Stop Saying the Robot Out-wrote A Reporter

You would think a top-notch news organization like NPR would know the difference between journalism and non-journalism. That wasn’t the case, however, in this NPR story titled ” ‘Robot Journalist’ Out-Writes Human Sports Reporter”.

The story really caught on on Twitter, where the original NPR tweet has been widely disseminated (just search for “robot journalist” on Twitter). Here’s a sampling:


The only problem is that the person out-written by the software program is not, and doesn’t even claim to be, a journalist. Instead, the human-produced piece in this case is a press release on the official George Washington athletics website — in other words, a PR piece.

Having read countless sports press releases during my career as a sports journalist, I know these things are kind of hit-or-miss. Some schools have very well written releases that tend to be fair, while other schools produce horrible homerish material. I’ve even seen a number of examples of releases burying great performances by opposing teams, though not quite to this degree. In this case, even as a PR piece for the team that came out on the wrong end of the perfect game, the writer made a bad decision in burying the rare feat. If a journalist reported about this game, however, that mistake in judgement would never have been made because this is as easy a judgement call as it gets for a journalist. There’s no way you’d bury the perfect game. The way the NPR story distorted the facts to call the sports information writer a journalist is a slap to the face for journalists everywhere.

But it gets even better. In case you missed it, Poynter addressed that bad gwsports.com piece a good three weeks before the NPR story (and here’s a bunch of sports journalists discussing it). The money quote from the Poynter piece (emphasis added by me):

“This is the George Washington website,” GWU sports information director Dave Lubeski tells Romenesko. “We’re in the business to promote our athletes and our team. We’re not claiming to be journalists.” What some call “the buried lead” was discussed after the story was posted, says Lubeski, and it was mentioned that the perfect game could have been noted in the sub-hed. But “we’re not in the newspaper business,” notes the SID.

So despite the quote above from the sports information director saying, “WE’RE NOT JOURNALISTS,” whoever wrote NPR’s “robot journalist” story decided to call the person who wrote the piece on gwsports.com a journalist anyway. It’s hard to imagine that anyone could mistake a school’s official athletics site for a journalism site, and I would certainly expect someone working at one of the top journalism companies in the country to know the difference. The question, then, is what’s the motivation for distorting the facts? Is it just because “robot out-writes journalist” makes for a better story than “robot out-writes PR piece trying to hide home team’s embarrassment”? If so, then whoever wrote that NPR piece has no business calling him or herself a journalist. Or if the NPR writer really couldn’t tell the difference between gwsports.com and a journalism site, then he/she should be replaced by a computer program.

Update 1 (4/18)

I should add that I’ve known about Narrative Science’s sportswriting software for a while and I’ve checked out some of its work from the Big Ten Network’s website. For the most part, the stories are actually not bad and definitely usable. If I was putting together a non-rev sports roundup from press releases, I would definitely take these computer-generated stories over some of the ones I’ve gotten in the past because they are a lot easier to extract useful information from. However, there are signs in the computer-generated stories that are relatively easy to pick up that would lead you to wonder if they were written by software, such as the odd placement of a particular factoid.

Update 2 (4/18)

If you were going to ask me which of all the posts I’ve written that’s going to shatter the single-day record for visitors to this blog, I would not have guessed this one, but this story apparently grew legs as it has been picked up by a few other sites and retweeted a number of times (I know, a few hundred views really aren’t much, but considering this blog probably averages around 100 views a day, it’s a big number for me). I’ve had traffic spikes before, but nothing quite like this:

Of course, much more important to me than my traffic numbers is the notion of correcting bad information that’s been widely disseminated. So, thanks to everyone who took the time to check out this post and pass it on. Also on that front, thanks in part to King Kaufman and the folks at the news bug-catching service mediabugs.org, NPR has changed the headline on the story. You can read more about this at mediabugs.org, where Kaufman reported the error and the service contacted NPR about it. Like Mark Follman said in the discussion there, I think NPR does need to post a more prominent correction than just a note in the comment thread on the story.

Update (4/20)

I saw this last night but didn’t have time to update the post until now. NPR has indeed gone back and added a correction at the beginning of the story, as well as on its corrections page.

Bookmark the permalink.

Leave a Reply