How AI Could Act as Boost for Investigative Journalism

2024-01-10

源 稿 窗
在文章中双击或划词查词典
字号 +
字号 -
 折叠显示 
 全文显示 
When The New York Times reported on Israel's bombardment of Gaza, it relied on the latest technology to reinforce its findings.

The paper's journalists used artificial intelligence to track satellite images that showed more than 200 craters in densely populated civilian areas, which experts say are likely caused by 907-kilogram (2,000-pound) bombs.

The Times' use of AI shows how the technology can assist media in quickly and accurately processing large data sets. More importantly, some analysts say, it could herald a new era of smaller newsrooms being able to take on investigative work.

Jared Schroeder, a journalism professor who specializes in emerging technology at the University of Missouri's School of Journalism, said it would have taken an army of journalists a long time to examine and isolate that much satellite data.

"That story is almost not possible without AI," Schroeder said. "It would be very difficult for humans to be able to count bomb craters off of satellite imagery and do it accurately. And so, AI almost makes that report possible."

Investigative journalism is a costly enterprise at a time when the industry is in financial decline. The use of AI offers a cost-effective way to aid investigative reporting. And while technology has long been used to aid journalism, the arrival of Open AI's ChatGPT has renewed focus on the possibilities that AI offers.

"Projects like this have been done for years. There was one that came out of Ukraine in 2018, and it was a textbook case of how to use machine learning and AI to analyze large amounts of satellite data," said Jonathan Soma, a professor of data journalism at Columbia University's School of Journalism.

That investigation revealed that tens of thousands of Ukrainians were illegally mining hundreds of millions of dollars' worth of amber.

"This was so technical it took a multisession course to learn how to do all the technical bits and pieces. But nowadays, I teach people to do this in 10 or 15 minutes," said Soma, adding that the simplicity is a "game changer."

Perhaps the best-known case of AI assisting investigative journalism is the 2016 Panama Papers case, when reporters trawled through 2.6 terabytes of leaked information.

The science news website LiveScience calculated that the leak would be the equivalent of 650 million pages.

"When we think about something like the Panama Papers, that was really revealed through an AI technique called a vector data base," said Ryan Heath, global technology correspondent for the news website Axios.

By using AI as a tool, news organizations and reporters were able to organize and use the Panama Papers' data to expose the offshore accounts and tax shelters created by some of the world's wealthiest people to dodge taxes, commit fraud and avoid government sanctions.

"What [AI] allowed the people involved in that investigation to do was to surface things no single human would be able to piece together," said Heath "When you have hundreds or thousands of shell companies operating and in hundreds of locations around the world changing their structures like chameleons, no one person can keep on top of that."

News equalizer

A more important element is the potential for advanced AI to act as an equalizer for small news organizations.

Significant investigative journalism was too costly and time consuming for most news organizations to tackle just a couple of years ago, said some media experts who spoke with VOA.

But everything is changing rapidly, said David Caswell, who has worked in journalism and technology, including at the BBC and Yahoo.

An AI expert, Caswell is currently working at the Oxford, England-based Reuters Institute for the Study of Journalism, training newsroom leaders on how to integrate AI in their outlets.

"With AI for data journalism, the impact is going to be a much broader impact of accessibility to data journalism - like with smaller newspapers, like in rural places - being able to do data journalism for the first time," said Caswell.

"Collectively, much more important, it is things like being able to analyze lots and lots of unstructured text documents," he said.

One example, Caswell said, is the ability of large language models to comb through thousands of police reports written in a way that couldn't be used as data previously.

"Extra magic"

Despite the advances, academics such as Soma say that technology will never replace journalists.

"The root of being a good journalist is doing journalism: interviewing, talking to people, all of those things," said Soma. "Knowing which stories to report out. And none of that really needs AI. AI is just a little bit of extra magic that you can sprinkle on top of traditional journalism skills."

While some media analysts have questioned whether AI will damage the credibility of journalism, most agree that it will never fully replace the role of reporters and instead can help enhance their role as public watchdogs.