An algorithm reads a computer
Illustration by Cathryn Virginia

The Robots Microsoft Assigned to Do My Job Can't Do It

MSN robots can't make fine discriminations for a startlingly vast audience of readers. What they can do, and why they're asked to do it, is unclear.
July 9, 2020, 3:29pm

The most important skill in journalism is not fucking up, and in early June, MSN’s new programming algorithm fucked up. MSN UK published a story about the British pop group Little Mix, mistaking one of its mixed-race singers for the other. It was an error that would have gone unnoticed 99 times out of 100 prior to the installation of AI programming, but trended vigorously because it came just days after the Guardian reported that 50 editors had lost their jobs to make way for this algorithm, which would handle the bulk of MSN’s programming going forward.


I was one of the editors who was let go, and it was bittersweet to see reporters at outlets like this one finally acknowledge the existence of MSN, a behemoth news-redistribution shop that trails only Google sites for total traffic, and is ahead of Facebook—though I never thought there was any mystery about why it was ignored. MSN had long since abandoned creating its own content in favor of licensing news from premium partners at the New York Times, the Wall Street Journal, the Washington Post, et. al. when not receiving it free from smaller operations that act as remoras via affiliate links in articles—a category that includes hundreds, if not thousands, of partners around the globe. If you use MSN at all, it’s likely indirectly, posting articles you found via a search or your default browser homepage:

My job was to create the MSN Money section every morning from these stories and monitor it, a job I called “fantasy newspaper” because of its relation to the pick-and-choose games of fantasy sports. I was good at it. I fucked up the same way the algorithm did when it misidentified the Little Mix member more times than I can count, but I either never got caught or, more likely, caught it myself. One thing about creating a morning news section by yourself for a site that gets more than six billion hits per year (and that’s just MSN Money, in America), is that you tend to be a bit peripatetic; I checked and rechecked my work repeatedly like a news junkie and full-blown digital addict. Imagine a lone old-time phone operator in a room where the ringing never stopped, and that’s pretty much what it felt like. It was a volume business with a skeleton staff, but it kept people—a lot of them—connected to the world, even though I had never really met these people. It was way more of a job than a career; just as smaller news operations feed off Microsoft, so do hundreds of companies and tens of thousands of individuals like myself, all happy enough to cash a check to push news around.

Here is where I’d like to get into salacious details about the politics of our programming, but they’re disappointingly bland. Like any operation of a certain size, MSN professes to be non-partisan but is in fact aggressively non-confrontational, to the point where a single user comment on any of our stories that raised a red flag would be brought to us at once. This is theoretically absurd, given the scale of things, but it happened rarely enough not to be too much of a bother. (By far the best one of these was a multimillionaire yacht owner whose boat had ended up in our photo system saying his ship was a nation unto itself that he did not allow to be photographed, so we needed to remove it.) Occasionally articles deemed too left- or right-wing would come down at the request of an editor, but there was such a deluge of content that you never needed to worry about finding a softer-edged replacement for them. It will not surprise you that virtually everyone I worked with was vaguely center-left, politically, and it will shock you even less that some of the smarmier full-time employees were sure to remind us not to let it affect our coverage. For my part, at least, being in the Money vertical allowed me to run as much left-leaning content as I wanted, given that the entirety of the rest of the section skewed right. I never had to elucidate it like this, though it came up once or twice; I just showed up and worked and that was it as long as no one was embarrassed.


About the only other time I can remember someone citing MSN was when certain outlets objected to the editorial disclaimer we were asked to put on articles that bent toward Scary Opinion, on the grounds that said articles (I believe they were from the Real Gawker family) were not, in fact, opinions, but facts. They were right, of course, but the language and tone of most of their stories precluded us beaming them to virtual Middle America in the first place.

This record of successful blandness, combined with the internal head-slapping after the Little Mix-up, does not bode well for the future—or the present—of news. MSN never got into real hot water for messing up a headline or story, which was a minor miracle but was at least the goal. Now, at a time where context is needed most, MSN is handing the reins from editors to engineers. A Microsoft spokesperson said the contract lapses had nothing to do with the pandemic, and I believe them; Microsoft has haltingly moved toward automation for years, usually in a cycle of job cuts followed by partial rehirings when they realized (as the editors did immediately) that they had bitten off more than they could chew.

This time feels different, though. Based on how far they’ve come down this road, the algorithm will sink or swim on its own, which is to say it’ll probably sink and take down the whole of MSN with it. Maybe that’s overstating things, but MSN is low enough in the Microsoft hierarchy that its existence has felt like it was on the chopping block for years. (This is all ultimately speculation, because I had no idea, after nearly a decade, who makes the decisions there, nor did my boss or talent agency.)


