Madison journalists debate AI bots versus human touch

Isthmus co-hosts forum on artificial intelligence and journalism

by

News

The panel on AI & Journalism co-hosted by Isthmus could not have been more timely.

“When we started planning this event, we were not anticipating the headline of this week: that AI might be our ultimate undoing,” said Isthmus editor and president Judith Davidoff in kicking off a Sept. 17 community conversation on the implications of artificial intelligence in journalism.

Hosted at Cargo Coffee East, the panel also featured Omar Waheed of Madison365 and Mike Wagner, director of UW-Madison’s School of Journalism and Mass Communication. Angela Russell, host of the Black Oxygen podcast and founder of consulting practice The Undercurrent, moderated.

In her greeting, Davidoff was referencing a round of recent media coverage sparked by Jacob Coxen, a former staffer at OpenAI, maker of ChatGPT, and of Anthropic, maker of Claude. Recently, Coxen posted on X that “Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives.”

“The people building AI earnestly believe that it could kill us all by the end of the decade,” he added, and the national headlines went wild. Yet existential fear of AI is part of why both OpenAI and Anthropic exist. OpenAI was founded with the explicit premise to "benefit humanity as a whole.” Anthropic was formed by a group of ex-OpenAI “Doomers,” who broke off to build what was, in their view, a more ethical alternative.

Those within the tech industry compare their P(doom) numbers: the probability that AI will lead to catastrophe. Thursday’s discussion had a narrower focus, examining the P(doom) for local journalism. As Davidoff said, “We were focused on a more narrow discussion, which is how AI is impacting local newsrooms, what people see as the benefits, the risks, [and] how we’re all grappling with it.”

Panelists warned that a key risk of AI is that publishers would erroneously see the technology as a way to enhance productivity and reduce staff, leading to even more employee cuts and lower-quality journalism.

“We are told the AI revolution is here and cannot be stopped,” said Wagner. “This is a claim that serves a small segment of very wealthy and powerful people, and we have no reason to believe that it’s true. But if we do believe that it’s true, it serves the interests of the very wealthy and powerful people, so they can hire fewer journalists, and produce more things faster, even though all of it will sound the same.”

The panel was sparked by Isthmus’s own investigation into a local reporter’s prominent use of AI in a Wisconsin State Journal story that contained a hallucinated source and inaccurate information. The reporter, who was fired, later outlined to Isthmus the mix of factors she said led to the inaccurate story, including that she was new to Madison and assigned a story outside her beat. She also noted a breakdown of editorial oversight, and lack of training on the AI software on her company computer.

Isthmus’ reporting led the newsroom to connect with Tomás Dodds, who researches AI in journalism and directs the Public Tech Media Lab at UW-Madison, and to develop a policy on AI, which applies to all aspects of its work.

Adopted in August 2026, the policy allows for limited uses of AI, including transcription software, but prohibits using the technology for writing or editing stories or creating images. During the panel discussion, Davidoff cautioned that the “the worst possible scenario” resulting from AI use was for newsrooms to emphasize quantity over quality — demanding staffers do more through the use of AI, and producing “slop” in the process.

Journalists have had to “shape-shift” to emerging technology, noted Russell in her first question to the panel, and a pattern has emerged over the last century: the emergence of new media that threatens to displace its predecessor. Wagner can list them off quickly: radio threatened to kill newspapers, television threatened to kill radio, cable threatened to kill broadcast television, the internet threatened to kill cable, and social media threatened to kill the internet.

“We still have radio. We still have newspapers. We have all of it. They have different power than they used to, and different utility than they used to. But we have all of it,” he said.

Each evolution, though, complicated the landscape. Panelists pointed out the limited use cases where AI can be useful. For example, Isthmus staffers use Otter.ai to transcribe long interviews. So does Waheed, who added that he pays for a more-expensive tier to avoid transcripts being used as training data — the service’s free, pro, and business plans are used to train its AI, but enterprise plans are not.

