Big Data and Political Propaganda
May 8, 2017 1:08 AM Subscribe
The great British Brexit robbery: how our democracy was hijacked. Details about Cambridge Analytica, Robert Mercer's big data company that helped Trump and Brexit achieve political victory. How Facebook and marketing data are creating a new kind of precision electioneering.
I wanted more details about what, like, they actually did. What DID they do in Trinidad? Report to me quasi-named source!
posted by DeepSeaHaggis at 1:51 AM on May 8, 2017 [3 favorites]
posted by DeepSeaHaggis at 1:51 AM on May 8, 2017 [3 favorites]
In the context of Brexit I think Jo Maugham's tweet hits on the central issue here: "A Government whose election may have breached criminal law is delivering a referendum result which may have breached criminal law."
posted by robself at 2:30 AM on May 8, 2017 [13 favorites]
posted by robself at 2:30 AM on May 8, 2017 [13 favorites]
I tried reading this on my flight from London to New York yesterday and found it impossible to get through, and I'm basically a caricature of the kind of person who should lap this up -- I live in Brooklyn, read the NYT, the Guardian/Observer and Le Monde, I'm pro-EU, anti-Trump, anti-Brexit, have a graduate degree, love international travel, blah blah blah.
The whole piece to me was written in an over-the-top breathless conspiracy tone that made it difficult to follow; I'm all for journalism that sets the stage with an interesting/breathless lede, but it felt like the details never got explained and the allusions were never clarified. It also felt to me like it was written for an audience that was already somewhat familiar with Cambridge Analytica, Aggregate IQ, etc.
I'm not necessarily disputing the truth behind the article -- I can very well believe that a shady data analytics firm had a lot to do with Trump and Brexit. But this wasn't the article to do that for me.
posted by andrewesque at 2:40 AM on May 8, 2017 [24 favorites]
The whole piece to me was written in an over-the-top breathless conspiracy tone that made it difficult to follow; I'm all for journalism that sets the stage with an interesting/breathless lede, but it felt like the details never got explained and the allusions were never clarified. It also felt to me like it was written for an audience that was already somewhat familiar with Cambridge Analytica, Aggregate IQ, etc.
I'm not necessarily disputing the truth behind the article -- I can very well believe that a shady data analytics firm had a lot to do with Trump and Brexit. But this wasn't the article to do that for me.
posted by andrewesque at 2:40 AM on May 8, 2017 [24 favorites]
There are so many issues entangled in this mess, but the most irksome is this. If people on the whole are not longer capable of forming "the will of the people," what's left of the democratic commonwealth?
We're approaching the age when the information asymmetry can no longer be bridged. In this informational dystopia the have-information can deal with the have-not-information in whatever way with impunity. It's no longer merely socialism for the 1% and capitalism for the 99%. It's humanism for the former and behaviorism for the latter.
The assumption of a democratic republic is that the will of the people, however diverse and flawed, must be the source from which legitimate sovereignty is derived. The will of the people is not identical to power. It is however the first cause that sets the democratic process in motion, and ultimately the judge of its outcome.
What if the will of the people is substituted with a simulacrum? There's no longer the capability for a public opinion to form. Public opinion must be nurtured by public sensory input, and training in its digestion. We no longer have the former, for the media is no longer public. Rather, it is tailored algorithmically to reinforce personal bias and distract from our real condition. The latter, education of the mind, is being dismantled and replaced by standardized metrics that measure and reward conformity. The entire business cycle of surveillance-advertisement-propaganda capital flow reinforces the behaviors of everyone tied to it.
If this is not the present, it might be the future. We're seeing the pieces falling in place.
posted by runcifex at 2:49 AM on May 8, 2017 [42 favorites]
It is substantially true, that virtue or morality is a necessary spring of popular government. The rule, indeed, extends with more or less force to every species of free government. Who, that is a sincere friend to it, can look with indifference upon attempts to shake the foundation of the fabric ?(George Washington: Farewell Address)
Promote, then, as an object of primary importance, institutions for the general diffusion of knowledge. In proportion as the structure of a government gives force to public opinion, it is essential that public opinion should be enlightened.
------BEGIN DEPRESSIVE SPECULATION------We're approaching the age when the information asymmetry can no longer be bridged. In this informational dystopia the have-information can deal with the have-not-information in whatever way with impunity. It's no longer merely socialism for the 1% and capitalism for the 99%. It's humanism for the former and behaviorism for the latter.
The assumption of a democratic republic is that the will of the people, however diverse and flawed, must be the source from which legitimate sovereignty is derived. The will of the people is not identical to power. It is however the first cause that sets the democratic process in motion, and ultimately the judge of its outcome.
What if the will of the people is substituted with a simulacrum? There's no longer the capability for a public opinion to form. Public opinion must be nurtured by public sensory input, and training in its digestion. We no longer have the former, for the media is no longer public. Rather, it is tailored algorithmically to reinforce personal bias and distract from our real condition. The latter, education of the mind, is being dismantled and replaced by standardized metrics that measure and reward conformity. The entire business cycle of surveillance-advertisement-propaganda capital flow reinforces the behaviors of everyone tied to it.
If this is not the present, it might be the future. We're seeing the pieces falling in place.
