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

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

#101

Post by jpalomaki » Sat, Mar 31, 2018, 1:13 AM 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…

It is reasonable to assume that a marketing message written with the profile of the targeted person in mind works better than generic message.

In Facebook campaigns you can use certain things, such as user's interest, to select who sees your message.

I'm not an expert on Facebook analytics, but I believe you can get pretty good stats on how your campaigns are working, how much promoted posts get shared etc.

This sounds like the holy grail for advertising. You get to write your message for certain profile and get quick feedback how it worked. Even if the system is not perfect, you would have an advantage compared to somebody else who is spending the same amount of money and not using similar targeting.

Maybe their model also allowed them to find social influencers with many followers. Being able to targer these people and get them to share your message would be really good.

The article compares this to the effectiveness of traditional voter targeting methods. I'm not sure what the parameters used on those are, but maybe all of them are not available on FB, justifying the need for something else.

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

#102

Post by neuronexmachina » Sat, Mar 31, 2018, 1:16 AM UTC

vuln wrote: post #77

frozenlettuce wrote: post #66

Earlier quoted context omitted.

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

This is exactly it. At least it stops the news from droning on and on about Russia. I thought Clinton spent large amounts of money on data and the Democrats admitted the data was bad or at least that was their excuse. How much did CA pay for this data? I still find it crazy that Trump campaign spent 30% of what Hillary did and still won. The Russians used 100k$ worth of ads to sway the election. This stuff doesn't t…

Do you have any sources for your claim about how the Clinton campaign acquired FB data and how they used it? Was any of it acquired fraudulently and/or in violation of FB's ToS, like CA's data was?

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

#103

Post by pps43 » Sat, Mar 31, 2018, 1:17 AM UTC

rjurney wrote: post #5

godelski wrote: post #4

> has revealed that his method worked much like the one Netflix uses to recommend movies. I'm not sure this is the model you want to emulate. The suggestions are terrible and continually getting worse.

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? :)

Yes, it's a hard problem, but no, Netflix recommendations are nowhere near state of the art. They were at the time of Netflix prize, but since then the field has advanced a lot, while Netflix only dumbed down their recommendation models.

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

#104

Post by wybiral » Sat, Mar 31, 2018, 1:58 AM UTC

darawk wrote: post #90

wybiral wrote: post #89

Earlier quoted context omitted.

Unless the data is completely random it's not crazy to say that the data can be reconstructed from a reduced version. If you have a million points that largely fall on a 3-dimensional line and you project that into 2 dimensions, you can easily recover that lost dimension with losses relative to the deviation. And that loss may not even matter depending on the kinds of data and margins of error you're working in.

Except that that toy example bears no resemblance to the actual situation.

How many dimensions were they working with and how much variance and correlation was there in the features? What's the margin of error for the end product?

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

#105

Post by mistermann » Sat, Mar 31, 2018, 2:02 AM UTC

vuln wrote: post #77

frozenlettuce wrote: post #66

Earlier quoted context omitted.

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

This is exactly it. At least it stops the news from droning on and on about Russia. I thought Clinton spent large amounts of money on data and the Democrats admitted the data was bad or at least that was their excuse. How much did CA pay for this data? I still find it crazy that Trump campaign spent 30% of what Hillary did and still won. The Russians used 100k$ worth of ads to sway the election. This stuff doesn't t…

It doesn't have to add up, most people are too busy with their real lives to manually search and find reliable details (what we get from the media is not reliably unbiased or true), and then read and understand them, so they believe what they see and hear repeated over and over on the TV, radio, and newspaper: the American President is controlled by Vladimir Putin. Even most smart people don't seem to care about actual evidence.

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

#106

Post by alexcnwy » Sat, Mar 31, 2018, 2:02 AM UTC

stochastic_monk wrote: post #84

darawk wrote: post #68

Earlier quoted context omitted.

> 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…

That's only accurate in the sense that because an LSTM's hidden layer is much smaller in dimension than the data on which it is trained, there is less information in it. However, it concisely represents a manifold in a much larger dimensional space and effectively captures most of the information in it. It may be (and is) lossy, but don't underestimate the expressive power of a deep neural network.

You're throwing out buzzwords instead of addressing the response.

It's dimensionality reduction. You cannot recover the original object. It's like using a shadow to reconstruct the face of the person casting the shadow.

Note this has nothing to do with the expressive power of a deep neural network. You are by definition trying to throw away noisy aspects of the data and generalize a lower dimensional manifold from a high dimensional space. If it's not lossy, it won't generalize.

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

#107

Post by hueving » Sat, Mar 31, 2018, 2:02 AM UTC

jacquesm wrote: post #23

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.

Which US legislation?

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

#108

Post by mistermann » Sat, Mar 31, 2018, 2:04 AM UTC

neuronexmachina wrote: post #102

vuln wrote: post #77

Earlier quoted context omitted.

This is exactly it. At least it stops the news from droning on and on about Russia. I thought Clinton spent large amounts of money on data and the Democrats admitted the data was bad or at least that was their excuse. How much did CA pay for this data? I still find it crazy that Trump campaign spent 30% of what Hillary did and still won. The Russians used 100k$ worth of ads to sway the election. This stuff doesn't t…

Do you have any sources for your claim about how the Clinton campaign acquired FB data and how they used it? Was any of it acquired fraudulently and/or in violation of FB's ToS, like CA's data was?

Sigh....the technical legality of obtaining the data is not the point of contention. Do you think that is what this is about, whether CA "broke the law"?

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

#109

Post by hueving » Sat, Mar 31, 2018, 2:04 AM UTC

neuronexmachina wrote: post #102

vuln wrote: post #77

Earlier quoted context omitted.

This is exactly it. At least it stops the news from droning on and on about Russia. I thought Clinton spent large amounts of money on data and the Democrats admitted the data was bad or at least that was their excuse. How much did CA pay for this data? I still find it crazy that Trump campaign spent 30% of what Hillary did and still won. The Russians used 100k$ worth of ads to sway the election. This stuff doesn't t…

Do you have any sources for your claim about how the Clinton campaign acquired FB data and how they used it? Was any of it acquired fraudulently and/or in violation of FB's ToS, like CA's data was?

Do you have any sources for your claim that the parent poster claimed the democrats purchased Facebook data?

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

#110

Post by cirgue » Sat, Mar 31, 2018, 2:04 AM UTC

rjurney wrote: post #5

godelski wrote: post #4

> has revealed that his method worked much like the one Netflix uses to recommend movies. I'm not sure this is the model you want to emulate. The suggestions are terrible and continually getting worse.

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 problem with all ML-based content recommendation, whether it's advertising, movies, whatever: the people who make the content have strong opinions about what should get shown to whom and when. If those people have negotiating leverage over your organization, your business team will compromise your recommendations to placate those people. That means that after you've build this beautiful model that minimizes whatever cost function you've chosen, there will be a bevy of business rules and content-specific score adjustments that will be overlaid. Over time, this shaggy, bad-assumption-laden system will dominate the user experience, and unless your management has a vested interest in maintaining the integrity of the recommendation system, your model will eventually be lost in a sea of human-generated noise. Not that I'm bitter about this or anything.