Psychographic Snake Oil
April 24, 2018 1:13 PM Subscribe
The Big Data Panic. "Cambridge Analytica said it could move the minds of American voters. Science tells a different story."
But I agree with the fundamental premise to some degree, even as I don't think we should sweep the actual problems under the rug.
posted by symbioid at 1:26 PM on April 24, 2018 [7 favorites]
posted by symbioid at 1:26 PM on April 24, 2018 [7 favorites]
If science is telling a different story, I would have liked to hear it spelled out scientifically rather than in vague claims about "unlikeliness" and in choosing to dismember connections between targeting and racism/sexism instead of treating them as feeding into each other.
Saying people involved in the election had serious doubts about Cambridge Analytica's effectiveness at the start of the campaign is pretty pointless, while gesturing to other factors that also had possible effect isn't really addressing the issue just denying CA absolute power over voters, which no one suspected they had. Psychometric targeting may not switch people's preferences, but it may intensify those preferences and shut off further reasoning. There too is some problem with claiming voters couldn't be swayed because they are highly partisan and then trying to argue that voters have agency so microtargeting isn't likely to work. And on and on.
It's a pretty empty piece that only glances at part of the problem, chooses to define it in ways that make any measurable influence difficult, and then move on to generalities about social media that aren't all that interesting. Pundits gotta pundit though, so good job on seeming all analytical like.
posted by gusottertrout at 1:32 PM on April 24, 2018 [27 favorites]
Saying people involved in the election had serious doubts about Cambridge Analytica's effectiveness at the start of the campaign is pretty pointless, while gesturing to other factors that also had possible effect isn't really addressing the issue just denying CA absolute power over voters, which no one suspected they had. Psychometric targeting may not switch people's preferences, but it may intensify those preferences and shut off further reasoning. There too is some problem with claiming voters couldn't be swayed because they are highly partisan and then trying to argue that voters have agency so microtargeting isn't likely to work. And on and on.
It's a pretty empty piece that only glances at part of the problem, chooses to define it in ways that make any measurable influence difficult, and then move on to generalities about social media that aren't all that interesting. Pundits gotta pundit though, so good job on seeming all analytical like.
posted by gusottertrout at 1:32 PM on April 24, 2018 [27 favorites]
Oh ok, sure, then, no worries.
This might be becoming an annoying hobbyhorse for me but I often feel like people are attracted to the idea that some "cheat" or anomaly explains Trump because it's easier to confront than the Big Worry, which is that Trump winning might actually make sense in context of American politics since the 80s, if not American politics since always.
I also think some of the claims for psychographic targeting smell funny - the obvious reason to be upset anyway though is that there's no way these people should have your data in the first place.
posted by atoxyl at 1:45 PM on April 24, 2018 [31 favorites]
This might be becoming an annoying hobbyhorse for me but I often feel like people are attracted to the idea that some "cheat" or anomaly explains Trump because it's easier to confront than the Big Worry, which is that Trump winning might actually make sense in context of American politics since the 80s, if not American politics since always.
I also think some of the claims for psychographic targeting smell funny - the obvious reason to be upset anyway though is that there's no way these people should have your data in the first place.
posted by atoxyl at 1:45 PM on April 24, 2018 [31 favorites]
I feel as though this article misses the forest for the trees. It's not like a million people needed a huge mental shift in order to shift the outcome of the presidential election. Targeting a relatively small number of people in swing states with falsehoods was enough, reiterating and the amplifying the garbage they've been primed with by other sources.
posted by exogenous at 1:47 PM on April 24, 2018 [17 favorites]
posted by exogenous at 1:47 PM on April 24, 2018 [17 favorites]
Was it the source of Trump winning?
Well, I mean considering he lost the popular vote, I really don't understand how this argument keeps coming up that "such and such" secured him the nomination. Nothing secured it other than the Electoral College. Period.
However, despite that the whole argument that CA helped him "win" seems like a farce on its face when it looked more like just taking the EC seriously as the final word on the Presidency, I think this article might be overselling itself a little.
