Fear of Observation

The unsettling nature of Newcomb's paradox indicates that suspicion about observation isn't just for paranoid schizophrenics. This game is paradox because of the perfect relationship between past observation and present prediction, a situation which (we're pretty sure) can't occur in the real world.

But observation can move closer and closer to successful prediction. If this seems to contradict my earlier post about irreducible complexity, it's because I'm now requiring only predictions based on generalizations from patterns in data, not explanations for the phenomena in question. This modification lowers the epistemic threshold low enough for contemporary computers to cross it--and then succeed spectacularly.

Datamining allows computers to predict, better and better, future outcomes within the field described by the data. It internally solves the problem of complexity by evaluating only digitized data points triggered by more complex behavior, then predicting which data will be triggered in the future.

Insofar as "important" aspects of people's lives can be translated into these data points, the results are deeply disturbing. Snapshots of social interactions can predict, with shocking accuracy, the future of relationships (student-teacher, dating couple, etc.) for months and years to come. Reducing behavior and faces to easily coded elements of a dataset works.

This predictive capacity brings the problems with free will uncomfortably close to home. Computers know better than you do what you are going to do in the domains they understand. And they understand almost every domain that you consider important.

Predicting human behavior deductively from systems logic failed. There is no perfect economic, social, or political human and its probably impossible to find anyone who represents a perfect synthesis of the three. But predicting human behavior inductively works all too well, suggesting that the regularities are out there, even if we lack the tools to explain them.

Humans are helpless to fend off incursions into traditionally private spheres. AT&T knows about your telephone conversations and the calling networks in which you play a small part. Facebook knows who your real friends are, who is really stalking whom, and when relationships are going to develop. Amazon, Google, and countless others know what you're about to buy.

A seemingly unique set of preferences that fit into categories encompassed by datamining will no longer feel authentic or interesting. In some ways, the fight for subjectivity against capabilities of mass prediction must involve (subjectively) meaningful activity that cannot be encompassed in a data set or made analogous to analogies by others. This activity can occur in groups, but these groups must shatter mechanical and corporate categories for prediction. As individuals succumb more and more to statistical control, the important thing is the act which makes true difference.

The capacity of datamining to predict the behavior of individuals reveals more than anything else the insight of Deleuze's take on eternal return. Those differences which can be subordinated to the taxonomy of a databank fail the test of true difference, for they can be subsumed, accounted for, explained, and defused. Only a reconfiguration of the past that operates autonomously beyond mundane configurations of history and personhood can escape the neutralizing power of prediction.

Thus we should be unafraid of observation and unprotective of the private--as long as we are willing to act in ways that sever the inevitably link between observation and prediction. In a capitalist world growing ever more efficient, prediction and control go hand in hand. As predictions become better and better, we must ask ourselves which aspects of life we are willing to cede to control... and which we will make autonomous.