------END DEPRESSIVE SPECULATION------posted by runcifex at 2:49 AM on May 8, 2017 [42 favorites]
It's also worth remembering that the President was surprised at his own victory, and that their models were similarly predicting losses. If your model is protecting results as badly as everyone else, you aren't geniuses. This isn't to discredit the notion that Cambridge Analytica may have some ties worth investigating - but it's different from saying that they were exceptional at their jobs.
posted by Going To Maine at 3:22 AM on May 8, 2017 [5 favorites]
posted by Going To Maine at 3:22 AM on May 8, 2017 [5 favorites]
As an article it's about Brexit, not Trump, and about breaches of UK election law, particularly in terms of breaching (very modest but adequate) spending limits. Read from that point of view, it makes a lot more sense.
posted by ambrosen at 3:27 AM on May 8, 2017 [15 favorites]
posted by ambrosen at 3:27 AM on May 8, 2017 [15 favorites]
It would be great if we could keep UK politics threads about the UK and not about the US. For one thing, there is a different legal system in play.
posted by GenjiandProust at 4:05 AM on May 8, 2017 [27 favorites]
posted by GenjiandProust at 4:05 AM on May 8, 2017 [27 favorites]
Mod note: A few deleted. Yes, please. I realize the article brings up Trump a lot, but it's in the context of how Brexit is connected to Trump. The US politics stuff just completely sucks every bit of air from any discussion that leaves even the smallest sliver of room to wedge it in there (not to mention so much of it already having been said a billion times, and having ongoing permanent threads dedicated to talking about this shit), so please, let's try to avoid it here.
posted by taz (staff) at 4:24 AM on May 8, 2017 [20 favorites]
posted by taz (staff) at 4:24 AM on May 8, 2017 [20 favorites]
I think we're gonna be seeing more about Mercer and CA in the coming months.
posted by sutt at 4:58 AM on May 8, 2017
posted by sutt at 4:58 AM on May 8, 2017
Fuuuuck. I am only part way through it, but it has occurred to me:
Thank God for Metafilter. Thank God that 90% of my Internet experience is mediated throught this hokey little site. I know its provenance, I know and trust the people who own it, who have worked here, who work here now.
I feel like I am in a very lucky bubble of trustworthiness, a tiny little bit of facts and sanity and sober debate in a sea of evil bullshit. A remnant of the Idealists' Internet of the 90s.
Strap me to the mast.
oh and PS, Thank you, Great Firewall of China, for cutting me away from Facebook. I was never super comfortable with that place but it was just a lucky happenstance that geography forced me to quit.
posted by Meatbomb at 5:01 AM on May 8, 2017 [25 favorites]
Thank God for Metafilter. Thank God that 90% of my Internet experience is mediated throught this hokey little site. I know its provenance, I know and trust the people who own it, who have worked here, who work here now.
I feel like I am in a very lucky bubble of trustworthiness, a tiny little bit of facts and sanity and sober debate in a sea of evil bullshit. A remnant of the Idealists' Internet of the 90s.
Strap me to the mast.
oh and PS, Thank you, Great Firewall of China, for cutting me away from Facebook. I was never super comfortable with that place but it was just a lucky happenstance that geography forced me to quit.
posted by Meatbomb at 5:01 AM on May 8, 2017 [25 favorites]
Every time this stuff comes up there's someone who breezily dismisses it because they're not "geniuses" or whatever. That seems to miss the point about the data sets they have. If they've built a detailed personal, political and psychological profile of you, attached to your name and address and any other actual signifiers, specifically for the purpose of identifying political allies, enemies, and persuadables, does it fucking matter if it's 100% accurate? The article says they used these same tools to develop a criminal database in Trinidad and Tobago. Meanwhile, in the Philippines, Duterte's government is sending out death squads to people's homes based on a list of names.
If the authoritaritarian surveillance state kicks in your fucking door one day because something in their database tells them to, does it matter if they were geniuses about it?
posted by schadenfrau at 5:06 AM on May 8, 2017 [22 favorites]
If the authoritaritarian surveillance state kicks in your fucking door one day because something in their database tells them to, does it matter if they were geniuses about it?
posted by schadenfrau at 5:06 AM on May 8, 2017 [22 favorites]
Also of interest # Snowman Series 1 - Leave.EU faces inquiry over Cambridge Analytica.
The rest of the Snowman series.
For those a bit lost the FPP article is Carole Cadwalladr's third article chasing this down.
posted by adamvasco at 5:11 AM on May 8, 2017
The rest of the Snowman series.
For those a bit lost the FPP article is Carole Cadwalladr's third article chasing this down.
posted by adamvasco at 5:11 AM on May 8, 2017
It's an appallingly-written mess, which is a shame as the central point about really rather Byzantine campaign funding / spending is a good one and points to, at best, some pretty shady activity. The Cambridge Analytica / data stuff is sort of a red herring, as at no point has anyone managed to prove that it's anything more than buying a bunch of Facebook ads.
posted by muggsy1079 at 5:18 AM on May 8, 2017 [4 favorites]
posted by muggsy1079 at 5:18 AM on May 8, 2017 [4 favorites]
Here's the assessment of George Washington university David Karpf on Cambridge Analytica's capabilities and role in recent elections: "I have described them as the Theranos of political data: I think they have a tremendous marketing department, coupled with a team of research scientists who provide on virtually none of those marketing promises."
posted by honest knave at 5:26 AM on May 8, 2017 [16 favorites]
posted by honest knave at 5:26 AM on May 8, 2017 [16 favorites]
Can anyone explain the significance of the electoral laws they are supposed to have broken?
My fuzzy understanding of the accusations against the tories at the last general election is that they were using national spending in local races? Presumably that’s not what’s going on here, since the referendum was an inherently national campaign.
If several campaign groups who are all on the same side of the same national race were coordinating their spending, that doesn’t immediately sound that sinister… but I assume it is illegal for a reason, can anyone explain what the issue is?
posted by Bloxworth Snout at 5:36 AM on May 8, 2017
My fuzzy understanding of the accusations against the tories at the last general election is that they were using national spending in local races? Presumably that’s not what’s going on here, since the referendum was an inherently national campaign.
