Working consciously with AI in newsrooms
Artificial intelligence tools have proved tremendously useful for AP, ICIJ cross-border investigations, Yle, France TV, and Deutsche Welle. However, several factors must be considered while developing and implementing systems. Six editors discussed the opportunities and challenges of AI and automation for the newsroom at the Artificial Intelligence and the Future of Journalism conference in Portugal.
The conference took place online on May 11 through May 12 2021. The panel "How are AI-based technologies used in newsrooms? Practical cases and opportunities" was moderated by Charlie Beckett, Head of Polis at London School of Economics.
Augmenting journalism
Automation is key to speed up some of the tasks that reporters tackle, said Emilia Díaz-Struck, research editor and Latin American coordinator at the International Consortium of Investigative Journalism (ICIJ).
The tech team at ICIJ developed the systems that, among other projects, drove the FinCEN Files investigation. This involved handling vast amounts of unstructured data in an array of file formats.
“DataShare is an open source tool that lets journalists safely explore the files and records. It has an AI component that extracts people, organizations, and locations. It does what reporters do, but it does it faster,” Díaz-Struck explained.
Katie Bremme, Executive Product Manager Innovation, France TV, agreed:
“The technology should liberate journalists from tedious tasks to build more interesting content.”
France TV built its own systems to, among other things, generating metadata, text recognition, automatic transcription, and categorization of topics.
Requires human validation
The editors agreed that machine learning is not always the right answer. Developing and testing systems takes resources and human expertise.
ICIJ does a lot of validation and factchecking, Díaz-Struck pointed out.
“It has been a learning curve. It is not all just a machine learning effort. Standard data analysis is also a possibility. Machine learning is very trendy, but it requires time.”
Jarno M. Koponen, Head of AI and Personalization at Yle News lab elaborated:
“We cannot trust automated system to run by themselves, he said.
“Being true to our journalistic values is the key. Whatever tools we use, we need to evaluate them according to our values. If we apply our methods and develop these systems ourselves, we can control the quality of them.”
All panelists stressed how editorial and technical efforts must be combined. Díaz-Struck argued that editors who understand the technologies and at the same time poses the right questions, are crucial.
“If you do not get 100 accuracy, what do you do? How do you report bias? Asking Critical questions when developing will ensure the same critical thinking for tools as for journalism.”
Challenging tech companies
Bulding in-house systems are key to maintain control and quality of the data and analysis, the editors agreed.
At Yle News Lab, their work with recommender systems and the personal news assistant, Voitto. Head of AI and Personalization Jarno M. Koponen, recommended that media organization implementing AI in the newsroom ask the following questions:
“Are we truly innovating on our own? Are we asking who we are, what is our core business, and what are the uses of AI and machine learning? Or are we doing stuff inspired by large tech companies?”
The panelists included: Emilia Díaz-Struck, research editor and Latin American coordinator, International Consortium of Investigative Journalism. Francesco Marconi, computational journalist and CEO at Applied XL. Jarno M. Koponen, Head of AI and Personalization at Yle News lab. Katie Bremme, Executive Product Manager Innovation, France TV, and Wilfried Runde, Head of Research and Cooperation Projects, Deutche Welle.