Tuesday, July 16, 2019

Facebook, why are you showing me adverts for [!!!]

So, I started getting adverts for [!!!]*** in my feed recently, and my first thought after I haven't been searching for that, was I haven't been talking about that. That pub conversation about the creepy ads that pop up suggesting something discussed earlier that day, like an over-eager personal assistant trying to anticipate needs you never thought you had, that happens nowadays. And I don't think it's just internet pareidolia, seeing order and intent in the random straw-scatter of commercial trolling. Although I'm still not sure why the internet is bringing me [!!!].

So, I'm fairly loose with my permissions on my phone - I'm a Google Guide, I have a timeline, I run phone games with sweeping permissions, linked up to my Facebook. Bixby's only partially set up (my assistant app) but I use google to identify music quite a lot, so it's used to listening. I should be a nice visible smear of data to the main data-brokers, with a location, interests, soundtrack and oh, waaay more. But could it really be listening?

Sure, I think so. I'm not even sure it would be hard. I'm while I'm not an expert in this area by any means here is my hypothesis.

All listening apps (and Facebook is one, because you can use it for phonecalls) keep a small amount of audio in buffer at all times (and Google is one, so you can run audio searches) so it can check for activation phrases (and Apple is one so it can run Siri) and become better at understanding your voice (and honestly, I could go on here for a while). Google knows where these files live. Possibly, lots of programmes know where these files live. If your phone and operating system are aligned (oh, and they have to be, so that audio can be improved on phone calls with poor connections) then those files can be accessed, analysed and used, perhaps to improve speech recognition technology (Google are open about doing this) and perhaps to give you a pair of jeans that might match the conversation you had about them earlier.

You'll notice I'm not saying this in a tone of any great panic. Part of the nature of buffered information is that it is both disposable and rapidly disposed, and audio is data-heavy. The vast bulk of this audio information will be dumped, though (as various experiences with web-cam data breaches have taught us), there are probably some bits and bobs retained as baseline, sample and reference, blah blah blah. That little ripple of activity around the data food source of the noise you're making is a weak and evanescent signal, not strong enough to do more than tweak an ad or improve a location report.

Who's listening? Mainly algorithms and analysis software, though good old Amazon puts samples to human ear. And what's it linked up to? If you visit that article (again - you probably read it once already) you'll note that it came from a whistle blower worried that they had neither reporting mechanism nor capability for troubling audio content. They couldn't identify the users as data was stripped of identification data on the way to the human listener, so no mechanisms for reporting data according to concerns about the individual was included in the business process. Protecting privacy and avoiding responsibility often do go hand in hand; and I suspect that you may insert qualifiers into that sentence. But it isn't always about what's possible. It's also about what's practicable.

So: algorithms filtering my audio wormcast for actionable data that my be used to fine-tune my advertising offer before said audio is dispersed by the tides of automated data cleansing? Practicable.  My phone listening in on my every word and alerting a third party when I make produce concerning content? Possible, and sadly I do know that there are apps for that, which are often used in the context of abuse, grooming and domestic violence. But doing that for everyone? Neither practicable (we don't have the resources for follow-up) nor possible (the processing power required would not be available) nor profitable (signal to noise ratio all wrong).

Tech giants listening to everything I say and using it to build a profile of me that could be used to deepfake my identity? In my dreams. No, seriously, I dream of having that much processing power dedicated to replicating my identity. It's a sort of tech immortality. Or not, I guess. Would you care?

To return briefly to [!!!], ****I should clarify that [!!!] in this case does not refer to the band. It's just a thing, you know. But not a thing I'd discuss in a public context. Never mind here's some music