In the past year, a number of articles have come out warning us that robots are going to replace our livelihoods. More than a third of U.S. jobs could be at risk by 2030, according to an analysis by PricewaterhouseCoopers, and “virtually no job is safe,” predicts the International Data Corp. The consulting firm McKinsey & Company puts the number even higher, writing in a recent report that “We estimate that about half of all the activities people are paid to do in the world’s workforce could potentially be automated by adapting currently demonstrated technologies.”
Eek! This takeover seems inevitable — 77 percent of Americans surveyed by Pew think it’s realistic that robots and computers will be able to do many of the jobs currently done by humans, but at the same time, most of us think that it won’t be our jobs that will be automated. The same Pew study says that the “majority of workers anticipate that their jobs or professions will not be impacted by automation to a significant degree.”
So what does that all mean for journalists? That was the question that Sarah Kessler, the deputy editor of Quartz at Work, started with when she set out to determine what parts of her job could be replaced by robots or automated, and what parts couldn’t. Part of the way through her research, she realized that Quartz was already automating her work through its Bot Studio, which creates bots that help journalists do their jobs.
“Before this point, I’d thought about the [bots from the bot studio] as a cool project,” she wrote. “I hadn’t linked them to the type of automation that is happening across other industries.”
But John Keefe, a developer in the Quartz Bot Studio, told Sarah that the bots his team was making weren’t designed to replace Sarah’s work — they were designed to “enhance” it.
I wanted to learn more about ways bots could help our work — and what to do to prepare for a possible bot takeover of my beat, so I reached out to Sarah and John. Our conversation is below.
Sarah, I’ll start out with the scariest question: Are journalists going to be replaced by bots?
Sarah Kessler: Not today. There are some computer programs that write newspaper articles, but typically they work on articles that wouldn’t be written by human reporters.
I’ll quote some examples from my story: “The Washington Post built a tool called Heliograf that combines databases with editor-created story templates to generate stories. During the 2016 election, it created more than 500 articles about house, state, and gubernatorial races in every single state. The AP partnered with a company called Automated Insights to write thousands of stories about financial quarterly earning reports, and the Los Angeles Times has used bots to write about earthquakes and homicides.”
“These are stories that, even when written by humans, are already structured like a game of Mad Libs. Most of them wouldn’t be written at all if they weren’t written by robots. For a human with limited capacity, it makes more sense to focus on stories that seem particularly interesting to a wide group of people rather than try to cover every single earnings report or athletic event.”
It sounds like automating parts of your job made you more efficient and more productive. What tools did you employ that helped you do your job better?
Sarah Kessler: To be clear, the exercise was hypothetical. I was asking companies that don’t work with journalists (but work with other industries to automate work) what they would do with an unlimited budget to automate my job, and I gave them an unlimited budget.
The point was to try to explain what terms like “automation,” “digital workforce,” and “bots” mean in practical terms by using my job as an example. I chose my job because I wanted to explain the story in a personal way and because most people can imagine what a journalist typically does. I think it’s important to point out that the argument I made in the story is easier to make when it comes to knowledge work than manufacturing jobs or low-skilled office work.
Most of the technologies weren’t a good fit for journalism (like “robotic process automation,” which specializes in drudgery). Other experts suggested things that I already used, but had never considered to be automation because I think of them as tools, which is a distinction that has always been a bit blurry.
From the article: “We call some of these technologies tools and others, like the interviewing bot, automation. But the divide between the two has always been a moving one. In the 1860s, The Metropolitan Record, a newspaper, scolded “very many young women” for fearing the arrival of French’s Conical Washing-Machine. “This machine will lighten the labor, save the hands, and relieve many of the wearing and disagreeable features of hand-washing,” it argued, “but is not designed to, and will not, take the place of a single young woman at service, we feel confident.”
It’s hard to say whether a “single young woman” ever lost her job to the washing machine, but we’re no longer afraid of the prospect. Same for the spreadsheet, which was once considered dangerous, accountant-automating, job-threatening technology. Today the accounting profession is projected to grow faster than the average occupation, and spreadsheets are just a boring piece of office software.
What do you think newsrooms should automate?
Sarah Kessler: Transcribing interviews. Nobody I’ve ever met enjoys doing it, and it takes up a ton of time that could be spent reporting or writing. A few companies are working on this.
What parts of journalism can’t be automated?
Sarah Kessler: Never say never, but putting together a story based on information from sources rather than data isn’t something that programs for automating news writing currently do well.
John, you work for Quartz, which has a Bot Studio, which creates bots that alert reporters when something unusual happens. I’m wondering what a smaller newsroom might be able to do without those resources? Are there low-cost or free tools available that they could use?
