Can AI programs be trusted to report the news?

As local newspapers grapple with dwindling budgets and overloaded journalists, some newsrooms are experimenting with an idea that skeptics say threatens the very role of reporters: integrating artificial intelligence into the newsroom.

Editors remain cautious about the use of AI in reporting, one major reason being that it cannot distinguish between fact and fiction. But used responsibly, they say, it can provide a cost-effective toolkit to ease the load on local journalists while augmenting their coverage – for instance, with AI-produced summaries of city council meetings.

Silicon Valley AI firm OpenAI helped fuel the interest – and debate – related to AI writing and reporting when it released its conversational chatbot, ChatGPT, in late 2022. The AI-powered program can quickly respond to text commands and then write essays, summarize books, and produce financial reports. Its release garnered national attention – and additional funding from Microsoft.  

In California’s Humboldt County, 300 miles north of Silicon Valley, Hank Sims and his local newsroom, the Lost Coast Outpost, started experimenting with ChatGPT last year. The online-only newsroom used the program to develop its own version, dubbed LoCOBot. The program downloads and summarizes agendas of local public meetings.

Mr. Sims says LoCOBot replaces the human need to scan through scheduled lengthy agendas of city council and other meetings, and frees up reporters to investigate larger stories.

“Our reporters love it because it notifies them automatically when a new agenda is published and it provides a sort of quick, you know, at-a-glance view on our own side of what the council’s going to be talking about,” says Mr. Sims.

LoCOBot is open to the public and easily accessible through Lost Coast Outpost’s website. Beyond summarizing the agenda items in concise, professional language, LoCOBot can elevate dry, bureaucratic prose into humor and verse.

For example, a City Council meeting agenda for the small town of Arcata had its Cinderella moment when LoCOBot transformed it into a bedtime story: “Once upon a time, in the City of Arcata, there lived a Finance Director named Tabatha. She was a hardworking and dedicated employee, who was responsible for overseeing the city’s budget.” 

In all likelihood, the finance director is indeed “hardworking and dedicated.” But of course, LoCOBot never interviewed the director or looked into her work record – making that amusing fairy-tale presentation a bit of a cautionary tale about AI’s shortcomings. Many experts say that real journalists who ask questions and follow leads still have an essential role to play in unbiased reporting of the news and uncovering of the truth.

Conversational chatbots, or large language models like ChatGPT, have no way of distinguishing between true and false, explains Nir Eisikovits, director of the Applied Ethics Center at the University of Massachusetts Boston. He says they cannot be trusted because there are no built-in guardrails. To use AI responsibly in newsrooms, a human reporter is always needed to double-check AI-produced work.

“I think there’s a great irony that journalists – the profession whose reason for existence is uncovering the truth – would rely on a tool that has no capacity to distinguish between truth and falsehood,” says Mr. Eisikovits. “The more important the story and the higher the stakes, the less I would rely on AI.”

Despite such concerns, the Lost Coast Outpost has not had any problems with misinformation or the accuracy of LoCOBot, according to Mr. Sims.

This is mainly because LoCOBot has simple, direct commands to summarize texts, not to generate open-ended responses. He adds that their publication is cautious and intentional with bylines – transparently labeling work performed by the program. 

As part of the ongoing effort to help local newsrooms integrate automation and AI technology, The Associated Press and other news organizations are funding the expansion of AI technology across the country, from Michigan to Puerto Rico.

In Michigan, reporter Dustin Dwyer is responsible for covering nearly the entire west side of the Great Lakes State for Michigan Radio’s WUOM-FM.

When newspapers were thriving, Mr. Dwyer relied on local news as a resource. But as newspapers dwindled, Mr. Dwyer was stretched thin and his coverage became strained, he says.

“For example, in Cedar Springs, Michigan, their little newspaper just closed down,” explains Mr. Dwyer. “There’s no one covering Cedar Springs City Council meetings right now and I can’t get there. So is there another way to get coverage of all these other localities? Is there a way to still give people the tools they need to be informed about the meetings with the realities that we’re facing in local news and in staffing?”

For the last three years, Mr. Dwyer and fellow reporters at Michigan Radio have been using their own program developed in-house, called Minutes, to create transcripts of city council and subcommittee meetings for over 40 cities throughout the state. Minutes has transcribed over 5,000 recordings and made transcripts available to reporters.

Mr. Dwyer says that accuracy with AI remains a huge concern. Transcriptions help reporters search for key words and phrases, but Minutes is far from foolproof, meaning fast talkers and grumbled phrases are susceptible to mistranscription. And yet, Mr. Dwyer sees AI as a viable tool for reporters, not their replacement. “It allows a reporter like me to get a lot more coverage to the audience than I could just doing it by myself.”

With AP’s help, Michigan Radio will expand and improve Minutes’ audio transcription capabilities with OpenAI’s speech-to-text program, Whisper.

Renee Richardson, managing editor at the Brainerd Dispatch in Brainerd, Minnesota, is the latest trailblazer for AI integration in her local newsroom. Their AI experiment will launch in June.

By working with a developer to create an AI program to automate public safety announcements from police blotters, Ms. Richardson hopes to maximize efficiency for her staff’s workflow and give something invaluable back to the Dispatch’s reporters: time.

“We’re constantly asking our staff to do more and provide more information in many more ways. Whether that be social media, video podcasts, audio segments, all of our photography, or all those pieces that go into it. Rarely do we do much that gives them time back. The benefit I see for this is finally giving them that time back.”

And the idea that AI might one day turn reporters into an extinct species?

“When you think, ‘Is this something that would take over what I do for my job?’ the answer for us is that, no, that’s just going to free you up to do even better, more impactful work.”