Robots and Content: Calculators or Creators?


Will a Robot Take Your Job?” has become a ubiquitous headline in the news in recent months. It seems that no profession is safe. Content creators are not immune; in 2016, The Washington Post announced that it would use robotic software, Heliograf, to compose news reports about the Rio Olympics. It works by parsing through a template of keywords and relevant phrases (which are created by Post editors) and linking this vocabulary with uploaded datapoints it deems relevant. The result is a computer-generated narrative that is published across platforms. The Post also uses Heliograf to provide real-time voting results during elections. To set up the system for reporting on this type of event, editors created an indexed template by choosing metadata unique to each election being covered, such as the names of candidates and voting patterns across precincts. They also had to ensure that every potential outcome was considered, so they needed to enter duplicate phrases for each outcome scenario (i.e., “Smith Wins,” “Smith Loses,” “Adams Wins,” “Adams Loses”).

Other media outlets using bots to create news stories include The Associated Press, which uses data analytics company Automated Insights to collate quarterly company earnings announcements, and Forbes, which uses Narrative Science to prepare this same type of content. Forbes not only lists it as the author, but also refers to it as a partner. In earthquake-prone California, the Los Angeles Times uses Quakebot, which deploys an algorithm when a report comes in from the U.S. Geological Survey (USGS). The resulting product is basically a notation of the time, magnitude, and epicenter of the earthquake.

Usually, when robots are discussed in conjunction with journalism, “reports” is the term used to describe the final product. However, the reports that they generate are basically just lists or ledgers with an accounting of statistics regarding events. Ultimately, the use of these AI algorithms is a quick way to generate a list of descriptors that characterize an event. To that end, are they really creating content? It is probably better to call these bots “calculators” rather than true creators, and their work product reports, lists, or notations rather than stories. Are these accounts a real threat to the work of human content creators? Hardly, not only because it is debatable whether or not they inform, enlighten, challenge, and entertain the reader and hopefully inspire thought and discussion (which all good stories do), but also because they can’t be developed, operated, or monitored without extensive human interaction.

Each of the aforementioned AI content creation bots relies heavily on human input and monitoring. In the case of Heliograf, the editors must create the templates on which the software relies. This is trickier than it seems; if the appropriate metadata is not listed in the template, the datapoints chosen by the bot might be described by terms that are irrelevant or even totally wrong. Also, the system uses Slack to alert reporters of any instances of outlying data. When reporters receive these notifications, they need to investigate and, when necessary, rewrite the stories. Call me crazy, but the creation of the metadata for the template is an extra step that isn’t necessary if reporters end up writing the stories themselves. Also, it seems as if it might be kind of time-consuming to enter every potential scenario of an election outcome into a template so that a robot can link them to relevant data, when a reporter can easily look at the numbers and determine who won and who lost.

In a previous column (“Combating Fake News With a Human Touch”), I wrote about the rise of fake news and the need for critical, analytical voices to be involved in the content creation process in order to ensure—at a minimum—accuracy and, ideally, a quality, informative, authoritative, and hopefully entertaining product. The same process holds true for AI and journalism. Rather than reduce the workload of journalists, AI algorithms add to and complement their existing skill sets, since journalists are creating even more content every time they write the metadata that is needed for the bots to be deployed.    


Related Articles

The assumption that organizations would buy a single-vendor DXP based on the number of parts within said solution is at odds with what our analysts and consultants were seeing across companies of all sizes and industry sectors. We decided that it was time for voice-of-the-buyer, fact-based insight that would separate DXP myths from reality.
Marketers have done well by pairing technological advancements with the best AI tools—such as personalization based on browsing history. This trend should continue with the new era of global chatbots. When companies move toward marketing by chat, they need to use this channel wisely or risk damaging their customer relationships. As technology advances, user expectations rise.