If several campaign groups who are all on the same side of the same national race were coordinating their spending, that doesn’t immediately sound that sinister… but I assume it is illegal for a reason, can anyone explain what the issue is?
posted by Bloxworth Snout at 5:36 AM on May 8, 2017
"This is one of the smartest computer scientists in the world. He is not going to splash $15m on bullshit.”
Why not?
posted by escabeche at 5:42 AM on May 8, 2017 [9 favorites]
Why not?
posted by escabeche at 5:42 AM on May 8, 2017 [9 favorites]
Bloxworth, the current Twitter thread explains a bit of it, but to quote the most salient bit: "To be clear: co-ordination between campaigns is AGAINST THE LAW, unless expenditure is declared jointly..."
posted by muggsy1079 at 5:47 AM on May 8, 2017 [3 favorites]
posted by muggsy1079 at 5:47 AM on May 8, 2017 [3 favorites]
I think what cambridge analytica is supposed to be doing is not as opaque or weird as people here are making it out to be. As I understand it, the process basically goes like this:
tldr don't do facebook personality quizzes.
posted by advil at 6:13 AM on May 8, 2017 [14 favorites]
- Gather as much information as you can on as many relevant people. The innovation here (except I'm sure they're not the only ones) is using viral facebook personality quizzes, which as I understand it also allow access to friends of friends on default privacy settings. They're supposedly getting data on huge amounts of people linked to exact identities, but not atypical numbers for online advertising.
- Identify the most persuadable individuals in their data set on some issue, and what it would take to persuade them. (Say, the most persuadable 1%.)
- Saturate those individuals with advertising and/or propaganda customized to what you know about them.
tldr don't do facebook personality quizzes.
posted by advil at 6:13 AM on May 8, 2017 [14 favorites]
Yes, I understand that it is against the law, and I completely support as much transparency as possible in elections. But in the specific case of campaigns on the same side of a national yes/no referendum it doesn’t seem as sinister as if, for example, two political parties were secretly colluding during a general election. I was wondering if it gave them some non-obvious advantage: helping them skirt around spending limits, for example.
posted by Bloxworth Snout at 6:20 AM on May 8, 2017 [1 favorite]
posted by Bloxworth Snout at 6:20 AM on May 8, 2017 [1 favorite]
Every time this stuff comes up there's someone who breezily dismisses it because they're not "geniuses" or whatever. That seems to miss the point about the data sets they have.
The dismissals come up because what they're doing isn't dark wizardry. The examples we've seen have been on the level of "If someone likes a bunch of stuff about Haiti, show them an ad about Clinton being mean to Haiti." Even to the extent this stuff is successful, it's not "They have deep models that super-mega-understand you!!" so much as "Even very simple models can be useful."
Where there's more interesting things going on, it would be more along the lines of "You liked a bunch of stuff that people who liked a lot of Haiti stuff also liked, so maybe you're Haitian too?" Which sounds a lot like basic factor analysis or similar simple data-reduction techniques, not deep models that understand you better than your spouse does.
posted by ROU_Xenophobe at 7:01 AM on May 8, 2017 [7 favorites]
The dismissals come up because what they're doing isn't dark wizardry. The examples we've seen have been on the level of "If someone likes a bunch of stuff about Haiti, show them an ad about Clinton being mean to Haiti." Even to the extent this stuff is successful, it's not "They have deep models that super-mega-understand you!!" so much as "Even very simple models can be useful."
Where there's more interesting things going on, it would be more along the lines of "You liked a bunch of stuff that people who liked a lot of Haiti stuff also liked, so maybe you're Haitian too?" Which sounds a lot like basic factor analysis or similar simple data-reduction techniques, not deep models that understand you better than your spouse does.
posted by ROU_Xenophobe at 7:01 AM on May 8, 2017 [7 favorites]
The campaign limit depends a bit on details, but for all these groups, it was £700,000 (in PDF, lower down the page). Which, well, I've had the misfortune of business dealings with people who've spent more than that on a primary for the Texas state legislature.
posted by ambrosen at 7:03 AM on May 8, 2017 [2 favorites]
posted by ambrosen at 7:03 AM on May 8, 2017 [2 favorites]
The dismissals come up because what they're doing isn't dark wizardry
Yes, and this fantastically misses the point. They don't need to be scarily accurate to do terrible things with these data sets.
posted by schadenfrau at 7:13 AM on May 8, 2017 [3 favorites]
Yes, and this fantastically misses the point. They don't need to be scarily accurate to do terrible things with these data sets.
posted by schadenfrau at 7:13 AM on May 8, 2017 [3 favorites]
This article is ridiculous, heavy breathing onanism. "Data and micro targeting" equals "basic 21st Century marketing." All the links among various individuals and companies cited = folks who share beliefs and business ties. It's always a "dark conspiracy" when folks on the other side do anything together.
posted by twsf at 7:32 AM on May 8, 2017 [3 favorites]
posted by twsf at 7:32 AM on May 8, 2017 [3 favorites]
The examples we've seen have been on the level of "If someone likes a bunch of stuff about Haiti, show them an ad about Clinton being mean to Haiti."
To be fair, the OCEAN personality model is a bit more sophisticated than “SELECT FROM voters * WHERE mentions("$liberal_keyword")”. And the models can only get better.
