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

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

#21

Post by emodendroket » Fri, Mar 30, 2018, 9:41 PM UTC

CiPHPerCoder wrote: post #8

daenz wrote: post #3

Am I understanding this correctly? Facebook user data (likes/profile info) was scraped to produce low-dimension feature vectors for users (similar to word2vec). These feature vectors were then run through some ML model to predict...what exactly? Targetability for effective political ads?

[ Deleted. Nothing I say on HN ever matters. Move along ]

Very few "undecided" voters truly are; elections are won and lost by getting your supporters to go to the polls. So if you wanted to use scurrilous, fake news to help your candidate, you'd be better off sending stories that will get your supporters really fired up and eager to vote and get their friends to vote, not trying to persuade the practically nonexistent undecided demographic.

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

#22

Post by NoodleIncident » Fri, Mar 30, 2018, 9:42 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…

Shortly after the election, I read something saying that the actual ads were targeted soundbites at specific demographics likely to vote Democrat, run shortly before the election with the intention of suppressing voter turnout.

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

#23

Post by jacquesm » Fri, Mar 30, 2018, 9:42 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…

There doesn't have to be a security breach for it to be a very bad example of using data collected in one way for a completely different purpose. It violates the 'lawful basis for processing' part of privacy legislation.

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

#24

Post by emodendroket » Fri, Mar 30, 2018, 9:42 PM UTC

acdha wrote: post #11

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? :)

There’s a bug confound with the catalog: Netflix’s streaming catalog is much smaller than their DVD selection was and it changes as licensing deals expire. No model can make up for that entirely.

I use the DVD one and I do find the suggestions are mostly pretty reasonable ones.

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

#25

Post by pavlov » Fri, Mar 30, 2018, 9:43 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

The EU should do a unionwide ad campaign pointing out the fact that the much-maligned eurocrats have actually been working for years to fix this very difficult problem that's only now becoming apparent to the wider public. The timing couldn't be better as GDPR goes into effect just after the Facebook/Cambridge scandal.

Unfortunately all EU institutions are terrible at marketing. If they did an ad campaign, it would probably be a TV commercial showing Jean-Claude Juncker giving a speech with subtitles in 15 languages.

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

#26

Post by godelski » Fri, Mar 30, 2018, 9:43 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...

Youtube does better than Netflix for me, but they still suggest stuff that I have already watched. Sometimes stuff I literally watched an hour ago. And I watched one Joe Rogan episode and it will not stop suggesting it to me but fails to notify me on things that I watch every episode on; like 3Blue1Brown, Robert Miles, or Rare Earth.

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

#27

Post by dazc » Fri, Mar 30, 2018, 9:44 PM UTC

danso wrote: post #16

dazc wrote: post #9

Earlier quoted context omitted.

How hard can it be? Any film that I didn't watch all the way to end ought to be a strong signal that I didn't enjoy it. Not that I want to watch other movies just like it.

OTOH, that you even chose to watch a movie is a sign that you are interested in the genre. Just because I abruptly stopped watching “Star Trek IX”, doesn’t mean that I have abandoned the sci-fi genre or even Star Trek

I have a friend who shares my account since I stay with him when I am in the UK. It isn't worth the effort to have two profiles.

So, he likes horror movies and will watch any horror regardless of any signal that it's going to be poor. You can look at the viewing history and see he rarely goes beyond 5 minutes of watching any of them.

I watch Netflix regularly throughout the year, he watches in phases that last a week or two and then nothing at all for months at a time (for reasons that should seem obvious by now).

As a non-horror movie aficionado, I can say with certainty that only 2% of horror movies are ever worth watching and only 50% of these are any good. As a consequence, my personal viewing history includes almost no movies in this genre.

My favoured genre is drama and I normally watch all the way through.

You should be able to guess by now that I should rarely be recommended horror movies but, alas, Netflix thinks otherwise.

Btw. I also rate movies I watch - my friend doesn't.

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

#28

Post by godelski » Fri, Mar 30, 2018, 9:45 PM UTC

danso wrote: post #16

dazc wrote: post #9

Earlier quoted context omitted.

How hard can it be? Any film that I didn't watch all the way to end ought to be a strong signal that I didn't enjoy it. Not that I want to watch other movies just like it.

OTOH, that you even chose to watch a movie is a sign that you are interested in the genre. Just because I abruptly stopped watching “Star Trek IX”, doesn’t mean that I have abandoned the sci-fi genre or even Star Trek

I feel like watching 20 minutes of a movie and then downvoting should be a strong indicator. This does not appear to be the case. In fact, it seems to be more likely to appear as the first suggestion when I do this.

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

#29

Post by jacquesm » Fri, Mar 30, 2018, 9:46 PM UTC

emodendroket wrote: post #20

daenz wrote: post #3

Am I understanding this correctly? Facebook user data (likes/profile info) was scraped to produce low-dimension feature vectors for users (similar to word2vec). These feature vectors were then run through some ML model to predict...what exactly? Targetability for effective political ads?

It seems like the purpose was narrowly tailoring messages, which is something political campaigns are really keen to do now (Obama's campaign was kind of a trailblazer here, right?).

> Obama's campaign was kind of a trailblazer here, right?

It's a pretty big gap between using and abusing social media and as far as I know Obama's campaign did not 'narrowly tailor messages'. They did target broad groups using generic messages and they did quite effectively use social media presence to build support.

But they did not - as far as I know, so please correct me if I'm wrong - go so far as to single out individuals or really small groups with the express intent of flipping their votes or targeting them with disinformation in order to try to stop them from voting.

And Cambridge Analytica seems to have been doing just that if the currently available information is to be believed.

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

#30

Post by lordnacho » Fri, Mar 30, 2018, 9:46 PM UTC

sebleon wrote: post #15

tomrod wrote: post #12

Earlier quoted context omitted.

> 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? Qualitatively: show things that get them angry. Quantitatively: test and control pop splits.

> Quantitatively: test and control pop splits. How do you actually do this? Presidential elections come once every 4 years.

And there's a big question mark over whether lessons learned (ie parameters) from one election are valid for the next.

What if all the sensitivities are dependent on the length of the candidates' hair? It seems the total hair length of the two candidates was a maximum at the last election. Another time you might be sampling more towards the middle.