Minority Report for 2014

22 June 2013

Minority Report for 2014

A sends:

Minority Report for 2014

The most insidious aspect of general surveillance is not illegal privacy invasion. It is even not the full personal history that enables indefinite retroactive incrimination.

The biggest danger lies from exclusive access to large data sets unavailable to others. Mining this data may provide novel actionable inferences, existence of which is out of grasp of the general public.

Just as hypothetical example, there may be a correlation between higher ratio of electronic contacts with close friends compared to other contacts, and number of people rallying on streets week later. Another artificial example would be stronger than random correlation between using certain words together in messages and anti-government sentiment developing several months later. Inferences like these would give government peek into possible futures and opportunity to proactively intervene before anything happens. It would be very hard for the general public to match such precog machinery.

This sounds like sci-fi, but data mining in newly available data sets does more often than not result in major discoveries. The first part of the discovery is observed causality. The actual mechanism may be discovered later or never, but that doesn't stop you from using the learned causality to your advantage. Insurance companies make real money from actuarial tables.

I do not know of a way to conclusively prove this by external observation. It's like proving someone is cheating on cards just by their winning pattern. I don't think that whistleblowers like Snowden are involved with such data mining, or even aware of its existence. But I cannot find justification for enormous spending on surveillance in any of the currently popular fears.

The real question is, then, should the government be trusted with exclusive access to this data. Or should it give us the API, so that people can also mine it, and help us make sense of ourselves. Frankly, with de-identified and provably true API, I'd be almost OK with snooping and retroactive crime capability. There is too few of them to screw us all. But I do want to know if what government is doing now is related to countering predictable sea change, and what that sea change might be.

What would be effective countermeasures (by effective I mean something better than requesting government to stop doing it) ?

One, as mentioned, would be making these data sets available to the general public. It's our data anyway. Imagine if physician wouldn't tell you your blood pressure, but would buy life insurance policy on you.

Randomizing personal behavior may help to break correlations we are even not aware of. How to do it on any significant scale is beyond me (making tossing coin for daily decisions a popular lifestyle.)