The big money wouldn't be in advertising if advertising didn't work at all. While it might seem weird to a lot of us because we are super aware of advertising and hate it, plenty of people don't even necessarily realize that a lot of what they interact with daily are ads. My point simply being ads (and by extension) political ads, can often work. Often enough to swing an election? Maybe not, but they do influence. Media is always an influence, even in minor ways, by exposing us to specific sets of ideas.
Also, he ends with this garbage:
Even so, according to him, one problem will remain: “No regulation will address the more fundamental problem that populists can use social media to spread their ideas by circumventing traditional media”
Oh, you mean like how Bernie Sanders is? How fucking awful that someone with different ideas that actually wants to help people should be able to communicate with the populace instead of it being like it used to be, where the media would tell candidates they didn't like that:
"You have not earned the right to media exposure, because you have not received enough media exposure."
Sorry, that last bit just really stuck in my craw. Populism isn't evil by definition for fucks sake, maybe sometimes the populace has real fucking concerns that aren't being properly addressed.
What we really need is regulation of fucking sockpuppet accounts. That's the god damned elephant in the room that is not being talked about that both fucking parties used during the election.
There will never, ever be a point in my life, where the idea that a person being paid to present a political idea, and is not disclosing that, and even moreso, is often using software to present themselves as a multitude of people, it ABSO-FUCKING-LUTELY SHOULD BE TREATED LIKE ANY OTHER POLITICAL AD AND BE LABELLED AS SUCH. We don't let people run political ads on television or the radio without saying they are an ad. We don't think its ethical to pay a bunch of people to go into a public square and shout down the civic opinion to be covered up with whatever the opinion they were paid for it, so why the fuck are we so ready to accept it when it comes to the internet?
The fact that that isn't even on the table tells you how serious any of the people involved actually take this shit.
posted by deadaluspark at 1:48 PM on April 24, 2018 [18 favorites]
Well, I mean considering he lost the popular vote, I really don't understand how this argument keeps coming up that "such and such" secured him the nomination. Nothing secured it other than the Electoral College. Period.
However, despite that the whole argument that CA helped him "win" seems like a farce on its face when it looked more like just taking the EC seriously as the final word on the Presidency, I think this article might be overselling itself a little.
The big money wouldn't be in advertising if advertising didn't work at all. While it might seem weird to a lot of us because we are super aware of advertising and hate it, plenty of people don't even necessarily realize that a lot of what they interact with daily are ads. My point simply being ads (and by extension) political ads, can often work. Often enough to swing an election? Maybe not, but they do influence. Media is always an influence, even in minor ways, by exposing us to specific sets of ideas.
Also, he ends with this garbage:
Even so, according to him, one problem will remain: “No regulation will address the more fundamental problem that populists can use social media to spread their ideas by circumventing traditional media”
Oh, you mean like how Bernie Sanders is? How fucking awful that someone with different ideas that actually wants to help people should be able to communicate with the populace instead of it being like it used to be, where the media would tell candidates they didn't like that:
"You have not earned the right to media exposure, because you have not received enough media exposure."
Sorry, that last bit just really stuck in my craw. Populism isn't evil by definition for fucks sake, maybe sometimes the populace has real fucking concerns that aren't being properly addressed.
What we really need is regulation of fucking sockpuppet accounts. That's the god damned elephant in the room that is not being talked about that both fucking parties used during the election.
There will never, ever be a point in my life, where the idea that a person being paid to present a political idea, and is not disclosing that, and even moreso, is often using software to present themselves as a multitude of people, it ABSO-FUCKING-LUTELY SHOULD BE TREATED LIKE ANY OTHER POLITICAL AD AND BE LABELLED AS SUCH. We don't let people run political ads on television or the radio without saying they are an ad. We don't think its ethical to pay a bunch of people to go into a public square and shout down the civic opinion to be covered up with whatever the opinion they were paid for it, so why the fuck are we so ready to accept it when it comes to the internet?
The fact that that isn't even on the table tells you how serious any of the people involved actually take this shit.
posted by deadaluspark at 1:48 PM on April 24, 2018 [18 favorites]
Was it the source of Trump winning?
