By Marina Bart (formerly aab) a writer and former public relations consultant, who thinks and writes about many things, including political economy, culture and communication
It has long been the case that “Big Data” has been treated as a magical, unstoppable force that will reap power and profits for those who can channel it effectively. In the 2016 Presidential election, the Clinton campaign and Democratic National Committee relied heavily on Ada (named after Lord Byron’s mathematician daughter, a perfect identitarian for Clintonian Democrats, combining as she does into one symbolic person aristocratic status, the creative class, feminism and computing). Ada was the Democrats’ attempt at a Big Data Election-Winning Machine, apparently created for it in secret in a dark cave at the top of a mountain by Eric Schmidt and unknown coding slaves who were probably killed as soon as Eric carried his prize down the mountain to Hillary’s waiting arms.
That last part is made up.
But the Democrats did have Ada, which only top aides were allowed to use or even see. Very little is known about Ada, because (spoiler alert) Clinton lost and the brain trust leading the party (if you can call it that) didn’t want anyone to focus on their incompetence and wasteful spending because RUSSIA. Ada said Wisconsin was a safe state. Ada said paying Jay Z to perform would win Ohio. Ada failed, along with Clinton.
Yet now there are rumblings that there is a REAL Death Star, a Big Data system perfect in its design, malevolent in its intent, and all-powerful in its capacity to segment and manipulate every human being on Earth. (No, not the NSA – how silly of you to worry about an arm of the government!) It’s Cambridge Analytica, of which Robert Mercer, right wing hedge fund billionaire, is a major investor. That Robert Mercer, who backed Donald Trump, the crazy-haired real estate hair and game show host who won the 2016 Presidential election.
The thinking is, now that Trump has access to all that otherwise completely benign NSA data, he and Mercer can conspire to feed all of America into the gaping maw of this algorithmic monster, and manipulate otherwise good-thinking Americans into…I’m not sure, exactly. Trump already won the Presidency. The Republican Party – the party notably less aligned with Silicon Valley, although that is rapidly changing because Silicon Valley boys know to go where the money is – is already so dominant, it controls not only every branch of the Federal Government,1 but such a large percentage of the states it only needs one or two more to call a Constitutional Convention. It did all of this without an Election-Winning Death Star.
This suggests a number of questions:
- How much should we fear Big Data?
- How much should we fear Cambridge Analytica?
- Are we on the verge having of a new, uniquely powerful propaganda tool?
- And the secret question, the one unsaid in most of these pieces: Can the Democrats get their hands on it to win back power?
The short answers are:
- A lot
- No more or less than every other Big Data operation
All this extensive data gathering and mining of people without their full knowledge and consent is bad. It’s bad when the NSA does it. It’s bad when Facebook does it. It’s bad when Cambridge Analytica does it. There is the potential for tremendous harm. There is also the potential (which is, to some degree, already occurring) for massive manipulation of people’s emotions in new ways. Targeted segmentation and messaging works for many purposes. Here’s a primer.
However, there are numerous potholes, landmines and rabid creatures lining the path to Big Data Death Star effectiveness. The first is the simplest: good, old-fashioned human error. GIGO (Garbage In, Garbage Out.) That was demonstrably a problem with Ada. Reports emerged after the election that grassroots organizers and local politicians were aware of and gathering information that there were serious problems in some of the crucial states that Clinton eventually lost. But their carefully recorded data and desperate phone calls were ignored, rather than be included in what Ada considered in her 400,000 simulations a day. Ada’s “advice” was filtered by what her masters allowed her to know. We have no way to assess if Ada really failed, or if she was just misused by Robby Mook and the rest of the campaign’s leadership. (Also, as a reminder, this was not Putin’s fault.)
Then there’s the other big, known problem with Big Data: it’s too big. The underlying assumption of the Big Data Death Star Election-winning machine is that you have all the necessary data on every voter, which you then slice into incredibly thin segments, test different messaging in all these different segments, watch for outcomes and responses indicating that messaging works, then hammer each very specific message to each very different segment to obtain your desired result. (Are you a native-born, Spanish-speaking Mormon man who watches Jane the Virgin, owns a gun, secretly identifies as queer, has a degree in business administration, and plays paintball? There’s a segment for that!)
But that assumes (rather like a can opener), that you have a way to sort and keep track of all that data, and also have accurate algorithms to evaluate that data at a very granular level, and also have enough skilled experts who can understand all these different segments well enough to craft messages that will drive all these different segments composed of living, breathing humans to do what you want them to do – in this case, win a national election.
