Walled Gardens of Self-Serve

Special counsel Robert Mueller’s investigation is “zeroing in on how Russia spread fake and damaging information through social media and is seeking additional evidence from companies like Facebook and Twitter about what happened on their networks,” reports Chris Strohm for Bloomberg Politics. House and Senate investigators are also seeking more information from Facebook.

Our friend and Civicist contributing editor Dave Karpf parses the latest Facebook-Russia-Trump news and says that the investigation is “a big deal” and the company has too much power, but the $100,000 spent by alleged Russian actors on Facebook ads had “virtually no impact on the 2016 election” because “digital advertisements don’t swing national presidential elections.” Even more clearly stated, “Political propaganda does not function like a hypodermic needle, injecting new opinions into the populace. Successful political persuasion, simply put, is really hard.” The real issue, he says, is whether the Trump campaign gave targeting criteria to the Russians to know where to place their ads—that would be evidence of a serious crime, regardless of whether it actually tilted the election.

More evidence of Facebook’s power: Five years after the fact, a group of Facebook research scientists published a paper this past April in PLOS One reporting on the impact of the company’s voter megaphone (the “I voted” button) on turnout in the 2012 general election. Unlike their original study on the 2010, which was published with the provocative title of “A 61-million-person experiment in social influence and political mobilization,” this paper was titled “Social influence and political mobilization: Further evidence from a randomized experiment in the 2012 U.S. presidential election.” And so, when it first came out last spring, it escaped widespread notice—including from me. I had written a piece for Mother Jones magazine in October 2014, taking Facebook to task for not sharing the results of its 2012 experiment, which reached 254 million people, according to this new paper. Imagine hiding a paper called “A 254-million-person experiment in social influence and political mobilization.”

The authors of this new paper focus on a subgroup of 14.5 million people living in 13 states whose voting records they were able to obtain and match precisely to residents’ Facebook pages. Roughly 2/3 of those people had checked in on Facebook on Election Day, they report. The net effect on these users of seeing, either in the form of a banner or in their news feed, that their friends had voted was to increase observed turnout by about a quarter of a percent, or, they estimate, about an extra 270,000 people voting. The actual number was probably higher because they excluded people whose voter records they couldn’t match accurately to specific Facebook accounts. The bottom line: as in 2010, Facebook measurably increased voter turnout. Or, as they write: “a simple message about friends continues to have a substantial effect on civic participation.”

Facebook researchers still haven’t published anything on another experiment they conducted in 2012, putting more hard news in the news feeds of 1.9 million users—who told them in a post-election survey that this had caused them to be significantly more likely to vote. As I noted on Twitter, “as in 2010, we have to trust its engineers and researchers when they say the ‘I voted’ prompt was applied in a neutral way.”

Adam Sharp, Twitter’s former director of media partnerships, tells CNN Money that he has questions about Facebook’s reported sale of $100,000 in ads to Russians trying to influence the 2016 election. “”If the interaction was limited to the walled garden of self-serve, Facebook can make the case that it was just the machine talking to the Russians. However, there are still questions,” he added. “Why weren’t there preventative measures in place to alert Facebook to the fact that this might be political advertising? And why didn’t this buy trigger an account executive to reach out to the buyer?” He also points out that if those ads succeeded in getting their viewers to “like” or engage with a post, they can continue targeting those individuals and their friends. Those users may then participate in spreading fake news or misinformation by sharing it on their timeline, in which case the advertising is no longer necessary.” As soon as they get page likes and engagements, they can use that to seed a fan-out of trolling,” Sharp explained, adding that Facebook would need “a more finessed approach” to prevent such trolling in the future.

ProPublica investigators Julia Angwin and Ariana Tobin discovered that until they asked Facebook about it, the company’s advertising platform sold promoted posts to people interested in topics like “Jew hater” or “How to burn Jews.” Algorithms! As they note, after Charlottesville, CEO Mark Zuckerberg declared that “there is no place for hate in our community.” But apparently no one told the advertising team. ProPublica also revealed last year that it was easy to buy ads that targeted people by race. One wonders if Zuckerberg shouldn’t just write ProPublica a large check for all the work it’s doing on the company’s behalf. Facebook’s response: “As people fill in their education or employer on their profile, we have found a small percentage of people who have entered offensive responses, in violation of our policies. ProPublica surfaced that these offensive education and employer fields were showing up in our ads interface as targetable audiences for campaigns. We immediately removed them. “

On a more positive note, junk news expert Craig Silverman writes for BuzzFeed that Facebook’s new collaboration with third-party fact checking sites should start paying real dividends, not because users see stories marked as “disputed,” but because “Facebook has found a way to create a reliable source of expertly-annotated data it can mine to create smarter artificial intelligence. Along with spotting completely false stories, the data may also prove useful in helping the platform identify common characteristics of low-quality websites. This is good news, and ultimately far more impactful than a label being shown to users.”

The Sunlight Foundation’s Alex Howard and John Wonderlich say that it’s time to require big online platforms to disclose who is spending what on political ads on their sites, or to make the purchasers of those ads make such disclosures.

Crypto-wars, continued: Former CIA deputy director Michael Morrell resigned his non-resident fellowship at the Belfer Center at Harvard’s Kennedy School upon the invitation of Chelsea Manning, who he calls “a convicted felon and leaker of classified information” to be a visiting fellow at its School of Politics.

In addition, current CIA director Mike Pompeo wrote a letter to the head of intelligence and defense projects at the Kennedy School, Rolf Mowatt-Larssen, arguing that “American traitor” Manning “stands against everything the brave men and women I serve alongside stand for” and informing him that he was withdrawing from an upcoming school forum.

Shortly after midnight this morning, the Kennedy School’s Dean, Douglas Elmendorf, issued a statement that first explained that Manning was simply invited spend a day at the school interacting with students and faculty, that the school has hundreds of such fellows at any time, that it often invites people with controversial views, and that the title “visiting fellow” isn’t intended to convey any honor. But despite all that, he declared that the “perceived honor” of being called a visiting fellow was problematic to “many people” and therefore he rescinded the title while maintaining the invitation to her to visit.

At 3am this morning, Manning tweeted in response: “honored to be 1st disinvited trans woman visiting @harvard fellow they chill marginalized voices under @cia pressure #WeGotThis.”

Among Harvard’s other visiting fellows not honored by not having their honorific rescinded: Corey Lewandowski, the one-time Trump campaign manager who has been charged with misdemeanor battery for manhandling a reporter. Ask not what you can do…

Credit due: Writing about the Equifax data leak, Open Markets fellow Matthew Stoller reminds us that the power of credit bureaus has a long history in America, shaped by relatively recent (and weak) laws regulating the use of credit data.

Independent security expert Bruce Schneier writes for CNN that the only solution to the Equifax security breach is tougher government regulation. He also notes, soberly, that “Yes, it’s a huge black eye for the company—this week. Soon, another company will have suffered a massive data breach and few will remember Equifax’s problem. Does anyone remember last year when Yahoo admitted that it exposed personal information of a billion users in 2013 and another half billion in 2014?”

Google is facing a class action lawsuit filed by three former female employees alleging gender-based pay discrimination, the Associated Press reports.

Opposition watch: The Mill Valley Community Action Network, one of the many new “resistance” groups, has launched Airlift.fund, aiming to help progressive political donors move their funds more strategically to front-line groups, using an innovative approach to bundling. As Amanda Terkel reports, the initiative is rooted in members’ frustration with endless fundraising emails from traditional groups who spend the bulk of their money on TV ads, and is deliberately playing on its members’ comfort with taking risks on start-ups.

Your moment of zen: Goodbye, Cassini.