I don’t know if that’s the right framing. Better is probably: “how many people had their vote changed because of things Cambridge analytica did?” And that answer is probably very very few.
posted by MisantropicPainforest at 1:52 PM on April 24, 2018 [5 favorites]
I don’t know if that’s the right framing. Better is probably: “how many people had their vote changed because of things Cambridge analytica did?” And that answer is probably very very few.
posted by MisantropicPainforest at 1:52 PM on April 24, 2018 [5 favorites]
On the one hand, if Cambridge Analytica and targeted ads didn't influence the election, then it shows that Facebook's (and Twitter's and Google's, et.al.) business model of targeted advertisements is fundamentally flawed.
On the other hand, if Cambridge Analytica and targeted ads did influence the election, then we have a new avenue for propagandists to exploit.
Either way, we should be having a discussion how Silicon Valley companies use and abuse the data they collect on us.
posted by SansPoint at 1:52 PM on April 24, 2018 [11 favorites]
On the other hand, if Cambridge Analytica and targeted ads did influence the election, then we have a new avenue for propagandists to exploit.
Either way, we should be having a discussion how Silicon Valley companies use and abuse the data they collect on us.
posted by SansPoint at 1:52 PM on April 24, 2018 [11 favorites]
The big money wouldn't be in advertising if advertising didn't work at all.
Of course, but the direct customer of advertising is the client, so the suggestion here is that Cambridge Analytica's business is selling clients on a fancy new kind of advertising that might not actually be more effective - could even be less effective - than existing kinds of advertising.
On the one hand, if Cambridge Analytica and targeted ads didn't influence the election, then it shows that Facebook's (and Twitter's and Google's, et.al.) business model of targeted advertisements is fundamentally flawed.
CA's innovation is supposed to be beyond just "targeted advertising," though - it's a particular approach to targeting.
posted by atoxyl at 1:55 PM on April 24, 2018 [3 favorites]
Of course, but the direct customer of advertising is the client, so the suggestion here is that Cambridge Analytica's business is selling clients on a fancy new kind of advertising that might not actually be more effective - could even be less effective - than existing kinds of advertising.
On the one hand, if Cambridge Analytica and targeted ads didn't influence the election, then it shows that Facebook's (and Twitter's and Google's, et.al.) business model of targeted advertisements is fundamentally flawed.
CA's innovation is supposed to be beyond just "targeted advertising," though - it's a particular approach to targeting.
posted by atoxyl at 1:55 PM on April 24, 2018 [3 favorites]
Better is probably: “how many people had their vote changed because of things Cambridge analytica did?” And that answer is probably very very few.
Or "how many didn't vote at all (or did vote) because of Cambridge Analytica's targeting?" And that answer might be quite a few. We've said many times there are no undecided voters. But there are people who are motivated to vote and people who aren't.
posted by The_Vegetables at 1:57 PM on April 24, 2018 [17 favorites]
Or "how many didn't vote at all (or did vote) because of Cambridge Analytica's targeting?" And that answer might be quite a few. We've said many times there are no undecided voters. But there are people who are motivated to vote and people who aren't.
posted by The_Vegetables at 1:57 PM on April 24, 2018 [17 favorites]
The evidence for this type of stuff increasing turnout is pretty thin too.
posted by MisantropicPainforest at 1:58 PM on April 24, 2018 [3 favorites]
posted by MisantropicPainforest at 1:58 PM on April 24, 2018 [3 favorites]
The_Vegetables: Right. Influencing an election doesn't just mean convincing someone who to vote for, but it also means convincing people to not vote. And you don't need Facebook and targeted ads to do it. For example, in 2008, someone distributed a phony flier in Virginia that said Democrats had to vote on November 5th, but Republicans had to vote on November 4th. There's a very small number of people this was likely to fool, but it may have fooled at least one person.
posted by SansPoint at 2:02 PM on April 24, 2018 [11 favorites]
posted by SansPoint at 2:02 PM on April 24, 2018 [11 favorites]
The evidence for this type of stuff increasing turnout is pretty thin too.