Yet Facebook, a behemoth designed to lure people into an addictively faux-intimate walled garden where their affinities can be identified, captured and controlled, along with all their purchasing information, daily activities, current thinking and mood, can’t even get its advertising and traffic metrics right. This is Facebook’s bread-and-butter: knowing its users well enough to push targeted advertising and content to them, in the hope this will be more effective than big, dumb (expensive, ignored) TV advertising. If it can’t even do something this basic and fundamental to its economic model correctly, does it seem likely they can actually find all those queer, Latinx, Mormon paintball players and understand them well enough to make going to the trouble and expense of doing the segmenting and message crafting worthwhile?
If Facebook2 can’t do it yet, when it has its users under a constant microscope, poking and prodding and dribbling out sugar syrup to watch them scurry, why would Cambridge Analytica, with its much less intimate and direct connection to user data, be more successful, even if given access to all the NSA data? We already know the NSA itself is struggling to manage its data hoard.
When Yves and I started talking about this, she mentioned the successful direct mail campaigns that Karl Rove made his bones on, which were purported to be highly segmented. Her thought was that if you can combine that with captured consumers who won’t throw out the mailer, and the ability to do brute force, constant, cheap a/b testing of messages, you don’t actually need to have clever people doing the messaging. Rather like monkeys typing Shakespeare, perhaps you could message test until you stumble across what works, then flood social media with it to get the results you want.
Remember, however, that Ada was running 400,000 simulations a day. Doing live a/b testing isn’t going to be any more effective if you don’t have processes in place to protect against bias and human error at every step of the gathering, evaluating, creating, dissemination, and assessment process. You also need the right tools to sort and evaluate the data, assess which pieces to prioritize for your specific purpose, and measure the effectiveness of your messages. There is nothing useful about testing numerous messages if you have no way to know which, if any, is working. There is no evidence at all that any of this exists yet. Manipulating someone’s mood is much easier than getting them to go to the polls, which delivers no immediate gratification and is often unpleasant. The bigger the data set, the more these problems compound.
With sales, this is not a problem. You want your targets to buy a product. If they buy the product, you know the message worked. You don’t need to develop a complicated system to evaluate message effectiveness. If you don’t need to successfully persuade the majority of your targets, having a high messaging failure rate doesn’t matter. The reason scam emails persist is that they’re so cheap to distribute, it doesn’t matter if you have a 99.9999999% failure rate. You only need a handful of suckers out of millions to make the scam profitable for the scam operator.
But the fearmongering around Cambridge Analytica is about the potential to manipulate large numbers of people into all believing the same false, emotionally inflammatory thing, so they will all move in the same direction. That is what propaganda is for: getting large numbers of people to do something that is not in either their communal or individual interests. If you don’t need that outcome, you don’t have to use such an ugly tool.
There is no question that modern social media facilitates highly segmented marketing. There is no question that political campaigns can benefit from this. Figuring out who might be receptive to your candidate and their policies, where they vote, and motivating them to go to the polls is fundamental campaign work. But that is not at all the same thing as manipulating people into voting against their interests, which is presumably what is feared (and possibly secretly hoped for) by the fretful Democrats. There is no evidence Cambridge Analytica did any psychological manipulations for Trump.
I’m not saying it’s impossible for Big Data highly segmented psychological manipulation to ever work. But it isn’t happening now; there’s no evidence it will work in the near future; there are many, many obstacles to overcome; and there are two very basic reasons why it cannot be the secret weapon I suspect the Democrats long for.
The most basic one is that voting is not the same as buying stuff. There is no direct connection between casting a vote and getting anything in return, not even the momentary pleasure of buying a candy bar. The Democrats seem to be fixated on the idea that they just need to find 100,000 more votes in the correct states to begin winning. That overlooks the problem that to govern, they can’t just win the presidency. They need to win back the Congress, governorships, and state legislatures. The insiders may not care if they regain power, as their own lives will be comfortable and well-compensated regardless. They are clearly hoping that two years of Trump (if they can’t remove him from power) will drive enough voters back to their cold, exploitative, all-talk-no-action embrace to at least show the results necessary to their owners to get the money to attempt to retake the White House in 2020.
But the “we suck less, vote for us” strategy hasn’t been working so far. Driving weakly motivated, weakly aligned voters to the polls is hard. Yet that is what the Democrats appear to be counting on: finding more people who wouldn’t vote for them this year while being told Trump was Hitler about to bring on Armageddon to come out for them in 2018 for nameless Congresscritters and Senators who have done nothing for them, and then have them march to the polls in 2020 for someone like Cory Booker, with his long history of praising and working with Betsy DeVos – all without offering any of the policies the voters need and have been clamoring for. They seem to think more fear, coupled with some Bill Clinton style small ball policies targeted like a laser at correctly identified segments will do the trick – with a population where half are already impoverished.