Eventually, there'll be an automated psychoanalysis, which will be used to control. Coupled with nation-state mass surveillance, it could be used not only for hunting down pre-crime (“sleeper cells”, “clean-skin deep-cover operatives”) or thoughtcrime (“radicalisation” is the accepted, appropriately racialised, term for it), but for punishing insolence at a fine-grained level.
posted by acb at 7:47 AM on May 8, 2017 [1 favorite]
To be fair, the OCEAN personality model is a bit more sophisticated than “SELECT FROM voters * WHERE mentions("$liberal_keyword")”. And the models can only get better.
Eventually, there'll be an automated psychoanalysis, which will be used to control. Coupled with nation-state mass surveillance, it could be used not only for hunting down pre-crime (“sleeper cells”, “clean-skin deep-cover operatives”) or thoughtcrime (“radicalisation” is the accepted, appropriately racialised, term for it), but for punishing insolence at a fine-grained level.
posted by acb at 7:47 AM on May 8, 2017 [1 favorite]
This article is ridiculous, heavy breathing onanism. "Data and micro targeting" equals "basic 21st Century marketing."
Yeah but "basic 21st Century marketing" is what was known in the 20th century and earlier as "propaganda" and "disinformation" and "indoctrination", and the only reason we accept active mass campaigns to manipulate and misinform the citizenry of a country as something which is normal and commonplace, or even somehow necessary, is because of propaganda to that effect and euphemistic terms like "marketing" and "public relations".
posted by XMLicious at 7:50 AM on May 8, 2017 [13 favorites]
There are so many issues entangled in this mess, but the most irksome is this. If people on the whole are not longer capable of forming "the will of the people," what's left of the democratic commonwealth?
Well, one of my takeaways from the Brexit vote in particular, is that a 50.0001% victory margin on a particular issue on one day of one particular week of one particular month of one particular year does NOT at all represent the real "will of the people".
The "will of the people" as measured by voting or polling is fairly fickle, changeable (within certain limits), and susceptible to misinformation, manipulation, etc etc etc. People vote (or abstain from voting) for a variety of reasons, including e.g. revenge voting.
The upshot of all of that is that a country should not make major constitution-level changes--which Brexit surely was--on the basis of one single bare-majority popular vote.
For sort of everyday lawmaking, where it is really more important to have the matter settled one way or the other and proceed, sure. A one-time 50% bare majority works just as well as many other decision-making procedures.
But when you are basically changing the direction of the entire ship of state by a major amount, the threshhold should be markedly higher.
A higher percentage requirement would be one way to do it--say, 55% support required.
Another, perhaps even better way would be two majority popular votes required, one or two years apart.
It would be an interesting question to explore--whether Brexit would have received a majority vote the second time around, a year later, after the country had an entire year to really consider the ramifications of the move.
posted by flug at 8:05 AM on May 8, 2017 [7 favorites]
Well, one of my takeaways from the Brexit vote in particular, is that a 50.0001% victory margin on a particular issue on one day of one particular week of one particular month of one particular year does NOT at all represent the real "will of the people".
The "will of the people" as measured by voting or polling is fairly fickle, changeable (within certain limits), and susceptible to misinformation, manipulation, etc etc etc. People vote (or abstain from voting) for a variety of reasons, including e.g. revenge voting.
The upshot of all of that is that a country should not make major constitution-level changes--which Brexit surely was--on the basis of one single bare-majority popular vote.
For sort of everyday lawmaking, where it is really more important to have the matter settled one way or the other and proceed, sure. A one-time 50% bare majority works just as well as many other decision-making procedures.
But when you are basically changing the direction of the entire ship of state by a major amount, the threshhold should be markedly higher.
A higher percentage requirement would be one way to do it--say, 55% support required.
Another, perhaps even better way would be two majority popular votes required, one or two years apart.
It would be an interesting question to explore--whether Brexit would have received a majority vote the second time around, a year later, after the country had an entire year to really consider the ramifications of the move.
posted by flug at 8:05 AM on May 8, 2017 [7 favorites]
Every time this stuff comes up there's someone who breezily dismisses it because they're not "geniuses" or whatever. That seems to miss the point about the data sets they have.
This article claims that the victories of Trump and Brexit are reasons to be afraid of these people. The point of saying these people are not actually geniuses is to say, although there may be reasons to be afraid of these people, there's not very good reason to believe that the Trump/Brexit victories are evidence of their power.
posted by straight at 8:17 AM on May 8, 2017 [3 favorites]
This article claims that the victories of Trump and Brexit are reasons to be afraid of these people. The point of saying these people are not actually geniuses is to say, although there may be reasons to be afraid of these people, there's not very good reason to believe that the Trump/Brexit victories are evidence of their power.
posted by straight at 8:17 AM on May 8, 2017 [3 favorites]
I do hope that the next 2016 campaign book does really focus on the way the campaign data shops DIDN'T identify the trends.
Everyone knew that Brexit would pass only with a surge of English-identified English voters outside of greater London. Everyone knew that Trump would win only with exurban and rural white voters in FL, PA, OH, MI and WI swinging over big time.
It's just that the technology didn't see the trend to those events happening. I can't quite speak to Leave/Remain, but Trump and Clinton saw the same numbers and analyzed them the same way. For Trump it meant no change: keep focusing on the key people in the key states because that's the only way to win, however unlikely. For Clinton, however, it led to the fatal, but entirely rational, reallocation of strategy to Electoral College realignment (AZ, GA) and a Senate majority.
posted by MattD at 8:51 AM on May 8, 2017
Everyone knew that Brexit would pass only with a surge of English-identified English voters outside of greater London. Everyone knew that Trump would win only with exurban and rural white voters in FL, PA, OH, MI and WI swinging over big time.