Then why are we still talking about 'butteremails'? If marketing/targeted marketing doesn't work, then that also had no effect. The NYT can print whatever it wants- it apparently has no effect.
posted by The_Vegetables at 2:08 PM on April 24, 2018 [1 favorite]
Then why are we still talking about 'butteremails'? If marketing/targeted marketing doesn't work, then that also had no effect. The NYT can print whatever it wants- it apparently has no effect.
posted by The_Vegetables at 2:08 PM on April 24, 2018 [1 favorite]
Assuming they were doing what they say they were doing is probably a mistake and leaves out the black money and any laundering of stolen data including voter rolls.
posted by Artw at 2:09 PM on April 24, 2018 [6 favorites]
posted by Artw at 2:09 PM on April 24, 2018 [6 favorites]
Targeted marketing is not the same as the media. Is anyone arguing otherwise?
posted by MisantropicPainforest at 2:11 PM on April 24, 2018
posted by MisantropicPainforest at 2:11 PM on April 24, 2018
Targeted marketing is not the same as the media. Is anyone arguing otherwise?
I think you are creating a distinction there that doesn't exist (ie: the media reporting of campaign happenings and drama beyond basic facts is marketing in my opinion, and creating targeted posts that look like the 'official' media blur that line even more) but fine if that is a distinction.
posted by The_Vegetables at 2:15 PM on April 24, 2018 [1 favorite]
I think you are creating a distinction there that doesn't exist (ie: the media reporting of campaign happenings and drama beyond basic facts is marketing in my opinion, and creating targeted posts that look like the 'official' media blur that line even more) but fine if that is a distinction.
posted by The_Vegetables at 2:15 PM on April 24, 2018 [1 favorite]
Again, the margins were so slim in the 2016 US election that any and all claims to effectiveness don't have to scale in order to matter. 5% of 30 million people is still 1.5 Million people.
There were so many things that went wrong that could have affected the 10,000s of voters that made up the margin, that everything becomes significant / causal. They were all necessary and all effective, despite how miniscule the effect could be. I have no doubt that this additional foreign campaigning helped Trump get elected, even if they were the crappiest campaigners that ever lived.
It does make it very confusing, then, on what to do next, because everyone will just assume that their reason is the reason, and it's difficult to gainsay.
posted by eustatic at 3:34 PM on April 24, 2018 [5 favorites]
There were so many things that went wrong that could have affected the 10,000s of voters that made up the margin, that everything becomes significant / causal. They were all necessary and all effective, despite how miniscule the effect could be. I have no doubt that this additional foreign campaigning helped Trump get elected, even if they were the crappiest campaigners that ever lived.
It does make it very confusing, then, on what to do next, because everyone will just assume that their reason is the reason, and it's difficult to gainsay.
posted by eustatic at 3:34 PM on April 24, 2018 [5 favorites]
Better is probably: “how many people had their vote changed because of things Cambridge analytica did?”
Guess what? I totally changed my vote thanks to a Facebook ad! (*) I consider myself a very high-information straight-ticket voter. I still changed my vote.
I could totally see people on the margin changing their vote in the Presidential race as a result of Facebook ads.
(*)The ad was against a bond issue I was going to vote for; it turned out the wording itself was questionable and it did a few other weird things. But I changed my vote.
posted by steady-state strawberry at 3:49 PM on April 24, 2018 [5 favorites]
Guess what? I totally changed my vote thanks to a Facebook ad! (*) I consider myself a very high-information straight-ticket voter. I still changed my vote.
I could totally see people on the margin changing their vote in the Presidential race as a result of Facebook ads.
(*)The ad was against a bond issue I was going to vote for; it turned out the wording itself was questionable and it did a few other weird things. But I changed my vote.
posted by steady-state strawberry at 3:49 PM on April 24, 2018 [5 favorites]
Didn't the Mercers want Ted Cruz? Wasn't Trump their hasty back-up plan?
posted by Lesser Shrew at 6:51 PM on April 24, 2018
posted by Lesser Shrew at 6:51 PM on April 24, 2018
This might be becoming an annoying hobbyhorse for me but I often feel like people are attracted to the idea that some "cheat" or anomaly explains Trump because it's easier to confront than the Big Worry, which is that Trump winning might actually make sense in context of American politics since the 80s, if not American politics since always.