I mean, it could work. It’s possible.
But generally speaking, the point of targeted marketing is to identify people who want what you have to offer them. If they don’t want what you’re selling, it doesn’t actually help that much. The Democrats apparently intend to not do that.
The other big reason is that most people talking about this are confusing marketing with public relations, and treating all public relations as if it is propaganda, which is simply wrong. Not morally wrong. Factually wrong. These are different disciplines, with different goals. Those differing goals hinge in part on human psychology, and how that impacts the process of persuasion under different conditions, in different circumstances. For example: a direct sale is a process of intimate manipulation and control. You use tactics to get your target to start agreeing with you – it doesn’t matter about what. Once they start agreeing with you, you use a combination of their growing emotional attachment and deference to you, and the information you are extracting from them, to dominate them while appealing to their self-interest and emotions to close the sale. One of the greatest strengths of phone sales back in the day was that the first hurdle was just keeping your targets on the phone. If they were so lonely, compliant or polite that they wouldn’t hang up, you were halfway home.
But public relations is not about that degree of control. It focuses on persuasion. Persuasion is softer. You use different psychological tactics, as well as strategies and channels, because you do not have the same degree of intimacy or control. Much of the public relations industry doesn’t actually do effective public relations, because clients don’t know what to look for or what benchmarks matter. So it’s easy for agencies to point at meaningless “media hits” or parties hosted, and two or three obscenely expensive contract years later, the client moves on, never having achieved their desired goals.
Propaganda, while technically a public relations discipline, is a different beast still. Propaganda, at its core, is about stampeding the herd – using intensely emotive messaging, usually fear or hatred based, to manipulate masses of people. While some argue that propaganda isn’t necessarily negative or deceitful, I think the common usage is most relevant to this discussion. Propaganda tends to be dishonest because, as mentioned earlier, the intent is to drive people to do something that is against their interest generally and individually. That is why you use the tool of propaganda to achieve your end. If you didn’t need people to act irrationally against their interest, you wouldn’t resort to propaganda.
Which brings us to one more key reason why the Big Data Death Star is not on the near-term horizon and may never be used: the psychological manipulation underlying propaganda is diametrically opposed to the psychological manipulation underlying targeted marketing. Propaganda works best when fear and/or hatred is employed to not just drive individuals to action, but to drive them together, and then, once together, all marching forward together.
Propaganda tells you the Germans want to eat your children, that the enemy is subhuman, that Muslims hate us for our freedoms, that the Communists are evildoers coming to kill us in our beds. Propaganda systematically dehumanizes people the propagandist wants you to give up your life to kill for him, and falsely proclaims the enemy both lesser than you and greater than you – less worthy of life, yet inhumanly dangerous, whether they’re smarter, crueler, richer, less moral – whatever it is, you can be sure it presents an existential threat to those being propagandized. Once they believe this, getting them to work together to do what the propagandist wants is the easy part.
Despite what the conservatives and neoliberals claim, humans did not evolve because of their competitive nature. They evolved because of their cooperative nature. And the best way to get them to cooperate with one another is out of fear or hatred of an existential threat.
Therefore, propaganda is most effective when everyone being propagandized is getting the same message, and believes the same thing. If you disseminate different messages to different segments to motivate them, you run into the problem of disagreement once you have herded them up. You can’t have infighting over who the enemy is or what their goals are; that breaks up the artificially created sense of unity that is absolutely necessary if you want your population to go die in great numbers so you can get your hands on a couple of oil fields.
So if it is unuseful to distribute different messages to different thin segments via social media, a major rationale for the Big Data Death Star just goes away. Which brings us to the real propaganda machine, being used right now, very effectively, by the Democrats.
If targeted, differentiated, segmented messaging is ineffective as a propaganda tool, what might work? What has historically worked? If you want ONE consistent message to reach great numbers of people, you want to control the main ways people receive information, particularly pre-existing, trusted sources. And our current corporate media landscape is perfect for this task. It is highly concentrated; just six corporations control at least 90% of the media outlets in the entire country3. Or you could say our media is controlled by fifteen people, all of them billionaires with shared class interests. Who needs the Creel Committee when a Hamptons dinner party will do the job?