It's just that the technology didn't see the trend to those events happening. I can't quite speak to Leave/Remain, but Trump and Clinton saw the same numbers and analyzed them the same way. For Trump it meant no change: keep focusing on the key people in the key states because that's the only way to win, however unlikely. For Clinton, however, it led to the fatal, but entirely rational, reallocation of strategy to Electoral College realignment (AZ, GA) and a Senate majority.
posted by MattD at 8:51 AM on May 8, 2017
You know Obama beat Romney by about 3% of the vote, and that was considered a big win. If you can influence an election by a percentage point or two that is a devastatingly effective tool.
posted by xammerboy at 9:14 AM on May 8, 2017 [1 favorite]
posted by xammerboy at 9:14 AM on May 8, 2017 [1 favorite]
Data people will all tell you the same thing: companies like Cambridge are 99% scam shops. I have worked near and for companies like this, and they have no value add besides cool marketing. Usually the work of the shitty data analysis they do costs them more than they gain.
posted by TypographicalError at 9:54 AM on May 8, 2017 [5 favorites]
posted by TypographicalError at 9:54 AM on May 8, 2017 [5 favorites]
I posted the article even though I agree it's problematic; too much emphasis on conspiracies about money and not enough on the technology of what organizations like Cambridge Analytica are doing. Perhaps I should have included some links to other things I've been reading; it's the context of all of this together that got my attention.
One companion piece to read is the New Yorker profile on Robert Mercer. He's a remarkably strange person for someone so influential.
Another contextual item here is the book Ratf*cked. It's a detailed story of how Republicans created the REDMAP initiative that took control of a bunch of seats in the House of Representatives but controlling the 2010 redistricting process. They employed technology that lets them draw district lines to literally move blocks of ~50 votes from one district to another with specific desired outcomes on the vote. A key part of that process is demographic and marketing data. Mercer was not connected to REDMAP, but it seems a similar sort of effort.
My impression is the state of the art in 2010 was just having basic data and doing some simple statistical modeling. Applying machine learning technology to very precise data gleaned from Facebook and online marketing allows for a very precise form of control. The idea this apparatus may have been turned on the Brexit election as well seems important.
posted by Nelson at 10:12 AM on May 8, 2017 [3 favorites]
One companion piece to read is the New Yorker profile on Robert Mercer. He's a remarkably strange person for someone so influential.
Another contextual item here is the book Ratf*cked. It's a detailed story of how Republicans created the REDMAP initiative that took control of a bunch of seats in the House of Representatives but controlling the 2010 redistricting process. They employed technology that lets them draw district lines to literally move blocks of ~50 votes from one district to another with specific desired outcomes on the vote. A key part of that process is demographic and marketing data. Mercer was not connected to REDMAP, but it seems a similar sort of effort.
My impression is the state of the art in 2010 was just having basic data and doing some simple statistical modeling. Applying machine learning technology to very precise data gleaned from Facebook and online marketing allows for a very precise form of control. The idea this apparatus may have been turned on the Brexit election as well seems important.
posted by Nelson at 10:12 AM on May 8, 2017 [3 favorites]
Why not?
Yes, Bob Mercer could probably find $15m in his couch cushions. I wouldn't put too much weight in his having put that much money into it.
posted by praemunire at 10:20 AM on May 8, 2017
Yes, Bob Mercer could probably find $15m in his couch cushions. I wouldn't put too much weight in his having put that much money into it.
posted by praemunire at 10:20 AM on May 8, 2017
I work in marketing for advocacy groups, which is close-ish to political campaigns. I think the thing with this targeting that they did is that it's ... basically what you're supposed to do. It's not black magic, it's pretty straightforward. As with so many things in both political campaigns and marketing, it's less about the sexy "strategy" and more just about having the funding and resources to actually do the best practices on a large scale.
I mean, in another breathless piece about Cambridge Analytica, there was a whole thing about how to they ... uploaded lists of supporters to Facebook! And created target audiences that had similar qualities! Which is a really bog-standard Facebook marketing tactic and I am 1,000% sure the Clinton campaign did the same. Hell, I do that all the time with my dinky FB marketing budget.
Two caveats to this, though:
- The quiz thing is really interesting and a bit nauseating. I haven't seen that done in that way by a campaign before.
- Most political marketing people I know are pretty skeptical about this kind of targeting (consumer data, facebook "interests") because it's usually not as effective as less sexy kinds (like basic demographic data).
I don't know, I am always trying to learn from what works for others, but there is SO MUCH snake oil in this field. My default setting is now "deep skepticism" for any wild claims about things like this.
posted by lunasol at 10:24 AM on May 8, 2017 [1 favorite]
I mean, in another breathless piece about Cambridge Analytica, there was a whole thing about how to they ... uploaded lists of supporters to Facebook! And created target audiences that had similar qualities! Which is a really bog-standard Facebook marketing tactic and I am 1,000% sure the Clinton campaign did the same. Hell, I do that all the time with my dinky FB marketing budget.
Two caveats to this, though:
- The quiz thing is really interesting and a bit nauseating. I haven't seen that done in that way by a campaign before.
- Most political marketing people I know are pretty skeptical about this kind of targeting (consumer data, facebook "interests") because it's usually not as effective as less sexy kinds (like basic demographic data).
I don't know, I am always trying to learn from what works for others, but there is SO MUCH snake oil in this field. My default setting is now "deep skepticism" for any wild claims about things like this.
posted by lunasol at 10:24 AM on May 8, 2017 [1 favorite]
If you can influence an election by a percentage point or two that is a devastatingly effective tool.
But that is an enormous "if" that lies completely unsubstantiated at the heart of this article.
posted by straight at 10:40 AM on May 8, 2017 [1 favorite]
But that is an enormous "if" that lies completely unsubstantiated at the heart of this article.
