How Cambridge Analytica’s Facebook targeting model really worked • Hacker News Board

Re: How Cambridge Analytica’s Facebook targeting model really worked

#61

Post by autokad » Fri, Mar 30, 2018, 10:35 PM UTC

etiam wrote: post #41

Spoiler warning. Article punchline ahead. "The whole point of a dimension reduction model is to mathematically represent the data in simpler form. It’s as if Cambridge Analytica took a very high-resolution photograph, resized it to be smaller, and then deleted the original. The photo still exists — and as long as Cambridge Analytica’s models exist, the data effectively does too." That's an eloquent piece of explanati…

> "will society grant me rights to have derivative forms removed or adjusted too? "

I am in favor of no. Imagine I build a gender classification model off public tweets, and then you later delete your twitter account and demand my model not be used because it was trained off 'your data'.

I am in the camp that so long as the data isn't traceable back to you specifically, then don't put any information out there you are not OK with sticking around.

Re: How Cambridge Analytica’s Facebook targeting model really worked

#62

Post by danso » Fri, Mar 30, 2018, 10:37 PM UTC

godelski wrote: post #49

danso wrote: post #46

Earlier quoted context omitted.

I’m confused — you honestly expect Netflix’s model to have figured out that your profile is actually 2 people based on what you believe to be regular and obvious cyclical patterns? I would have to imagine that this is a relative edge case for Netflix, and there is no obvious answer for what to do with someone who mostly likes drama but for some reason, goes on a periodic horror binge. I assume that Netflix’s model ha…

> as a test profile to see what happens when a user only seems to like the “Human Centipede” trilogy. And? You can't just leave us hanging on that.

After creating the account and immediately giving a thumbs up to the trilogy, I've only occasionally logged in and feigned "interest" by clicking on the movies as if I'm about to watch them, or that I enjoy re-reading the synopses. The recommendations are all normal and not noticeably feces-related. But maybe I haven't yet met the threshold for the model to consider me a particularly engaged (or real) user.

Re: How Cambridge Analytica’s Facebook targeting model really worked

#63

Post by factsaresacred » Fri, Mar 30, 2018, 10:39 PM UTC

nicoburns wrote: post #7

rjurney wrote: post #5

Earlier quoted context omitted.

It's a hard problem but Netflix's model represents the state of the art in machine learning for recommendations. Still scared of the singularity? :)

Really? Because in terms of actual usefulness, I find youtube's suggestions to better...

Spotify is king IMO. Their 'discover weekly' playlist turns up gems every time.

But that's because they dive into other user playlists that contain the same music you play. Pretty simple. I assume Youtube do somthing similar.

If Netflix had playlists or a 'want to watch' feature I bet their recommendations would improve.

Re: How Cambridge Analytica’s Facebook targeting model really worked

#64

Post by sg0 » Fri, Mar 30, 2018, 10:41 PM UTC

lordnacho wrote: post #10

Finally, I was waiting for someone to talk about the model itself. It makes sense that SVD or something like it (PCA, co-occurrence, etc) would be used. But I also wonder what exactly you are going to do with the predictions. What exactly do you show to someone to make them more likely to go and vote if they are inclined to vote your way, or make them stay at home otherwise? Is there evidence that whatever you're sho…

Before Netflix stopped showing the number of stars next to content, I used to sort of depend on it for choosing a movie. In fact, I sort of miss it now, and spend more time sifting through content undecided. That's because I am clueless about movies. I believe that there are people who are as unsure about electoral candidates (as in, at a given day, they don't favor one candidate above others) as I am about choosing movies. When push comes to shove (my wife's irritation quotient above threshold) in terms of making a decision, an advertisement that someone saw couple of days back can definitely assist in making a choice at the split second.

Re: How Cambridge Analytica’s Facebook targeting model really worked

#65

Post by dpwm » Fri, Mar 30, 2018, 10:42 PM UTC

daodedickinson wrote: post #57

lordnacho wrote: post #10

Finally, I was waiting for someone to talk about the model itself. It makes sense that SVD or something like it (PCA, co-occurrence, etc) would be used. But I also wonder what exactly you are going to do with the predictions. What exactly do you show to someone to make them more likely to go and vote if they are inclined to vote your way, or make them stay at home otherwise? Is there evidence that whatever you're sho…

Door to door canvassers these days carry devices that tell you what topics to bring up and what topics not to bring up at a certain address, even distinguishing between individuals at an address; some are told to demand a husband let them talk to the wife, for example.

I don't know about the specific campaigns that you are referring to, but in my experience a lot of the information used in campaigns I've been involved in comes from previous canvassing sessions. Political parties in most countries are involved at many levels where there are elections. Canvassing doesn't just take place for the big elections.