Once you have proper control of the channels, you just need a simple, terrifying message that can be pumped out non-stop by thousands of people, until the victims of this process accept and parrot it themselves, acting as messaging agents for their manipulators, reinforcing the notion that it must be true because “everybody” thinks it. And the best possible simple, terrifying message would be an evergreen: one that has worked before, that millions of people already think was true once, and are therefore primed to think is true again. That will give you a nice jump start on acceptance and voluntary dissemination in our distracted world.
Enter Vladimir Putin, leader of Russia, former KGB agent and supposed slayer of Clintons.
The propaganda surrounding “Russia stole the election and installed Trump as a puppet” is astoundingly fierce, even as it is empty of proof. Naked Capitalism has already covered this extensively, so I will just summarize: even the NSA wouldn’t say so, there’s no evidence of it, and there is literally no physical way they could have. Part of the current propaganda push is to pretend that Russia had a hand in releasing the DNC and Podesta emails, none of which have been disproved, and many have been validated. But a) there is still no solid proof Russia was involved in leaking the emails and b) even if they did, they are real communications that prove real actions and beliefs on the part of Hillary Clinton, her campaign, the DNC, and Democratic insiders generally, that voters recoiled from. Voters rejected Hillary Clinton for who she was, what she, her husband, Barack Obama and the rest of Democratic leadership had done, and what it was proposing to do. It wasn’t Russia.
And yet, despite the complete lack of facts, polling shows a rising number of Americans are worried about Donald Trump’s relationship with Russia. That is good, sound propaganda best practices in action. Facebook is useful, but only for its capture of attention, capacity to block out conflicting information, and tendency to reinforce group think (which is by design, because that delivers more users, more often, voluntarily offering up their data by being lulled into a false sense of community). There is no targeted messaging going on; no clever, secretive use of magic dashboards is necessary. All you need is everybody at the New York Times, and NBC, and Newsweek, and CNN, and Facebook all singing from the same hymnal, loudly and clearly; something familiar, something frightening, something to drive you to seek out your fellows to stop evildoers from eating your children and killing you in your bed.
We’ve got ourselves a good, old fashioned Red Scare. It’s a fine vintage, cellar-aged yet full strength. That is how propaganda works. You don’t need complex technology or massive data collection to achieve it. Why go to a lot of expense and energy to develop a system to destroy something with millions of hand grenades, if you already have an available and easy to launch bomb that will do the job? That is why it doesn’t matter whether Cambridge Analytica is in the hands of the Democrats or the Republicans. The Democrats actually have the upper hand right now in terms of propaganda delivery,4 because they’re the party of “communication,” these concentrated companies tend to be on the Democratically-controlled coasts, and media billionaires are mad at Trump for blocking TPP and other such neoliberal rent extraction schemes they so richly enjoy.
And yet, the Democrats’ dilemma remains. They may or may not be able to work with the CIA and the neoliberal faction of the Republican party to ramp up a war with Russia and tame Trump into irrelevance. They may even get him impeached. But since they are working with the neoliberal Republicans to go after Trump, removing Trump doesn’t do much for them electorally. They aren’t using this Red Scare to demonize the entire Republican Party, just Trump. (They seem to be intentionally rehabilitating George W. Bush, which is astounding.)
The Democrats’ brand identity is not military security. Their brand identity is economic security. If they succeed in terrifying Americans en masse, they are very unlikely to see those terrified Americans turn to them in the next election over this. Because while propaganda plays a role in elections, people still are unlikely to go to the polls to stop something they don’t like. If you want people to vote for you, you have to give them something to vote for. That was the genius of the Obama 2008 campaign. People thought they were voting for change – not just change from the Republicans, but change from the Clintons, change from the status quo. That trick is unlikely to work twice, whatever the Democrats may think.
People still want change. That is why Trump won. But the Democratic Party seems determined to make sure every single voter in the country knows they don’t want change. There is no Big Data Death Star that can help them with that problem.
Note: This is intended as the first in a series of pieces addressing persuasion practices and their relationship to political and economic issues, including the often poorly understood yet important differences between public relations and propaganda. Feel free to let me know in the comments section what aspects of that you would like to learn about in greater detail
1Yes, the Supreme Court is technically tied. For now.
2Even more algorithm problems at Facebook: https://www.wsj.com/articles/facebooks-trending-feature-exhibits-flaws-under-new-algorithm-1473176652
3This piece is a few years old. While the data is fundamentally accurate, Comcast has replaced GE as a major media owner.
4It’s interesting that people allied with the Democrats are so worried about a propaganda machine that doesn’t exist, instead of the real one they themselves are currently using. I wonder why.