Internal memos show Facebook knew about flaws and did nothing

Note: This was originally published as the daily newsletter for the Columbia Journalism Review, where I am the chief digital writer

In 2018, Mark Zuckerberg, co-founder and chief executive of Facebook, said that the company was rolling out a significant change to the algorithm that governs its News Feed, in an attempt to encourage more users to interact with content posted by their friends and family, rather than content from professional sources such as news publishers and other brands. One of the reasons for doing this, Zuckerberg said, was a growing body of research showing that consuming mostly content from brands and publishers was not good for the well-being of users. However, according to a report from the Wall Street Journal published on Wednesday, the algorithm change didn’t improve the well-being of users — in fact, it actually had the opposite effect. Internal memos describe how the company’s own researchers said the changes were making the News Feed “an angrier place” by encouraging outrage and sensationalism. And when they suggested changes, Zuckerberg turned them down because they would decrease engagement, the Journal says.

The Facebook researchers “discovered that publishers and political parties were reorienting their posts toward outrage and sensationalism,” the Journal report says, because this generated higher levels of comments and reactions, which the platform uses as indications that a post is successful and should be amplified. “Our approach has had unhealthy side effects on important slices of public content, such as politics and news,” data scientists at the company said in memos that the newspaper was able to read. “This is an increasing liability. Misinformation, toxicity, and violent content are inordinately prevalent among reshares.” These researchers worked on a number of potential changes that they hoped might ameliorate the algorithm’s tendency to reward outrage, the Journal says, but memos show Zuckerberg resisted many of these solutions because he was worried they might lead to people spending less time interacting with content on the platform.

The emails and memos the newspaper quotes from are part of what it calls “an extensive array of internal company communications” that it gained access to (although it’s not clear how), which so far have produced three investigative pieces on the company’s practices, of which the News Feed story is the third. The first, from reporter Jeff Horowitz, described a little-known system within the company that allowed VIPs to avoid any repercussions for breaching the platform’s terms of service. The program, known as XCheck (pronounced cross check) allowed celebrities, politicians, athletes, and other “influencers” to post whatever they wanted, with no consequences. Although an internal Facebook report seen by the Journal referred to “a select few members” as having this ability, the story says that as of last year, close to 6 million people were covered by the XCheck program.

Continue reading “Internal memos show Facebook knew about flaws and did nothing”

Backcountry camping trip to McEwen Lake

As we have done the past several years, we did a big backcountry camping trip in September, right around my birthday, with our neighbours and friends Marc and Kris. This year it was a little harder to find a campsite than it has been in previous years, because COVID restrictions loosened up and everyone wanted to be out in the woods apparently. But we found a great site on a little lake called McEwen, just northwest of Carnarvon, and we picked a great weekend to do it because the weather was fantastic for camping — a big rainstorm went by just as we were getting ready to put in, but we didn’t even get wet, and it was essentially warm and sunny the whole time we were there. It did get a little cool at night — like around 9 degrees Celsius — but we were fine.

We put into Margaret Lake, which is long and thin, but we crossed it sideways and then had a fairly long and rocky portage (about 300 metres or so) uphill to the next put-in. Unfortunately, it had rained not that long before we got there, so the trail was quite muddy, but we made it without any mishaps. We paddled out of a swamp and into Dan Lake, and stopped partway up for a lunch break at one of the campsites. We were thinking we would have another portage to get into McEwen, but as it turned out the water was still high enough that we just poled the canoes through a narrow swampy section and we were in McEwen. After that, it was only a 20-minute or so paddle and we made it to our campsite.

Continue reading “Backcountry camping trip to McEwen Lake”

How a story about ivermectin and hospital beds went wrong

Note: This was originally published as the daily newsletter for the Columbia Journalism Review, where I am the chief digital writer

Over the long weekend, one of the trending topics on a number of social-media platforms was a news item from Oklahoma with some terrifying information: it said that so many people were in hospital due to overdoses of ivermectin—a drug originally designed for horses, which some anti-vaccine sources have been promoting (incorrectly) as a defense against COVID—that there was no room in the intensive-care units for other patients, including those with gunshot wounds. The story, from a local news outlet called KFOR, contained quotes from an interview with Dr. Jason McElyea, a local physician, and the article quickly got picked up by a number of national outlets, including Rolling Stone magazine, the Guardian in the UK, Newsweek magazine, and Business Insider. A producer for MSNBC repeated the claim on Twitter (although she later deleted it), and so did the Rachel Maddow show.