It’s not a stretch to believe that a member of the global 1% found a way to influence this crucial 1% of British voters.It's also not a stretch to believe that they didn't. Did they change that many people's minds? And more importantly, did they change that many more people's minds than they would have by putting the same money into traditional TV ads? Nobody knows.
posted by straight at 10:40 AM on May 8, 2017 [1 favorite]
I fully understand that all this is shady, I am a privacy advocate and all that - but you could do a straight substitution of "Big Data" or "Data Analytics" or "Data Science" for "Witchcraft!!!1!!!" and have essentially the same article.
I have a pretty good idea what the state of the art is. And it is not witchcraft or anywhere near witchcraft. If I had those magic powers would I be sitting in my cubicle typing into a little Metafilter edit window?
Of course not. I'd be sitting on my yacht typing into a little Metafilter window. Do you see a yacht? Of course not. Therefore, "data science", as practised by Cambridge Analytica or anyone else, is not magic.
posted by tel3path at 10:44 AM on May 8, 2017 [2 favorites]
I have a pretty good idea what the state of the art is. And it is not witchcraft or anywhere near witchcraft. If I had those magic powers would I be sitting in my cubicle typing into a little Metafilter edit window?
Of course not. I'd be sitting on my yacht typing into a little Metafilter window. Do you see a yacht? Of course not. Therefore, "data science", as practised by Cambridge Analytica or anyone else, is not magic.
posted by tel3path at 10:44 AM on May 8, 2017 [2 favorites]
So fun to read but where's the beef
posted by Joseph Gurl at 10:47 AM on May 8, 2017
posted by Joseph Gurl at 10:47 AM on May 8, 2017
But what if witches prefer cubicles to yachts? Small rocks also prefer cubicles to yachts.
posted by XMLicious at 11:01 AM on May 8, 2017
posted by XMLicious at 11:01 AM on May 8, 2017
Data science is not magic; the difference is statistical machine learning actually works. Not 100%, and not entirely predictably, but it's a significantly powerful new tool. Similarly there's nothing exactly new about market research or using it for political targeting. But the volume of data that is now available in the Internet era, combined with the processing power and algorithms to find new patterns in that data, is a huge progression in the ability for politicians to control the electorate.
Also while you may not be sitting in a yacht, Mercer and his colleagues at Renaissance Technologies are. Or could be if they choose to. Their trading strategies are proprietary but rumored to be data science. You may laugh it off as witchcraft, but they have been enormously successful.
Political scientists definitely have the power to target and persuade the votes of very small groups of people in a precise manner; perhaps a group as small as a single individual. These tools are powerful and effective. The fact that the Guardian has found evidence that they were applied in a precise way to shape the Trump and Brexit elections at the whim of someone like Mercer seems worth paying attention to.
posted by Nelson at 11:04 AM on May 8, 2017 [3 favorites]
Also while you may not be sitting in a yacht, Mercer and his colleagues at Renaissance Technologies are. Or could be if they choose to. Their trading strategies are proprietary but rumored to be data science. You may laugh it off as witchcraft, but they have been enormously successful.
Political scientists definitely have the power to target and persuade the votes of very small groups of people in a precise manner; perhaps a group as small as a single individual. These tools are powerful and effective. The fact that the Guardian has found evidence that they were applied in a precise way to shape the Trump and Brexit elections at the whim of someone like Mercer seems worth paying attention to.
posted by Nelson at 11:04 AM on May 8, 2017 [3 favorites]
He's a remarkably strange person for someone so influential.
"delightfully eccentric".
posted by Going To Maine at 11:39 AM on May 8, 2017
"delightfully eccentric".
posted by Going To Maine at 11:39 AM on May 8, 2017
Gather as much information as you can on as many relevant people. The innovation here (except I'm sure they're not the only ones) is using viral facebook personality quizzes, which as I understand it also allow access to friends of friends on default privacy settings. They're supposedly getting data on huge amounts of people linked to exact identities, but not atypical numbers for online advertising.
Identify the most persuadable individuals in their data set on some issue, and what it would take to persuade them. (Say, the most persuadable 1%.)
Saturate those individuals with advertising and/or propaganda customized to what you know about them.
Building a human mal-info botnet basically.
posted by srboisvert at 11:39 AM on May 8, 2017
Identify the most persuadable individuals in their data set on some issue, and what it would take to persuade them. (Say, the most persuadable 1%.)
Saturate those individuals with advertising and/or propaganda customized to what you know about them.
Building a human mal-info botnet basically.
posted by srboisvert at 11:39 AM on May 8, 2017
There've been a bunch of breathless articles about this company and I still think a lot of it amounts to advertising for CA. Not that none of this stuff works, or will be important, but do these guys really have secret sauce nobody else has? I dunno I'm skeptical of that.
posted by atoxyl at 11:45 AM on May 8, 2017 [1 favorite]
posted by atoxyl at 11:45 AM on May 8, 2017 [1 favorite]
I should perhaps point out at this juncture that I am both a data scientist and a small rock.
posted by tel3path at 12:56 PM on May 8, 2017 [2 favorites]
posted by tel3path at 12:56 PM on May 8, 2017 [2 favorites]
The secret sauce in terms of this article is the ability to launder things so that multiple campaigns could spend an order of magnitude more than their campaign limit on targeted data. The national limits were set at £700,000 for secondary campaigns: that's 20p per voter. So the Electoral Commission's intent was clearly that any secondary campaign was to be a pretty much pure grassroots thing.