A lot of the friction in the changeover from human to robot involves negotiating the change from the contractor model, which Microsoft and others use to save themselves from paying benefits or severance or providing long-term work assurance, back to the full-time employee model, which takes a bunch of people who are harder to fire and forces them into jobs to which they may or not be suited. The tension in the old model was between the contractors like myself with news backgrounds, who produced the lion’s share of news, and the full-timers with Microsoft backgrounds, who “oversaw” it and did a lot of thinking about whether we should be pivoting to video to slideshows or vice versa that week. There were exceptions on both sides, but this is how it went until July 1, when what the Guardian cheekily called our “algorithmic robot” replacements stepped in, kicking the full-timers down to grunt duty and the contractors out the door, context —in and outside the newsroom—be damned.

That is to say that there are still human eyes on the site, just fewer, and that the people behind those eyes know the writing is on the wall. Even the process of losing my job was impenetrable; my boss couldn’t tell me who made the decision because he simply didn’t know. This is how announcements come at Microsoft: from on high. In that way, the algorithm’s ascent is just part of the natural order of things. If the decision-makers aren’t even identifiably human, why wouldn’t they rather wipe us all out?


The nice part of the job was having access to all the world’s news, though it gave me scant time to read it all, given the constant maintenance MSN Money—just one vertical!—required. Each section—News, Money, Sports, Entertainment, Travel, etc.—was its own fiefdom, as was the Home Page, which was and is the ultimate traffic driver for MSN. Much like the CD business thrived long after you and I stopped buying them, the MSN home page is still big money for legacy media of all shapes and sizes. It is common for popular stories there to get 15 million pageviews and have 15,000 concurrent users, and for super-popular stories to bury those numbers. I produced all these stories, and I still don’t know a person who actively tries to consume them. But someone does.

Who? From what I can tell, old people. The most consistently popular topic we ran in Money, and we did more or less daily, was “At what age you should take Social Security?” I could spend my whole morning agonizing over which Times story to feature (the Big Three old newspapers had strict limits on what we could take), follow it up with a Social Security story from another partner, and bang—the latter would do better numbers, invariably. I largely avoided learning about popular topics in the other verticals, but as Money and Lifestyle were next to each other on the home page, I can tell you the only true rivals to Social Security’s popularity as a topic were Duchesses Catherine and Meghan, about whom it was not possible to program too much.

The Social Security stories are a good example of how some partners gamed the system for the better. Most of them came from the Motley Fool, a financial publication that effectively parcels out stories on the same seven subjects a day, incredibly effectively— think Chipotle for popular retirement and savings topics. On the flip side, the same Motley Fool, along with several similar smaller operations from which we licensed content, was found several years ago to be publishing articles from writers who had been paid to write articles specifically to move stocks in a certain direction. All of the writers settled, likely because they were all caught dead to rights. Given that these are the exact type of articles we—humans, with knowledge of what makes a piece plausible or not—chose or did not choose to program at MSN Money, I think the problems are about to get much worse.

There are two problems, as I see it. First, I don’t think the algorithm will be able to keep things orderly. So much of the job was about maintenance that, against my better judgment, it reminds me of Hamlet: “Tis an unweeded garden that grows to seed, things rank and gross in nature possess it merely.” My job was often that of a gardener, and as someone who is now doing quite a bit of actual gardening in unemployment, the difficulty and necessity of it—the effort it takes to produce something worthwhile—is top of mind. The only difference is that I fear things “rank and gross in nature” will not merely possess MSN, but subsume it to the point Microsoft wants its name completely disassociated from it. I hope I’m wrong.

Second, the algorithm won’t really know the game, the way even I didn’t when I started. I can see it now, and if you’ll allow me another analogy I’d compare it to antichess, which is exactly what it sounds like: A game you’re trying to lose. The quirk that makes the game go is that you must take a piece if you are allowed, and as a novice, I quickly learned that this can lead to easy cascades from which there is no escape, far more so than regular chess in that, at least in a proper chess death spiral, you think you have agency over the moves. In antichess there’s none of that: It’s usually a cascade to certain finishes, with the fatal mistake so obvious to be easily perceptible. This feels like that mistake. MSN is up against forces it hasn’t reckoned with, and whoever’s left can only hope they get one more move to fight back, even if it can only all still end one way.

Correction: An earlier version of this story referred to Little Mix as a duo; it has also been updated to clarify who settled a case.

Follow Bryan Joiner on Twitter.