So far in this project I’ve been annotating traditional academic sources. These sources explore methods of machine learning, Natural Language Processing, sentiment analysis, and the tools used to mine social media for research purposes. But the literature hasn’t kept pace with the news, and social media data is being used for things other than academic research. Like maybe stealing elections.
Here begins a series of three annotations of investigative reports by London’s Channel 4 News. These are video stories about Cambridge Analytica and its methods and role in political campaigns in the U.S., Africa, Europe and beyond.
These are of course non-traditional annotations. But I consider the source credible, and given the subject (use of data to win elections) important to include. I’m writing this bibliography as the final project in IS452, Foundations of Information Processing, at the University of Illinois iSchool. But I plan to pursue this topic beyond, as it represents many overlapping areas of information science, political communications, social media, technology, and journalism. My research and professional work touches all these areas.
So to the story: Here we have the first report by the team at Channel 4 News that somehow tricked the supposedly brilliant minds at Cambridge Analytica into speaking candidly to a potential client. Unknown to them, the client was fictional and Channel 4 News was recording the entire series of four meetings. Which I regard as completely fair and important, because this is a story that has to be aired.
A more formal annotation for the bibliography project follows the video. Enjoy:
Channel 4 News. “Cambridge Analytica: Whistleblower reveals data grab of 50 million Facebook profiles.” YouTube, March 17, 2018, https://youtu.be/zb6-xz-geH4
In this video report, former Cambridge Analytica research director Chris Wylie explains how he company mined and used the social media data of 50 million Americans to target prospective voters with messages tailored to their fears and motivations. Wylie worked for Steve Bannon, who sought profile information on Facebook users for use in the Trump presidential campaign. Says Wylie, “Steve wanted weapons for his culture war…to change the culture of America.”
A great deal of research had been conducted over the past decade on analysis of social media data. Machine learning techniques have been developed to extract meaningful information from this data, such as sentiment analysis which determines positive and negative attitudes toward subjects and entities. Comparative opinion mining seeks to identify user preferences among alternative choices. Social network groupings can be mapped and key influencers identified within each group. A wide variety of demographic information and aspects of personality can be derived from data accessible from the Facebook Graph API. Content marketing can make use of this data to develop messages calibrated to influence specific users based on their profiles and the influence they have in their online social networks. Other academic research has determine that a small amount of leverage in the right places can have significant effects in voting behavior.
Bannon’s idea was to leverage academic research on social media personality profiling, and replicate it on a massive scale for messaging operations in the 2016 campaign. Under Bannon as vice president of Cambridge Analytica, the company contracted with Aleksandr Kogan, a data scientist at the University of Cambridge who developed a Facebook app. The app offered to pay Facebook users to fill out a personality survey. These users agreed to share not just their survey responses but also their profile information. Even more crucially for Cambridge Analytica, the app also crawled each user’s contacts and mined the profiles of all their Facebook friends: an average of 300 additional people per app user. The reach was so extensive that Facebook later clarified that while the original estimate was that 50 million profiles were mined, in reality it was closer to 80 million profiles.
Tens of millions of Facebook records could thus be pulled within a short time, with no restrictions or limits from Facebook. Many news reports terms this a “breach,” but the term is inaccurate given the accessibility of the data via the Facebook Graph API. Aleksandr Kogan has been accused of violating Facebook’s terms of service, a claim he disputes. Facebook essentially exposed the profile data to researchers, and asked them to delete the data set when their research was complete. But Cambridge Analytica received a copy of the data, and it remains unclear if it was ever erased.
Facebook has since suspended Kogan, Cambridge Analytica, and Chris Wylie from the platform as the investigation continues.
For their part, Cambridge Analytica claims that the data wasn’t useful and made no difference in the presidential election. An assertion Chris Wylie disputes:
“It wasn’t some tiny pilot project, it was the core of what Cambridge Analytica became. It allowed us to move into the hearts and minds of America voters in a way that had never been done before.”
Wylie’s assertion echos Cambridge Analytica claims during the 2016 election. Cambridge Analytica CEO Alexander NIX spoke openly about how the company was able to “microtarget” voters and deliver custom messages that resonate with their psychographic profile. For example, “someone who is neurotic, is someone who is quite emotional, and might respond in this case to stimulus of fear.”
After the story broke in March of this year, Cambridge Analytica has renounced the effectiveness of psychographic techniques to influence voters. The company claims that Aleksandr Kogan is responsible for any violation of Facebook policy, or breach of Facebook users’ privacy. But Wylie says Cambridge Analytica did just what it claimed during the election:
“If I am studying you, and I have enough information about you, because you curated your entire self online and I capture that, I can anticipate what are your mental vulnerabilities. What cognitive biases might you display in certain situations? And I can exploit that.”
Wylie acknowledges that political campaigns have always attempted to persuade people to think a certain way and vote for their candidates. But he asserts there’s a difference between persuasion and manipulation. “This gets at the heart of why is it you’re taking this psychological approach,” he says. “Why do you need to study neuroticism in people? What’s going to make them fearful?”
Wylie himself was at the heart of whatever Cambridge Analytica did in the 2016 U.S. election. As Cambridge Analytica’s research director he had a hand in all of it. He now claims he was naive and is coming forward to own up to his mistakes.
Adventures in Information Science 