This is why the article's all about the money. And the reason it's forensic (and not drawing all of the conclusions it could) may well be because it's partly intended as a dossier to nudge the police in to action. Here's one of the tweets Carole Cadwalladr announced it with: Full story below. I'm writing to the CPS and Met police tomorrow.…They broke THE LAW.
posted by ambrosen at 1:17 PM on May 8, 2017 [9 favorites]
This is why the article's all about the money. And the reason it's forensic (and not drawing all of the conclusions it could) may well be because it's partly intended as a dossier to nudge the police in to action. Here's one of the tweets Carole Cadwalladr announced it with: Full story below. I'm writing to the CPS and Met police tomorrow.…They broke THE LAW.
posted by ambrosen at 1:17 PM on May 8, 2017 [9 favorites]
There's a lot about this that's circumstantial, and speculative. But this is due to the lack of transparency, with the article showing that the elections commission declined to investigate, and that sounds pretty corrupt by itself. And I can't think of a scientific reason why machine learning techniques can't do the things claimed in the article. It's the obvious way to automate it if you had such large, exploitative data sets.
posted by polymodus at 1:36 PM on May 8, 2017
posted by polymodus at 1:36 PM on May 8, 2017
I'm really unsure about this, because I can't balance out so many conflicting impressions.
First, these sort of companies are indeed much better at snake oil than anything else. It doesn't take long writing about the business tools market to realise how few promises are realised, how much money just vanishes, how many big projects are quietly never talked about a year or so after the press releases. To some extent, this is all by design - to you and me, the idea of sinking a million dollars (or large multiple thereof) on something that promises revolutionary advantage but pans out to bupkiss seems to verge on the criminal. To everyone involved, it keeps the pipelines full of delicious cash, and that's the intended result. A lot - a very great deal indeed - of B2B big-project spending is a shell game, and everyone in the expensive suits is in on it. You wonder why so many big companies are so terrible at living up to their expensive marketing? Wonder no more.
And business data analytics is the creme de la creme here. So easy to sell, so easy to hide the holes, so much motivation by everyone to finesse 'results'. Default position: if you can't check it, it probably isn't true, and you can so rarely check anything these people say.
On the other hand - state intelligence agencies have always spent a huge amount of time on public domain data collection and analysis. Go through the CIA's declassified Cold War era reports, and you'll learn more about the grimy details of, say, oscilloscope manufacturing in Romania than you ever thought existed. Now, you can do this at home. So something like a private sector intelligence agency capability is evolving rapidly, and a lot of people involved are also involved in the state side - but with even less oversight and regulation, and a lot more ability to set up offices all over the place and mix'n'match data sets. It is, in short, very easy to collect and abuse data concerning people anywhere in the world, out of reach of their regulators and laws. Just as it is for so many areas of action online.
On the other other hand, if there is big nasty voter targeting going on, isn't it apparently aimed at the exact demographic of older, information-poor people who really don't get online very much? This looks like a bit of a dichotomy to me.
But my major conclusion is that yes, it's happening; how well it's working is unknown; whatever it is, it's completely opaque to regulators and the public; and that last bit absolutely needs to change. That is, I think, the somewhat-stated conclusion of the Guardian article.
That won't happen in the US, of course. It may happen in Europe, and there's a good track record of that spreading.
In the UK? Nobody knows anything any more.
posted by Devonian at 6:06 PM on May 8, 2017 [2 favorites]
First, these sort of companies are indeed much better at snake oil than anything else. It doesn't take long writing about the business tools market to realise how few promises are realised, how much money just vanishes, how many big projects are quietly never talked about a year or so after the press releases. To some extent, this is all by design - to you and me, the idea of sinking a million dollars (or large multiple thereof) on something that promises revolutionary advantage but pans out to bupkiss seems to verge on the criminal. To everyone involved, it keeps the pipelines full of delicious cash, and that's the intended result. A lot - a very great deal indeed - of B2B big-project spending is a shell game, and everyone in the expensive suits is in on it. You wonder why so many big companies are so terrible at living up to their expensive marketing? Wonder no more.
And business data analytics is the creme de la creme here. So easy to sell, so easy to hide the holes, so much motivation by everyone to finesse 'results'. Default position: if you can't check it, it probably isn't true, and you can so rarely check anything these people say.
On the other hand - state intelligence agencies have always spent a huge amount of time on public domain data collection and analysis. Go through the CIA's declassified Cold War era reports, and you'll learn more about the grimy details of, say, oscilloscope manufacturing in Romania than you ever thought existed. Now, you can do this at home. So something like a private sector intelligence agency capability is evolving rapidly, and a lot of people involved are also involved in the state side - but with even less oversight and regulation, and a lot more ability to set up offices all over the place and mix'n'match data sets. It is, in short, very easy to collect and abuse data concerning people anywhere in the world, out of reach of their regulators and laws. Just as it is for so many areas of action online.
On the other other hand, if there is big nasty voter targeting going on, isn't it apparently aimed at the exact demographic of older, information-poor people who really don't get online very much? This looks like a bit of a dichotomy to me.
But my major conclusion is that yes, it's happening; how well it's working is unknown; whatever it is, it's completely opaque to regulators and the public; and that last bit absolutely needs to change. That is, I think, the somewhat-stated conclusion of the Guardian article.
That won't happen in the US, of course. It may happen in Europe, and there's a good track record of that spreading.
In the UK? Nobody knows anything any more.
posted by Devonian at 6:06 PM on May 8, 2017 [2 favorites]
This is why the article's all about the money. And the reason it's forensic (and not drawing all of the conclusions it could) may well be because it's partly intended as a dossier to nudge the police in to action. Here's one of the tweets Carole Cadwalladr announced it with: Full story below. I'm writing to the CPS and Met police tomorrow.…They broke THE LAW.