There was. The cheat is the US’s insane electoral college system, coupled with decades of gerrymandering and voter suppression.
posted by His thoughts were red thoughts at 3:08 AM on April 25, 2018 [4 favorites]
There was. The cheat is the US’s insane electoral college system, coupled with decades of gerrymandering and voter suppression.
posted by His thoughts were red thoughts at 3:08 AM on April 25, 2018 [4 favorites]
This might be becoming an annoying hobbyhorse for me but I often feel like people are attracted to the idea that some "cheat" or anomaly explains Trump because it's easier to confront than the Big Worry, which is that Trump winning might actually make sense in context of American politics since the 80s, if not American politics since always.
More like the 60s, the 1860s specifically.
On the one hand, if Cambridge Analytica and targeted ads didn't influence the election, then it shows that Facebook's (and Twitter's and Google's, et.al.) business model of targeted advertisements is fundamentally flawed.
Not necessarily. First of all political and sales focused advertising is inherently very different. You don't get to test and tweak political ads in the same way because there's only one election. Most advertising that's trying to get you to buy things is A/B tested constantly, you can't do that with an election.
Also, CA was selling a very particular claim which is that they could use psychometric profiles to target political ads. The evidence that FB likes can be used to extract Big-5 psychometric characteristics is actually quite weak - it has been done and those results were academically interesting because they actually worked at all. That doesn't mean they can be used for practical purposes like targeting political advertising.
Most advertising targeting doesn't use any of this stuff. The most valuable information about you isn't actually *on* your FB profile, it certainly isn't in your private messages. Knowing where else you go on the web and especially where you buy things and what you buy are worth a lot. One of the most interesting things I got out of reading Chaos Monkeys is that FB came to the realisation quite early that most of their data was worthless to advertisers. What really makes them money is tracking you around the web and retargeting ads to you.
posted by atrazine at 4:06 AM on April 25, 2018 [3 favorites]
More like the 60s, the 1860s specifically.
On the one hand, if Cambridge Analytica and targeted ads didn't influence the election, then it shows that Facebook's (and Twitter's and Google's, et.al.) business model of targeted advertisements is fundamentally flawed.
Not necessarily. First of all political and sales focused advertising is inherently very different. You don't get to test and tweak political ads in the same way because there's only one election. Most advertising that's trying to get you to buy things is A/B tested constantly, you can't do that with an election.
Also, CA was selling a very particular claim which is that they could use psychometric profiles to target political ads. The evidence that FB likes can be used to extract Big-5 psychometric characteristics is actually quite weak - it has been done and those results were academically interesting because they actually worked at all. That doesn't mean they can be used for practical purposes like targeting political advertising.
Most advertising targeting doesn't use any of this stuff. The most valuable information about you isn't actually *on* your FB profile, it certainly isn't in your private messages. Knowing where else you go on the web and especially where you buy things and what you buy are worth a lot. One of the most interesting things I got out of reading Chaos Monkeys is that FB came to the realisation quite early that most of their data was worthless to advertisers. What really makes them money is tracking you around the web and retargeting ads to you.
posted by atrazine at 4:06 AM on April 25, 2018 [3 favorites]
Knowing where else you go on the web and especially where you buy things and what you buy are worth a lot.
Thankfully, it's pretty clear that Cambridge Analytica's UK political customers had all that data, and illegally collected, too.
The use of behavioural analytics in targetting that's the focus of the FPP post seems to be a red herring: presumably any behavioural things were used to hone the messages and monitor the successes of the war of psychological attrition that was behind exhausting (the fact-based) Clinton and Remain supporters and bolstering (the resentment-based) Trump and Brexit supporters.
And the targetting was done by the more traditional means.
posted by ambrosen at 4:41 AM on April 25, 2018 [2 favorites]
Thankfully, it's pretty clear that Cambridge Analytica's UK political customers had all that data, and illegally collected, too.
