Surveillance and data security

2018-02-12 · ~517 words

It seems near-inevitable that continued engineering progress will cause more and more personal data to be collected by computers, given the various social and economic benefits of that. Trying to stop adoption of these technologies is, I think, likely to fail, as almost everyone will trade a certain benefit now for a distributed, hazy risk in the distant future. And even if a particular piece of data isn't collected, AI-powered analytics will increasingly be able to infer it from other pieces, as everything interacts with everything else. Even if I don't take my phone to the grocery store, for example, I can still be picked up by the store cameras, by payment networks when I check out, by the phones of other people there, etc. (I've already had friends surprised at how much of their lives could be reconstructed from apparently random-looking Facebook and Instagram posts.)

Hence, I think it's necessary to reverse the usual privacy question, and ask: assuming a world where everything, every moment of every life, is stored in a computer somewhere, what could stop that from being abused? Right now, any number of people "could" kill me, in the sense that they have the physical ability to shoot me with a gun. But they don't, because it would be pointless - stopping shootings is an extremely high priority in Western societies, so this person probably wouldn't gain very much, and they'd soon be shot themselves and/or have their life made miserable. Historically, there hasn't been anything like this for data (except some government and military secrets), as it hasn't been important. But that will ultimately have to change. In the limit, with brain-computer interfaces and uploading, there's no real distinction between "a person" and "a person's data" - someone who knew everything about you would de facto have their own copy of you.

For example, some months back, a lawyer friend of mine wrote about how the usual "punishment" for American violations of the Fourth Amendment was to have the information excluded in a criminal trial against you:

https://www.facebook.com/alyssamvance/posts/10213113878151275

Obviously, trying to "punish" the police by making them let a guilty person go free is... perhaps not the best way to create an incentive. With much higher stakes, there needs to be something like a reverse-Prisoner's-Dilemma in place: a situation where even when a group can do something bad collectively, no individual is incentivized to do so, just as the American military "could" stage a coup but no individual member would gain from plotting a coup individually.


That's an interesting angle, but what about the scenario where something today is analogous to eg. being gay in the 1950s, when it was illegal (and the majority of people endorsed that)? In that case, there wouldn't be any public outrage at eg. the police using their power to arrest a gay person, since most people would approve of that. To be consistent across time periods, it seems like we'd have to have rules that punish defectors for privacy violations even when the result is "good" ("good" according to what people believe at that time).