As an aspiring journalist, I’m used to feeling like an anachronism. Many people assume that the revolution in digital technology will eventually kill newspapers, and leave a cacophony of independent bloggers and Twits in its wake. But at least that means people will still be doing the writing. If a new company called Narrative Science has its way, that could change too.
According to a recent story in Wired, Narrative Science has developed an algorithm that can write news as well as a human. The company is currently selling its services to businesses and the parents of Little League players; its computers primarily churn out quarterly reports and coverage of Little League games. However, company cofounder Kristian Hammond thinks that, in 15 years, 90 percent of news will be written by computers.
I’m not so sure about this, and not just because I want to keep my job. In order to write stories, Narrative Science’s computers need to be fed raw data. They can sift through that data much more efficiently than a human, but someone needs to compile it for them. In the case of the Little League games, Narrative Science relies on an app called GameChanger, which parents use to file all the statistics of their kids’ games.
Even with humans doing the leg work, Wired’s Steven Levy believes Narrative Science will capitalize on the importance of data in our lives. In theory, a system that can see every minute detail of an event, from a pitcher’s stance to the precise shade of orange of John Boehner’s skin at a Congressional hearing, can write a more accurate story than a person.
If we focus on data, then the machines win. Computers are unerring and un-judgmental, and they can sift through massive caches of numbers more efficiently than humans. However, there is hope for human journalists: information.
Data and information may seem like the same thing, but they are not. A datum is the building block that a piece of information is made from. That’s why people use the term “raw data:” it needs to be formatted into something that people can understand. That is where humans come in.
A machine can record a player’s batting average with extreme accuracy, but it will never be able to express the feeling of watching that player come out of a slump by hitting a home run. There are also situations that are too subjective for data. Can statistical analysis of past speeches predict what Newt Gingrich will say next? Can anything?
Another advantage information has over data is timeliness. Taking a poll of eyewitnesses at an event will not always be possible, so how will mechanized reporters get the data they need to compile a story? Having a human reporter on site to decide what readers need to know is still the only way.
You can learn a lot by crunching numbers, but it’s hard to tell what it all means. Reporters are more than just meat puppets who string sentences together; they are professionals who figure out what people need to know, and how best to tell them. There is no love lost between the public and the media, but that does not mean that computers can do this job better than humans.