One year they will have been round and had a lengthy discussion with Mrs X, but Mr X slammed the door in their face another time. This was somewhat lower tech: the information was printed out and attached to a clipboard.

Most of the time this information is correct. It's more interesting when it's really incorrect. That said, some of the best sessions I've been involved in were where there was no information.

Re: How Cambridge Analytica’s Facebook targeting model really worked

#66

Post by frozenlettuce » Fri, Mar 30, 2018, 10:45 PM UTC

emodendroket wrote: post #19

I am really puzzled by the Cambridge Analytica scandal. It's not particularly savory, but is there something happening here that it wasn't basically already known about how Facebook worked? By the protests of their own executive, the system was working as designed, and at worst Cambridge Analytica misled them about how they intended to use the data, right? There was no actual security breach here, as far as I can und…

It became a "problem" because it helped Trump win.

Re: How Cambridge Analytica’s Facebook targeting model really worked

#67

Post by salty_biscuits » Fri, Mar 30, 2018, 10:45 PM UTC

etiam wrote: post #41

Spoiler warning. Article punchline ahead. "The whole point of a dimension reduction model is to mathematically represent the data in simpler form. It’s as if Cambridge Analytica took a very high-resolution photograph, resized it to be smaller, and then deleted the original. The photo still exists — and as long as Cambridge Analytica’s models exist, the data effectively does too." That's an eloquent piece of explanati…

I guess the question is what are you trying to protect? The model is fundamentally lossy as it is a rank reduction method so your original data is gone (i.e. no one would be able to accuse you of liking a particular controversial post, just that you are likely to like that post). So it sort of has the differential privacy thing going on. I guess it is another question as to if such models should be built at all. I think the fidelity of the models will answer that in time, if they work really well it is scary, if they are poor models they will cease to be used. I suspect that it will be in the middle and highly sensitive to the quality of the original data and the quality of the implementation like all ml applications.

Re: How Cambridge Analytica’s Facebook targeting model really worked

#68

Post by darawk » Fri, Mar 30, 2018, 10:46 PM UTC

etiam wrote: post #41

Spoiler warning. Article punchline ahead. "The whole point of a dimension reduction model is to mathematically represent the data in simpler form. It’s as if Cambridge Analytica took a very high-resolution photograph, resized it to be smaller, and then deleted the original. The photo still exists — and as long as Cambridge Analytica’s models exist, the data effectively does too." That's an eloquent piece of explanati…

> when strictly speaking the raw data has indeed been deleted after being used to create a derivative work that can for all important purposes be used to recreate the original?

To be precise, you almost certainly cannot use this data to recreate anything remotely resembling the original dataset. This type of dimensionality reduction would throw away enormous volumes of data. There is no meaningful sense in which you can reconstruct the data from it.

What they have done is distill some insights about people from this data. It's arguable whether they should be allowed to keep those insights, but there's no privacy risk there really.

It's honestly kind of disingenuous to describe dimensionality reduction in the way that they do here. It is like reducing the resolution of a photo, but it'd best be described as reducing that resolution to say, the 20 most representative pixels. There's no real sense in which the photo still exists.

Re: How Cambridge Analytica’s Facebook targeting model really worked

#69

Post by pbalau » Fri, Mar 30, 2018, 10:50 PM UTC

etiam wrote: post #41

Spoiler warning. Article punchline ahead. "The whole point of a dimension reduction model is to mathematically represent the data in simpler form. It’s as if Cambridge Analytica took a very high-resolution photograph, resized it to be smaller, and then deleted the original. The photo still exists — and as long as Cambridge Analytica’s models exist, the data effectively does too." That's an eloquent piece of explanati…

If I tell you I like furry porn, is there a way for me to make sure you forget that? This has lots of implications, many of them placing the "blame" on me for telling you this.

Re: How Cambridge Analytica’s Facebook targeting model really worked

#70

Post by rossdavidh » Fri, Mar 30, 2018, 11:18 PM UTC

scottybowl wrote: post #2

I'm excited for GDPR.. The hard part is going to be getting the truth out of these companies about the actual extent of the data they hold on us

That "hard part" is going to be more than just hard; I think the word you're looking for there is "impossible". They don't have any way of knowing what data Google, Facebook, Amazon, or any other company has. As this article does a credible job of explaining, you have to understand a fair amount about statistics (PCA) and machine learning to even know if if you were looking at it, and they won't know where to look for it. They have no enforcement mechanism in mind, and they passed a law anyway, which effectively means "you can't admit to having this", which will mean that the more willing a company is to lie, the bigger an advantage they will have over their competitors.