Do not fear the robot-journalist
Most conversations about AI in journalism tend to revolve around the robot-journalist and the question, if and when machines will replace human writers. But as intriguing as these discussions are, they tend to miss the actual role AI will play in newsrooms and is indeed already playing.
In a broader sense, the term AI actually describes an accumulation of various historical definitions and methods, as well as science fiction imaginations. Today “AI” is mostly used as a stand-in for “machine learning” algorithms. A collection of concepts and systems, which — under human supervision — are able to analyze vast amounts of data, while also improving itself. And although those systems can achieve impressive results and solve complex issues, as well as write rudimentary texts, they’re still lacking one important skill journalists are well versed in: creating context. Machine learning systems are bound to their narrow field of expertise and are not able to handle problems outside of it.
The augmented journalist
The future of AI in newsrooms lies thus not in automation and replacement, but the augmentation of journalists. And the best thing about this future? It’s already here.
The New York Times, for example, uses an AI facial recognition system to help their reporters identify the over 500 senators and representatives leading the country. ProPublica is using machine learning to identify the most important topics spoken about inside the US-Capitol, thus helping their newsroom and readers to keep a finger on the political pulse. Or just imagine the hundreds of hours AI services like Trint or Otter.ai can save newsrooms by helping journalists automatically transcribe interviews.
But machine learning can also assist the investigative work of newsrooms. Texty, an outlet based in Ukraine, trained an algorithm to identify thousands of illegal amber mines all across the country’s vast forests on satellite images. Global Fishing Watch used a similar system to track down potentially illegal ship-to-ship transfers across the globe. Combining orbital photography with machine learning can actually yield a number of intriguing results. For example, are companies like Orbital Insight or ICEYE predict the stock price of major retailers by analyzing the number and type of cars in their parking lots. Or they can extrapolate the yield of a certain crop, by using radar satellites or they can measure the wealth of neighborhoods by looking at the number and size of swimming pools or even the wealth of a whole nation, based on the number of light sources illuminating its cities at night.
AI as a defence against the overwhelming flood of information
But machine learning can not only detect patterns, but also discrepancies and deviations in a given data set. An ability journalists of the L.A. Times used to identify a mismatch between the LAPD official crime stats and the actually number of committed crimes.
Considering the daily flood of information, machine learning might actually be the right tool at the right time to help make sense of an ever more complex world. Especially considering the increasing numbers of sensors and systems measuring every aspect of the global economy, society, and ecology. Today, big leaks like the Panama Papers contain terabytes of documents, e-mails, and images. Enough data to bring whole newsrooms to a standstill for months at a time. But the clever use of machine learning can help journalists not only to sort through this pile of raw data, but will also aid in finding and making sense of hidden patterns within the data. AI can thus help journalists to investigate faster and more thorough than ever before.
No algorithm is infallible
But machine learning does not come without problems and risks. Since most services are being cloud-based at the moment and tied to big software companies, journalists should be mindful about the risk of sending potentially sensitive data to foreign servers, where some legal protections might not apply. Furthermore, no algorithm, as advanced as it might be, is infallible. Especially machine learning struggles with the inert problem, that not every pattern detected by the algorithm is indeed a pattern. The consequence can be a biased or even flat-out inaccurate output. Journalists should thus not only understand how machine learning works and know about the risks involved, but also stay in the loop and in control.
It is, therefore, time to stop thinking about the robot-journalist and to start talking more about the cyborg-journalist and the future collaboration between man and machine instead. Because this future has already arrived and will continue to shape the future of journalism.
And what is your opinion on AI changing every aspect of our lives?
This article is part of a series of guest contributions on the topic of Artificial Intelligence. You also have an interesting thesis you would like to share with us? Feel free to contact us!
Articles already published:
A guest contribution by Johannes Klingebiel from the innovation team of the Süddeutsche Zeitung. His irregular “Zine” newsletter gathers knowledge and information on technology and the Internet and their impact on journalism and society.