After countless decades of considering myself an elite armchair quarterback, studying the gridiron greats from Jim Brown to Peyton Manning, I now come to learn how much I really know about professional football—nothing. Or so says the ghost in the machine that is Yahoo Fantasy Football.
After two hours of carefully deliberating on every pick in the Daily Dot’s fantasy football draft, I was pleased as punch with my choices: a great quarterback (Andrew Luck), powerful running back (Alfred Morris), and an elusive wide receiver (Brandon Marshall). I could almost taste victory in the DDUFL.
Then, the roof caved in. My draft report card, quickly and automatically generated by Yahoo’s content partner Automated Insights, was less than impressed with my draft skills. This marvel of transparent technology evaluated my picks, compared them to the others in my league, and pronounced me as someone “who hates winning.” I will finish dead last, according to the phantom talent evaluator. I should take up bird watching.
It gets worse. In commenting on my expert assessment of running backs, the stern hand behind my report card proclaimed: “They put together the worst group of RBs in the league, as they added Morris, Stevan Ridley, Darren Sproles, and Bryce Brown to their roster.”
The power behind these sharp-witted fantasy football synopses is Automated Insights, a Durham, N.C.-based company that specializes in the art and science of what is commonly referred to as “robot journalism.” For fantasy football players, the service is a clever added bonus that keeps people on site longer. It’s like having a hometown beat reporter covering your fantasy team: analyzing your draft picks, providing fun recaps of games, and insightful game-day analysis.
But Automated Insights, and contemporaries like Narrative Science, are having a major impact outside the realm of fantasy football. Such prominent news organizations as Associated Press are trialing this technology to churn out data-heavy stories on finance and sports.
The process of turning data into semantic magic has five steps:
1. Data from any number of sources—XML, spreadsheets, APIs (secure remote data connections), etc.—is sent to the Automated Insights Wordsmith platform for interpretation.
2. The data is then analyzed, and Wordsmith looks for patterns and trends and creates a context (general meaning and catalogization) for the content.
3. Next, the data is matched against other data at an aggregate level and benchmarked against other data to create general comparisons, which lead to general observations.
4. Using advanced Natural Language Processing, which derives meaning from language, a narrative or story is built around the key insights from the data analysis. The narratives can be created in any number of forms from short synopses to visualizations or even social media posts.
5. The stories are then published, either directly to a content partner’s site or to the partner’s content management system.
Providing Yahoo’s Fantasy Football combatants with pithy evaluations of their teams is but one manifestation of robot journalism, and a fun one at that. Implementations at large, incumbent news organizations such as Associated Press, the Los Angeles Times, and Forbes have received more media attention. The trend has forced the controversial question: Can journalists be replaced with machines—if not entirely, then at least in more data-driven subject areas?
In finance, for example, Associated Press uses the Wordsmith platform in conjunction with data from Zack’s Investment Service to create complex, deadline stories for earnings calls. The robot journalism approach would take advantage of being able to evaluate mountains of historical data on a particular company and create a context-rich, narrative-led story mere minutes if not seconds after the earnings calls end.
The same goes for sports stories. Narrative Science works with the Big Ten Network to provide digitally driven in-game summaries, halftime recaps, and post-game stories by analyzing live game feeds and transforming them into digestible narratives that also can be converted into tweets—the kind of short-form posts that have driven the success of sports site Bleacher Report.
With decades of journalism experience under my belt, I can understand the value of using machines to turn out stories that require analyzing dense sets of often difficult-to-understand data. It would take an army of interns to replicate the process employed by Automated Insights—and robots don’t miss deadlines.
From the reporter perspective, these sorts of assignments are tedious and offer little chance for creativity. For the publisher, some financial reporting can be dry and void of personality, but it’s a boilerplate topic that newspapers feel the need to cover as part of their charter. Employing a more cost-effective way to provide coverage makes business sense in today’s competitive marketplace.
Also consider the fact that data is becoming an essential ingredient in creating rich, illustrative narratives. Charts and infographics are vital to young, digitally inclined readers. The National Institute for Computer-Assisted Reporting (NICAR), part of the Missouri School of Journalism, has offered training courses to journalists in how to use tools such as mapping and databases to add more context to their work. Journalists armed with both reporting and data skills could make editors and publishers rethink their approaches to machine-delivered content.
As you might expect, industry voices have not yet reached a consensus on the issue of robot journalism. Some see the merit, while others miss the value it brings to the world of media.
Jim Romenesko, a well-known journalist and one of the more strident supports of traditional journalism, gave his take to The Kernel in four simple words: “not ready for primetime.” Romenesko pointed to the recent snafu when a computer-generated story that appeared in Game Changer referred to female Little League baseball player Mo’Ne Davis as a “he.”
From the other side of the debate, there’s Kevin Roose, a writer for New York magazine. “We should be celebrating the rise of machines that can supplement and assist us in our jobs,” he wrote in a recent article, “while doing the most banal parts of our workloads for us.”
To Roose, I counter with the 10,000-hour rule Malcolm Gladwell spoke of in his book Outliers—the premise that the key to success in any field is, to a large extent, a matter of practicing a specific task for a total of around 10,000 hours. As someone who has written more than his share of obits, created TV listing grids, and even gathered Friday night high school football scores, paying your dues has its rewards.
Then again, I can’t imagine anyone devoting serious time and energy to reporting on my fantasy football team. And even if someone did, it’d be hard to top this snarky headline I already received:
“With a Projected Finish of 14th, Gonzo Obviously Hates Winning.”
We only have 16 teams in our league.
Illustration by J. Longo