A well-known example of artificial intelligence being used in investigative journalism is commonly called “the Panama Papers leak”. The leaked documents became searchable on the website for the International Consortium of Investigative Journalists (ICIJ) in New York.
"They had to go through receipts and invoices to trace the money. Who it came from and where it went," says researcher Gustavo Mello at OsloMet Artificial Intelligence Lab.
Machine learning was incredibly useful in that context.
"This receipt-like text is normally semi-structured. These receipts aren’t structured like ordinary text in a news article for example" Mello explains.
The way the information in the receipts was set up meant that artificial intelligence was a suitable means of sorting the data.
"AI can quickly identify these patterns, retrieve information, and set it up in tables to make the journalist's work easier," says Mello.
Tracing the flow of money
Professor Roy Krøvel at the Department of Journalism and Media Studies has also researched the flow of money in the research project Making Transparency Possible, which was what really brought home the importance of incorporating artificial intelligence in investigative journalism.
"What I realised was that corporate structures are so complex that there can be hundreds of subsidiaries that regularly change their registration, move, and restructure, almost on a weekly basis. They can do so by using pretty advanced artificial intelligence," says Krøvel.
The tax authorities and journalists may also have to use artificial intelligence to keep up with the practices these companies use.
Journalism is useless if you don’t know how you found the answer – Roy Krøvel
"These companies’ practices make it unrealistic, if not practically impossible, for most tax authorities to keep up with or understand what is going on. It’s almost impossible for ordinary people to get their heads around it," says Krøvel.
The banking system already uses artificial intelligence to notify Økokrim – the Norwegian National Authority for Investigation and Prosecution of Economic and Environmental Crime.
"I think they receive between ten and fifteen thousand notifications of suspicious transactions every year. These notifications are often based on artificial intelligence. The transactions they look at are often part of a bigger chain of transactions. It’s impossible for Økokrim to investigate all of this without using artificial intelligence themselves," says Krøvel.
Artificial intelligence in investigative journalism
The research project is a collaboration between the Department of Journalism, Department of Information Technology and the AI Lab, and brings journalists, engineers and technologists together to make investigative journalism more efficient.
Does that mean we need fewer journalists?
Roy Krøvel teaches in the journalism programme and recognises that more teaching about artificial intelligence would benefit the programme. However, he doesn’t believe that using artificial intelligence will necessarily mean fewer journalists.
"We’ve actually been through a phase in which the number of journalists has decreased. But that’s due to other things. In recent years, a lot of money for advertising has left Norway and gone to other actors such as Facebook and Google. That’s left us with less money for proper journalism. It’s a big problem," says Krøvel.
Nor does he believe that auto-generated articles are now set to replace the work of journalists.
"Artificial intelligence will not replace human journalists for a long time to come. I hope it doesn’t happen at all, and I certainly don’t think it’s going to happen as a result of this research project," says Krøvel.
Gustavo Mello instead believes that artificial intelligence will prove a useful tool for journalists.
"Our philosophy is that reliable artificial intelligence is trustworthy, works with us and is not something that will replace us," says Mello.
"The idea is that we will develop tools that can increase the potential and speed of journalists’ work and automate processes that are laborious and boring."
How it works
"Most of the projects we look at will analyse text data from financial reports, government documents, public records etc. to identify patterns that may uncover problems and irregularities," says Mello.
As well as identifying patterns, artificial intelligence can:
automate the division of text into categories to speed up work
make text summaries based on topic sentences or other relevant information in the text.
divide the text into groups once patterns have been found, which enables the journalist to see the difference between the groups more quickly, and whether it is relevant to what they are investigating.
Journalists can generally save a great deal of time when they have to go through a lot of text.
There are already a lot of good tools around for using artificial intelligence, but most of them have been developed for the English language. The way we combine words in Norwegian differs from English. The same is often true for the length, meaning and writing style of a particular text.
"The plan is to make the English tools also available in Norwegian and adapt them for journalistic purposes. We can then expand them and develop new ones if they are not sufficient," says Gustavo Mello.