The use of behavioural analytics in targetting that's the focus of the FPP post seems to be a red herring: presumably any behavioural things were used to hone the messages and monitor the successes of the war of psychological attrition that was behind exhausting (the fact-based) Clinton and Remain supporters and bolstering (the resentment-based) Trump and Brexit supporters.
And the targetting was done by the more traditional means.
posted by ambrosen at 4:41 AM on April 25, 2018 [2 favorites]
I have this argument in my day job a lot.
90% of big data arguments are bunk. But there are so many 'experts', and so many innumerate people, they're making the world worse by adding so much NOISE.
posted by DigDoug at 6:01 AM on April 25, 2018 [2 favorites]
90% of big data arguments are bunk. But there are so many 'experts', and so many innumerate people, they're making the world worse by adding so much NOISE.
posted by DigDoug at 6:01 AM on April 25, 2018 [2 favorites]
What really makes them money is tracking you around the web and retargeting ads to you.
It's interesting that you say this because yesterday when I was reading this thread I was thinking about what form of online advertising I think is most likely to work on me and the first thing that came to mind is "something I specifically looked at (or something else from the same company/in the same category) following me around to remind me I wanted it."
posted by atoxyl at 9:07 AM on April 25, 2018 [1 favorite]
It's interesting that you say this because yesterday when I was reading this thread I was thinking about what form of online advertising I think is most likely to work on me and the first thing that came to mind is "something I specifically looked at (or something else from the same company/in the same category) following me around to remind me I wanted it."
posted by atoxyl at 9:07 AM on April 25, 2018 [1 favorite]
Cambridge Analytica knew your job, profession, income, religion, gender, reading habits, activities, personality type, and what issues were most likely to motivate you. They had Russian backed funding, an army of people creating ads, links to fake news sources, presented themselves as members of your social circle, and knew in advance of events that could be easily exploited. If ever anyone in history had a chance of crafting an ad that could sway a voter, it was Cambridge Analytica with Russian cooperation.
I understand that the experts have looked into this and say political beliefs cannot be swayed by ads. I also know (A) you don't have to change someone's political beliefs to sway their vote, (B) My doctor will prescribe a drug because a sales rep has given him a free pen, (C) If you subscribe the New York Time's Book Review, I can win money all day long betting on what else you believe, (D) I have seen plenty of elections swayed by ads (e.g. remember Willie Horton?)
Was the vote swayed? I know it was. I know people who didn't vote because they were put out by the non-stop email scandals. I know people that felt that Trump was essentially independent, wanting Medicare for all. I know people that felt Hillary's charitable foundation was a set up to take bribes from corporations. I could go on all day long with this stuff, listing dubious claims, or beliefs pushed to extremes, or differences claimed where there were none.
Did this stuff make a difference? What planet were you on during the election?
posted by xammerboy at 10:11 AM on April 25, 2018 [5 favorites]
I understand that the experts have looked into this and say political beliefs cannot be swayed by ads. I also know (A) you don't have to change someone's political beliefs to sway their vote, (B) My doctor will prescribe a drug because a sales rep has given him a free pen, (C) If you subscribe the New York Time's Book Review, I can win money all day long betting on what else you believe, (D) I have seen plenty of elections swayed by ads (e.g. remember Willie Horton?)
Was the vote swayed? I know it was. I know people who didn't vote because they were put out by the non-stop email scandals. I know people that felt that Trump was essentially independent, wanting Medicare for all. I know people that felt Hillary's charitable foundation was a set up to take bribes from corporations. I could go on all day long with this stuff, listing dubious claims, or beliefs pushed to extremes, or differences claimed where there were none.
Did this stuff make a difference? What planet were you on during the election?
posted by xammerboy at 10:11 AM on April 25, 2018 [5 favorites]
« Older "... the sad truth is most of the time we call her... | Why suffer a flawed family when you can hire a... Newer »
This thread has been archived and is closed to new comments
Oh ok, sure, then, no worries.
posted by symbioid at 1:26 PM on April 24, 2018 [19 favorites]