> I should have a way of asking them to delete all data they have on me including meta data from sms/phone calls they might have collected.

With GDPR they will have a legal requirement that "delete your account" deletes all of your data if you're in the EU. Currently they don't delete conversations with other people, and don't delete any ML data which was based on your data, but hopefully that will be ruled illegal under the GDPR requirements.

I think the problem is that most of the data they have on me is implicit in the data they have stored about other people. They definitely have my phone number, but I'm not sure they store the result of a computation they could easily do on demand with what effectively would be a join statement (no ML required), namely reconstructing my call history with everyone that has the facebook app installed. Again I'm not sure if the GDPR has any provisions against that, because it affects two parties one that has consented to the storage and another that hasn't. Also it is not clear to me how an average citizen would be aware of the fact that even if they only have your phone number stored they are most likely able to reconstruct most of your call history as well.

> namely reconstructing my call history with everyone that has the facebook app installed

IANAL, but I think that would be okay, as long as they don't associate any/strip all personal identifiable information from your phone number, and don't store the phone number in plaintext.

Crazy thought. When a company with a ton of accumulated assets (data) craters, do the rules of bankruptcy and/or mitigating their losses to investors force them to just sell these assets outright?

This may not happen in the next few decades with the likes of Facebook, but imagine what kind of data that much smaller companies have that will just be sold off to the highest bidder, possibly de-anonymized to increase it's value. Does a user privacy contract hold any weight when a company is about to dissolve?

That's impossible with the GDPR for privacy relevant data. That data cannot legally be an asset of a company because a company cannot own it. A company may be a data processor or data controller, but not a data owner for other peoples private data. They may choose to ask each individual if they agree to let another company use the data, but that's a different thing.

> Does a user privacy contract hold any weight when a company is about to dissolve?

It would hold about the same weight as it does now: Effectively ignored.

My knowledge in this area is limited, so maybe others can pipe in, but assuming massive amounts of data on millions (or most?) Facebook users is out in the wild, which cannot be 'put back':

1. how well can this be actually used currently to create a proper psychosocial(?) profile of an individual? What's the current state of the art? Any suggestions for books or papers to read on the matter?

2. do we have any idea of how much our ability to do 1 will improve over time, given the same data? I often find it difficult to separate the Derren Brown style 'profiling' bullshit from what's actually possible. And I also keep reading that 'big data' is not quite as easy to use as it's often made out to.

3. how valuable does this 'stale' data remain? Personality is by definition supposed to remain stable, but I recall learning that even things like friendships, political/world views, other non-personality characteristics are relatively unchanging for most people after somewhere (early?) in their twenties. This could mean that even if Facebook does privacy 'properly', or somehow disappears, there's a rich dataset of individuals that, even if only in the future, be shockingly powerful to manipulate. And we're already scarily susceptible to manipulation using 'conventional' means.

I use Whatsapp only for communication with office colleagues. I had my Instagram linked to my mobile number(2fA). Soon my office colleagues' profiles started showing up on my Instagram follow suggestions. This happened despite the fact that I have not uploaded my contact list to Instagram. I had to block all my colleagues on insta, and then remove my number from Instagram profile.

"how well can this be actually used currently to create a proper psychosocial(?) profile of an individual? What's the current state of the art? Any suggestions for books or papers to read on the matter?"

I suspect quite poorly.

It's 2018 and I've been hearing pitches and narratives to do with fancy targeting of advertisements and user profiling now for 22 years.

And what do we have ? If I search for something obscure related to an (product), I will see unbelievably blunt and almost comically generic (product) ads for a week. Even if my search makes it obvious that I am not a prospective customer.

What advertisers receive for their money on these platforms is laughable and I am sure there's a complicated sales pitch with graphs and fancy terms of art and mentions of "AI" ... and it's all just bullshit, just like it has always been.

I had one incident which convinced me that their algorithm is easily fooled even by single misclick.

I hate ads propagating alcohol so I always report them to FB. Once I misclicked and instead of reporting it I opened linked article. Based on that click FB added alcohol to my interests. It did not matter that I reported all other similar adds before and also after that.

Whether the algorithms seem effective to you, as a single consumer of the advertisements, is irrelevant.

Facebook ad campaigns most definitely work (though I'd imagine there's some variance of effectiveness based on what exactly you're trying to sell with the ad campaigns), and lookalike audiences are an effective way of determining who to show your ads to.

It's really quite simple, as a company you can spend money on Facebook ads and you will get traffic/conversions/etc.

I agree. But I reacted on comments disqusing if this algorithm is capable of doing more than targeting ads to make them cost effective.

That's part of the reason I asked. Based on the ridiculously bad suggestions I'd often get (from search, netflix and the lik, etc.), it all seemed a bit too much hot air.

But a few things made me (sort of) change my mind over time, and worry more actively:

1. the ability to properly use the data available is a skill in itself. The fact that this 'new, non-private' internet is still relatively young (really just a decade, post-smartphone I'd say) could mean that most companies/governments simply didn't know yet how to properly use it. Considering the many examples of the slowness with which organizations tend to adapt, that doesn't seem unlikely.

2. There's a huge difference between the data available, and the targeting achievable pre-smartphone, and post-smartphone. On top of that, I distinctly remember that throughout most of my youth (so up until a few years before world-wide Facebook) it was still very much a 'default' attitude to be very careful about your personal information online. Not quite the "don't go into chatrooms" of my childhood, but still the "don't use your real name, ideally, on <national social network>".

3. We've made quite a number of advances already in storing and analyzing 'big data', from what I understand. And a decade or two of research on this new world of personal internet data must have led to some significant advances, no? I mean, so far we've gotten really good at manipulating people with whatever new technology entered the scene. Just not right away, necessarily.

4. from personal experience, and conversations with people who do/did targeted advertising on Facebook, it really does seem shockingly easy to target specific messages to specific groups. I find it hard to believe that this does not fundamentally change the whole game of manipulation, compared to things like TV and newspapers. Hell, 'old fashioned' direct marketing / door to door sales is perhaps a precursor of all this. I remember reading quite a while back how advanced the techniques were to decide what neighborhoods to visit or snailmail-spam for particular products or services. If companies were willing to spend the amount of money necessary for that, it must have some effect. Being able to target essentially large parts of the entire world in that same way has got to be at least somewhat more powerful, which is scary enough maybe.

I'm not sure of any of this, though. Just thinking out loud and very curious to hear what others think (although ideally I'd also love some suggestions of where to find the best research/articles on the matter).

regarding 1, I recall reading that 23 likes are enough to guess a persons sexual orientation with better accuracy than close colleagues can. Couldn't find a source on this in 5 minutes though.