John Keefe: Absolutely. The one I turn to most is IFTTT.com, which stands for “if this, then that.” It operates on the core bot concept that if something (a trigger) happens do something else (an action). The trigger can be a tweet by the president, a certain word in an RSS feed, a particular change in the outside temperature. The action can be adding a row to a spreadsheet, sending a Slack message, even turning on a light on your desk!
We have journalists at Quartz using IFTTT to keep tabs on lots of things, including stories we post ourselves (Quartz has a suite of IFTTT triggers, as do several other news organization).
I’ve posted an example, from a talk at the 2015 Online News Association conference, which still works today (though the IFTTT graphics are a little different today). It’s essentially how WNYC got a “mini scoop” using IFTTT to watch for certain words in the RSS feed of federal court actions in the Southern District of New York.
And it’s all free.
Early in the new year, we’ll be adding more sophisticated features to Quackbot, the Slack tool Quartz and DocumentCloud have built for journalists. It will include features to watch a website for changes that are interesting to you, similar to what Klaxon does but right within Slack. Quackbot is available now and is free to journalists with a DocumentCloud account (which is also free).
What bots do you wish existed to help journalists?
John Keefe: I wish there were bots that helped individual journalists cover their beats or obsessions in ways they didn’t anticipate. Today I can say, “Computer, watch this website and let me know if something changes.” Tomorrow, we might say, “Computer, you know me. We’ve been working together for some time now. Keep an eye on my areas of interest and let me know if you detect something newsworthy I should look into.” And Sarah’s right; we’re thinking about that every day. With time, code, and a little luck, we’d love to add that kind of ability to Quackbot.
There are so many things that could be automated to help reporters become more efficient and save them time, and it seems like a lot of them might be once-offs. I’m curious how you decide what to prioritize, what to automate, and how you work with the reporters to design the bots.
How has the response been in your newsroom? Do you have everyone clamouring to automate parts of their work? When do you point them elsewhere?
John Keefer: The whole concept of bots and technology that augments our lives is very “Quartzy,” so it may not be surprising that the response has been fantastic. As I write this, there are 83 people in our #bots Slack channel, there are several bot-related sub-channels, and we have a few channels occupied primarily by bots, such as #bot-preschool (“where bots are free to learn, play, and occasionally break things before they’re ready for the real world.”)
People throughout the company and from around the world have suggested a long — and growing — list of bot ideas and requests. We’re trying to turn as many of them into action as possible. Some are easy and quick, others are far more complicated. One of the most common requests is a bot that can answer grammar and style questions. That turns out to be pretty tricky, though we’re working on it.
If someone has a simple trigger-action request, it’s surprising how often we can get pretty far with IFTTT. That said, usually the request is more complicated than IFTTT can handle, so we add it to the list. For example, you can get an alert whenever a particular user posts a YouTube video. You can’t, however, detect when the number of views on the video crosses, say, a million. We’d need to find another way to detect that.
How can automation help reporters?
John Keefe: That whole list of things you should check or do regularly to stay on top of your beat? A computer can help do some of those checks, helping ensure you don’t miss something. Ideally, that helps you get good story leads and scoops.
It’s also clear that we can use computers to detect things a reporter can’t see on their own. A great example is ProPublica’s use of machine learning to detect what’s uniquely of interest to individual members of Congress. A human would have had a hard time detecting those nuances from hundreds of thousands of press releases.
What can’t automation replace?
John Keefe: Great writing. Rich storytelling. Journalistic ethics. Sorting fact from fiction. Knowing truth from misinformation.
What bots have you worked on in 2017 that you’re most proud of?
John Keefe: Definitely most proud of our Quartzy Bot — though it’s not the kind of bot you’re thinking of! It’s a chat-based interface for some feature experiments associated with our Quartzy lifestyle edition. Emily Withrow, our Quartz Bot editor, has done an amazing job writing very cool experiences (making no-knead bread, watching Stranger Things, etc.), and I’ve had the pleasure of coding some of the back end and natural language systems to make it as conversational as we can.
I also really like Quackbot, though its real smarts (and utility) are really set for the next iteration.
It seems like the Bot Studio has made reporters at QZ much more efficient. I’m curious if you’ve seen your efforts replicated in other newsrooms, and if so, how?
John Keefe: I’ve seen several parallel efforts, at places like Vox and the Washington Post, where folks are using Slack-connected bots to help them do their work. I also know people have been looking to replicate data-monitoring projects like @actual_ransom, which was made by Keith Collins while he was at Quartz and which had a bona-fide scoop.