Not long afterward, the story started to spring some major holes. As detailed on Twitter by Drew Holden—a public-affairs consultant in Washington, DC and former assistant to a Republican congressman—and by Scott Alexander on his popular blog, Astral Codex Ten, the first sign that all was not right came with a statement from a large Oklahoma hospital, which said that there was no bed shortage due to ivermectin overdoses, and that the doctor quoted in the KFOR report didn’t work there. Others pointed out that in his original interview with the Oklahoma news outlet, McElyea hadn’t said anything about ivermectin cases crowding out other patients. He mentioned that there had been some ivermectin overdoses, and he said that beds were scarce, but the connection between the two seemed to be a leap that the news outlet and subsequent reports had added.

This was all it took for the story to catch fire with right-wing Twitter trolls and other conservative groups, as yet another example of the mainstream media‘s tendency to make up news stories to either make citizens of rural areas look stupid, or to overstate the risk of non-mainstream COVID treatments. Many latched on to the tweet from the Maddow account, and used it as evidence that no one fact-checks their statements any more, especially when they serve the purpose of making right-wing anti-vaxers and COVID denialists look bad. Others used the Rolling Stone story as an excuse to revisit the magazine’s infamous investigative story from 2014 on an alleged rape at the University of Virginia, which collapsed after the single source it was based on retracted some of her statements.

Continue reading “How a story about ivermectin and hospital beds went wrong”

Facebook plans to show users even less political news

Note: This was originally published as the daily newsletter for the Columbia Journalism Review, where I am the chief digital writer

In February, Facebook announced an experiment designed to test how much political news users wanted in their News Feeds, a test that would remove that kind of content for a small group of users in the US, Canada, Brazil, and Indonesia, and then survey those users for their reactions to the removal. According to an update published on Tuesday, the company saw what it called “positive results” from the experiment, and as a result is now expanding the test to cover users in Ireland, Sweden, Spain, and Costa Rica. In addition, Facebook said it is tweaking the way it measures user behavior when interacting with political content: “We’ve learned that some engagement signals can better indicate what posts people find more valuable than others,” the company said. Instead of looking only at whether someone is likely to comment on or share a political post, Facebook said it will now put more emphasis on newer signals, such as how likely a person is to provide negative feedback about a political post or topic that happens to show up in their News Feed.

Facebook’s announcement is just the latest in a long series of algorithm changes aimed at de-emphasizing not just political news but professional news sources in general. Mark Zuckerberg, Facebook’s chief executive, said in 2018 the social network would be changing the News Feed to prioritize content shared by a user’s friends and family, rather than content from professional publishers and brands. “I’m changing the goal I give our product teams from focusing on helping you find relevant content to helping you have more meaningful social interactions,” he said. Some hoped the changes might spur some media companies to stop relying on Facebook for their traffic, but it’s unclear whether that has happened to any significant extent. Meanwhile, some disinformation researchers have pointed out that Facebook’s prioritization of content from friends and family—including a focus on promoting the use of private groups—may actually have made the problem worse.

Whenever Facebook tweaks its news algorithm, media outlets and publishers around the world tend to hold their breath, because even a small change in such a large and influential platform can impact a publisher’s traffic significantly. When the company made a similar tweak to its algorithms designed to down-rank professional news content in favor of personal posts, some publishers saw traffic declines as high as 30 percent. According to Facebook’s note, the latest changes “will affect public affairs content more broadly [and] publishers may see an impact on their traffic.” The company didn’t say how big an impact they might see, but did add it is planning a “gradual and methodical rollout” for its experiment, and expects to announce further expansions in the coming months.

Continue reading “Facebook plans to show users even less political news”