions reduced to online fictions by the presence of predictive data mining algorithms. It is a space where I need not divulge facts about myself in order for them to be known.
It need not be this way, and ought not to be. Unless we start to understand exactly how public spaces differ from hyperpublic ones, we can expect to see more and more of the hyperpublic, and less of the public in our lives.
We would be wise to start thinking about how best to behave ourselves in a hyperpublic space, since it seems clear that though we may have misunderstood privacy (e.g. the rhetoric referred to above), we have not yet given up on it.