Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

How Cambridge Analytica’s Facebook targeting model really worked

niemanlab.org

Post reply on HN
Print view View original

51–60 of 210 posts

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

#51

Post by wu-ikkyu »

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…

>I don't know how effective it is, I'd like to learn more.

Here is an interesting Ted Talk which discusses an FB experiment that details how effective minor UI changes can be on voter turnout (13:40)

https://www.ted.com/talks/zeynep_tufekci_we_re_building_a_dy...

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

#52

Post by godelski »

codezero wrote: post #42
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.

Are you saying all of Netflix suggestions are getting worse? How are you measuring that? Or are you saying your netflix suggestions are getting worse? Mine are pretty good and have held steady for a while, at least in terms of my own preferences, though I also am probably using it a little less as I've got Prime and Hulu now as well, so there's probably less times I'm randomly searching through Netflix and finding no…

I can only tell you what my personal experience is. I laid it out more in another comment under rjurney's reply.

I'll also add that I am more frequently searching for 15 minutes then switching to another service. I used to find a movie to watch in 5 minutes. I am a big movie person too, and will watch most things. But I am also more aware that if I watch one show that is just "meh", then I am going to be bombarded with shows of similar quality for the next few weeks.

Also, there is a fairly obvious pattern that shows up from the movies in "my list". Those do not seem to be weighted more heavily.

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

#53

Post by joncp »

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 sure seems that way. It's almost as if Netflix is giving in to pressure from content owners (which now includes themselves) to downplay or even weaken their suggestions.

First, they got rid of those wonderful ranked lists that made us love Netflix in the first place, replacing them with the much more opaque cover art carousel view. Then they started mixing in lower-ranked items into the carousel. Finally, they switched from the five-star rating to the thumbs up and down buttons, which can't possibly give them as much information about your opinion.

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

#54

Post by tunesmith »

Isn't the accuracy of the predictions kind of orthogonal to the fact that they were basically lying in their attempts to change behavior?

Using lies to convince someone to do something is going to be more effective than using truth, if that "something" is not in accordance with the truth.

The comparison with netflix really breaks down there. You're not going to be able to convince me that I liked Crash, so recommendations based off of that aren't going to be very useful to me.

But if you reinforce my false belief that Obama and Soros are gonna use the deep state to invoke Sharia Law on the 2nd amendment, then that might better convince me to vote for so and so.

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

#55

Post by godelski »

joncp wrote: post #53
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 sure seems that way. It's almost as if Netflix is giving in to pressure from content owners (which now includes themselves) to downplay or even weaken their suggestions. First, they got rid of those wonderful ranked lists that made us love Netflix in the first place, replacing them with the much more opaque cover art carousel view. Then they started mixing in lower-ranked items into the carousel. Finally, they swi…

Currently I have zero trust in their rating system. Actually, zero is a bad number because if I see a high ranking I think I am going to dislike the suggestion. This is opposed to me previously trusting the system a lot.

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

#56

Post by golemotron »

RpFLCL wrote: post #34
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…

One example I can provide is of gun control topics. If you understand someone's mentality on the subject you can decide if they see: 1) An ad with someone breaking into a home and the homeowner defending themselves with a firearm (sell insurance?) 2) A grandfather and grandson on a hunting trip (hunting supplies?) 3) Or maybe gun violence hotline with powerful images. The people seeing these ads are under the assumpt…

> The people seeing these ads are under the assumption that everyone else sees them, not that it's specifically targeted at their personality type.

How long will that be true? Do people make that assumption about search results?

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

#57

Post by daodedickinson »

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.

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

#58

Post by dazc »

danso wrote: post #46
dazc wrote: post #27

Earlier quoted context omitted.

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…

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…

2 people but only one consistent regular user. My recommendations are heavily biased towards the occasional user - who is also giving very strong negative feedback?

It doesn't bother me since I know what I like. There are not that many good films that I'm not going to find them anyway.

Someone, or a group of people, are being paid for nothing though. I don't know anyone who subscribes to Netflix because of their recommendation algorithm.

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

#59

Post by frenchie14 »

golemotron wrote: post #56
RpFLCL wrote: post #34

Earlier quoted context omitted.

One example I can provide is of gun control topics. If you understand someone's mentality on the subject you can decide if they see: 1) An ad with someone breaking into a home and the homeowner defending themselves with a firearm (sell insurance?) 2) A grandfather and grandson on a hunting trip (hunting supplies?) 3) Or maybe gun violence hotline with powerful images. The people seeing these ads are under the assumpt…

> The people seeing these ads are under the assumption that everyone else sees them, not that it's specifically targeted at their personality type. How long will that be true? Do people make that assumption about search results?

Outside of the tech bubble, simply saying "yes" would be disingenuous. They're not even asking the question in the first place

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

#60

Post by demirev »

alexbeloi wrote: post #31
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?

They used it to predict political affiliation of people that don't explicitly state a party preference. The two parties already have a list of registered party members (and they can see who on Facebook explicitly states their party preference), for those members the main goal is higher turnout (they are the training data). The other voters they're interested in are unregistered (e.g. independent) voters that are like…

Did the app really grant them continuous access to user info? I thought it was a one-off thing - they get your data at the time of use (and your friends') and that's it.

Plus the approach you outlined would require the user to like/dislike things based on the add they saw, so CA can observe a change in the predicted affiliation (they didn't have access to posts as far as I know). I don't think it would have that effect (even if the add influences you, I doubt that it would make you go unlike Obama's page for example). Not to mention that by any likelihood you shouldn't be able to verify that a particular add was shown to a given individual.

I suspect it was a simpler use case - they would group users into segments, and then craft different add strategies for each one (maybe based on other research or just expert opinion).

Post reply on HN