Features:

Quakebots and Pageview Quotas: Bot or Be Botted?

Matt Waite on Daft Punk, hamster wheels, and journalism’s Rushkoff moment


Sisyphus and the perpetual deadline (Wikimedia Commons)

One week separated their announcement, but in a lot of ways, they flow from the same semi-unsettling spring, blending into a stream of logical conclusions.

On March 17, a small-yet-widely-felt earthquake woke Los Angelenos, including LA Times developer-journalist Ken Schwencke. Schwencke rolled out of bed, logged into the Times content management system and published a story. Wholly uninteresting, except for the fact that a bot he programmed wrote the story that went up three minutes after the quake. For the rest of the day, Times journalists–humans all–updated the story with information humans are good at, like scene details and quotes from other humans. By the end of the day, the bot-written paragraphs were consumed by the other details, assigned to the dustbin of hours-ago history.

Schwencke would spend the rest of the week being interviewed by other journalists curious about his bot. A lot of posts were written. That set off the entirely predictable round of “bots r stealn’ r jawbs” me-too blog posts.

A day after an interview with Schwencke aired on NPR’s Weekend Edition–replete with do-bots-get-surly-with-editors jokes–another gush into the stream happened: The Oregonian, a widely respected newspaper with its share of Pulitzers in its past, was changing how reporters would be evaluated and even compensated. The short version: Post more, interact more, drive more pageviews, and you’ll get a bonus (maybe, if there’s money left at the end of the year). Don’t and…well, company documents don’t discuss what will happen, but we can surmise it’s not good.

Why am I connecting these two ideas? And how?

Both are taking lines from an old Daft Punk song: “Harder Better Faster Stronger.” Pageviews are the coin of the realm, so post more so more people click. Or post faster. Or both. We need more stuff on the thing so people can click.

Schwencke’s bot, while useful because it writes a story that would bore most human reporters and does it much faster than they could, is really about getting a story online as fast as possible to catch the pageviews when people are most interested. It’s a good idea that’s good for readers and business. Can’t click on a story telling you about the earthquake that just happened if there’s nothing to click on.

The Oregonian’s strategy is similar, except not nearly so surgical. Everyone post more so there’s more to click on. Engage more so more people tune in and click. Click click click.

Neither idea is a new strategy. Schwencke’s bot isn’t the first or most prolific at the LA Times, and they’re downright lazy compared to companies like Narrative Science and Automated Insights writing thousands to, say, a billion stories. Online news organizations like Gawker have used traffic-based compensation methods in the past, but the Oregonian is getting attention because its the first time a place that still produces a dead trees edition has done it. Reactions range from it’s a terrible idea to it gets some things right.

Neither strategy is necessarily evil: They’re merely a result of logical conclusions a page view based system. I’m not about to start a rant about the consequences of a page view focused business model (not here, anyway). Others have done that better than I can. Some are calling for a new way and others argue pageviews aren’t all bad. And there’s some really brilliant things going on in this space. Read those.

Given the attention given to Schwencke’s quakebot in the same week as the Oregonian announcement, more publishers are going to start asking developers “can you make a bot that writes a story?”

Here’s BuzzFeed-esque quiz to determine if you too can write a story bot (and you won’t believe the answer!):

  1. If you breathe on a mirror, does fog appear?
  2. Do you understand enough programming to match a pattern?
  3. Do you understand enough programming to substitute a variable into a sentence at a given spot?
  4. Can you use if/then conditional logic?
  5. Did you play Mad Libs as a kid?

Congrats! You can write a simple bot that writes stories! Go forth and steal a reporter’s job or destroy journalism!

Okay, okay, laying it on a little thick. But the point is, it’s not hard to get in the door, but it opens up some interesting questions. Can people tell the difference between the work of a bot and a human? Maybe. Do people care if they’re reading the work of a bot or a human? Sort of. Does it matter? Answer: No, publishers are going to do it anyway.

Another interesting question: What are the ethics here? Preliminary (and criticized) research shows people think a bot-written story is more objective than a human written one. But the truth is a bot is as objective as the person who programmed it. It’s as opinionated as the algorithms that power it. In pseudo code:

if change > 10 percent:
    phrase = "alarming rise"
else:
    phrase = "small increase"

Like human-written stories, who decides what’s alarming and what’s a small change? Who said 10 percent is the line? Alarming to who? And is 10 percent always where the line gets drawn? And is it always alarming?

Less scrupulous companies might seek to capitalize on that false belief of robotic objectivity. How do we spot that? Should there be a robotic story Code of Ethics? And if so, there’s no argument that it should be an API, right?

As soon as I saw Schwencke’s bot blowing up, I started to wonder is there a journalism school class here? What would it look like? And what does it say that a class about automating journalism is coming from inside a journalism school? I’m sure buggy whip schools had auto mechanic classes too, right?

It didn’t take long for me to come to the decision that this class had to happen, and soon. Took me a day to pitch it–one email, one conversation, and it was on the books. Current title: An Introduction to Storybots. The syllabus, which isn’t remotely close to being done, is on GitHub, along with some other materials. I’ll be working on it up until class starts in the fall.

I’m teaching the class because I want to–I’m already teaching the classes I’m contractually obligated to teach. I don’t have to do this, but I feel like someone should be–this is the program or be programmed moment for journalists. If you’ve never heard Douglas Rushkoff’s argument, go watch this and think about bots writing stories.

This is just getting started.

Credits

  • Matt Waite

    Matt Waite is a professor of practice in the College of Journalism and Mass Communications at the University of Nebraska-Lincoln and founder of the Drone Journalism Lab. Since he joined the faculty in 2011, he and his students have used drones to report news in six countries on three continents. From 2007-2011, he was a programmer/journalist for the St. Petersburg Times where he developed the Pulitzer Prize-winning website PolitiFact.

Recently

Current page