When a publisher decides to deploy tech for automated content, the primary reason is generally a business one: The robots generate additional coverage, which, in turn, drives advertising and reader revenues. However, introducing data-driven robot journalism has a secondary effect: It constitutes a new editorial tool — in more ways than one.
As the term suggests, introducing robot journalism involves new tech. But more interestingly, it involves the newsroom. Having spent some four years working with publishing organisations to implement our text robots, I have seen first-hand how the process of integrating automated content positively impacts reporters and editors — in their daily work as well as in a more fundamental way.
Let’s start with the latter. This touches on something I mentioned in my previous column, namely the misconception that there’s an inherent danger with robot journalism because the robots are somehow in control. They are not; the newsroom is. In fact, you could say the newsroom creates the robot. While we at United Robots provide the programming and natural language generation (NLG) expertise, it’s the journalists who set the framework and rules around the automated content — for their news publication.
This process often acts as a catalyst for a healthy newsroom discussion around how and why its journalism is created. What language is used and why? How do we identify stories? How do we determine which should be published? What constitutes a good headline? And so on.
Let’s take sports as an example. At the beginning of the implementation process, we work closely with the sports department to create a platform for the automation of texts for that domain. We collaborate with the journalists to create a framework based on how they think.
This can be a bit of a hurdle, as reporters are not generally used to qualifying, quantifying, and structuring their work in this way. The positive outcome is not just the automatically generated content, but also a deeper understanding among journalists of how the work is done and why. The process creates a new level of ownership — of the language, the values, the dos and don’ts.
Once the robot has been “given its orders” by the newsroom, it becomes a useful editorial tool in the daily work in a number of ways:
- Surfacing stories in the data. A computer can find patterns in the data that even the most knowledgeable reporter would struggle to identify (not least due to how time-consuming such a process would be). If the newsroom has specified the robot should look for “a record suite of victories” for teams in a given football league or “the most expensive property sold” in a certain region this year, the robot will consistently look and not make any mistakes.
- Making data-based analyses. The robot can look through huge amounts of data, and several data sets in parallel, which allows it to make comparisons, identify trends over time, and generally provide analyses based on the data, which would be hard for a reporter to manage.
- Alerting the newsroom to anomalies. The robot alerts editors to any instance when the information in the data is outside the norm, as set by the newsroom. For example, if a hat trick has been scored or a house has sold for over a certain price in a certain area.
In this way, deploying automated content can benefit not just the business of publishing, but also newsroom practices and the use and understanding of data to lift the journalism.