I think it makes sense for the article to be about the money, but Cadwalladr notes up top that the second article in her series got people to start investigating that. This article seems to be about tying all of that together more thoroughly by exposing the AggregateIQ ties, but then adds a number of additional suspicious murmurs on top. Cadwalladr brings Trump into the story herself, and I’m really not clear why he even matters beyond also being connected to Mercer & Bannon. Maybe I wasn’t reading closely enough?
Similarly, there’s interesting notes about how the connections of Cambridge Analytica and Palantir to the military-industrial complex, and how it seems strange and concerning that the same software used by the military is also used in running political campaigns, and is also being funded by defense grants, but that idea isn’t actually explored anywhere. It’s mentioned and then passed on, as if that’s sufficient to prove wrongdoing. Like, I find the diagram Cadwalladr posted in the tweets linked by ambrosen interesting but not evocative of any particular thesis beyond “all of these things are related”. I mean, hell, Ted Cruz knows all of these people, and Mercer backed him before Trump. Why isn’t he on the spooky chart?
There’s meat here, and she’s found it, but I’m really staring askance at her editor for not cutting this down…
posted by Going To Maine at 6:55 PM on May 8, 2017 [2 favorites]
I think it makes sense for the article to be about the money, but Cadwalladr notes up top that the second article in her series got people to start investigating that. This article seems to be about tying all of that together more thoroughly by exposing the AggregateIQ ties, but then adds a number of additional suspicious murmurs on top. Cadwalladr brings Trump into the story herself, and I’m really not clear why he even matters beyond also being connected to Mercer & Bannon. Maybe I wasn’t reading closely enough?
Similarly, there’s interesting notes about how the connections of Cambridge Analytica and Palantir to the military-industrial complex, and how it seems strange and concerning that the same software used by the military is also used in running political campaigns, and is also being funded by defense grants, but that idea isn’t actually explored anywhere. It’s mentioned and then passed on, as if that’s sufficient to prove wrongdoing. Like, I find the diagram Cadwalladr posted in the tweets linked by ambrosen interesting but not evocative of any particular thesis beyond “all of these things are related”. I mean, hell, Ted Cruz knows all of these people, and Mercer backed him before Trump. Why isn’t he on the spooky chart?
There’s meat here, and she’s found it, but I’m really staring askance at her editor for not cutting this down…
posted by Going To Maine at 6:55 PM on May 8, 2017 [2 favorites]
Couple more links on Cambridge Analytica: No, Big Data Didn't Win the U.S. Election, The myth that British data scientists won the election for Trump.
posted by TheophileEscargot at 9:01 PM on May 8, 2017
posted by TheophileEscargot at 9:01 PM on May 8, 2017
Also the article says:
If I put the website into the Wayback Machine, the website definitely existed in Jan 10 2016 though it looks pretty cursory. On the basis of "the work of moments" I don't think it's impossible someone found them on the Internet, though as far as I know there's no way to duplicate a historical Google search exactly.
posted by TheophileEscargot at 9:28 PM on May 8, 2017
And to finally answer the question about how Vote Leave found this obscure Canadian company on the other side of the planet, he wrote: “Someone found AIQ [AggregateIQ] on the internet and interviewed them on the phone then told me – let’s go with these guys. They were clearly more competent than any others we’d spoken to in London.”When I search with a date filter aggregateiq canada I don't see a whole lot, but there is a business addreess for "Aggregateiq Data Services Ltd - Victoria, BC" dated 23 Jan 2015 on the first page, and the usual 10 pages of results beyond that. Searching "Aggregate IQ" with a space just gives me a deluge of unrelated results.
The most unfortunate aspect of this – for Dominic Cummings – is that this isn’t credible. It’s the work of moments to put a date filter on Google search and discover that in late 2015 or early 2016, there are no Google hits for “Aggregate IQ”. There is no press coverage. No random mentions. It doesn’t even throw up its website. I have caught Dominic Cummings in what appears to be an alternative fact.
If I put the website into the Wayback Machine, the website definitely existed in Jan 10 2016 though it looks pretty cursory. On the basis of "the work of moments" I don't think it's impossible someone found them on the Internet, though as far as I know there's no way to duplicate a historical Google search exactly.
posted by TheophileEscargot at 9:28 PM on May 8, 2017
fwiw...
-Brexit: victory over the Hollow Men
-How Brexit was won, and the unreasonable effectiveness of physicists
-Brexit in the Multiverse: Dominic Cummings on the Vote Leave campaign
-Dangerous Knowledge and Existential Risk (Dominic Cummings)
-Management by the Unusually Competent
posted by kliuless at 10:19 PM on May 8, 2017 [3 favorites]
-Brexit: victory over the Hollow Men
-How Brexit was won, and the unreasonable effectiveness of physicists
-Brexit in the Multiverse: Dominic Cummings on the Vote Leave campaign
-Dangerous Knowledge and Existential Risk (Dominic Cummings)
-Management by the Unusually Competent
posted by kliuless at 10:19 PM on May 8, 2017 [3 favorites]
Follow the data: does a legal document link Brexit campaigns to US billionaire?
But the Observer has seen a confidential document that provides clear evidence of a link between the two campaigns. More precisely, evidence of a close working relationship between the two data analytics firms employed by the campaigns – AggregateIQ, which Vote Leave hired, and Cambridge Analytica, retained by Leave.EU.posted by Nelson at 10:27 AM on May 15, 2017
« Older "Today we are talking with Sid Meier, who needs no... | State of the Hive, May 11 1917 Newer »
This thread has been archived and is closed to new comments
posted by Going To Maine at 1:37 AM on May 8, 2017 [4 favorites]