Davidoff acknowledged that Isthmus has not experimented with using AI to draft articles, and that she “doesn’t want to trash it without having tried it.” Still, she warned that using AI for the more creative processes of writing could atrophy creative muscles. “I just think it’s a lazy way of writing, and you lose creativity,” she said. “If you continue doing it that way, you lose the ability to write.” As for those who didn’t learn those skills before the advent of AI, she fears “we could lose a whole generation.”

Waheed said he did not buy the claim that journalists need to acclimate to AI. “It’s optional. It’s completely optional. And when we think about trust in the media, does the lying machine make you trust the media more?” Waheed attended a series of trainings through the Online News Association last year as part of a Madison365 quest to see if AI was worth an investment.

They decided it wasn’t. “Anything it can do we can do… the human component is important for us.

“The bot is not showing up to the grieving family’s wake after their child tragically died,” Waheed said. “They’re not going to replace me showing up to your community meeting. They’re not going to replace me showing up to the protest. They’re not going to take photos. They can’t do everything I can do…It’s a program.”

“They live in a computer,” Waheed added. “I live on the south side.”

Still, AI can compound the reputational problems already plaguing the field of journalism, including a general lack of trust from the public. Said Davidoff: “The profession has been demonized, from the president on down.”

She took specific issue with the proliferation of public relations professionals who are, in her view, increasingly emboldened to act as gatekeepers to sources and information. Doing so leads to “not very interesting responses,” which can dull a story.

Nationally, there are now more than six public relations professionals for every one working journalist, according to a newsroom index from Everything PR released in August. Those professionals can outflank reporters and lead to a poorer product.

“It’s now become customary to not answer reporters' questions, demand questions before an interview, only return a statement when you have follow-up questions,” said Davidoff. “Frankly, I think much more than AI, that's impacting…local journalism.”

But when journalists do write something, people might not trust it. Last year, Gallup reported that 28% of Americans trust the news to report “fully, accurately, and fairly.” (Republican trust is in the single digits).

They might even trust algorithms more, erroneously believing that those are “neutral,” according to two studies recently conducted by Wagner and colleagues on “motivationally relevant messaging.” The study was funded by a federal grant focused on disinformation that has since been canceled.

In the studies, Wagner and his team obtained a list of Facebook accounts whose history suggested they were skeptical of vaccines, then ran ads recruiting them to participate in a study. When they showed participants different content about HPV vaccines and asked whether they’d be more likely to have their kids vaccinated, participants were more likely to trust messaging if it was disclosed that the information came from an algorithm.

Yet an ample body of research shows that algorithms — which are built by humans — reflect the biases of their creators. "Any algorithm is a product of the information that's fed into it, and the directions that you give it,” Wagner tells Isthmus.

The event’s program promised time for questions and audience members were ready. “How do you ensure your stories are trustworthy?” asked one.

Some measures are “basic,” and even “old-fashioned,” said Davidoff, noting that Isthmus has established practices for interviewing, sourcing and verifying information. The paper is also accessible and responsive when told about an error. Corrections are noted on stories with an “editor’s note.”

Still, journalists could be more explicit when it comes to their methods and practices, said Waheed, adding that they have “operated in a realm of mystery.” Waheed said he tries to demystify the process when dealing with sources. (Madison365 publishes a cheat sheet for sources preparing for an interview.)

Waheed added that he is a card-carrying member of the Society of Professional Journalists. The group’s Code of Ethics directs journalists to, in part, “explain ethical choices and processes to audiences,” and “encourage a civil dialogue with the public about journalistic practices.”

But methods and practices can and do vary outlet-by-outlet and even person-to-person. And there is no credentialing body for reporters.

 “You don’t get called in front of a board with the ghost of [former Washington Post publisher] Katharine Graham looking down upon you,” said Wagner. “You can’t lose your license to be a journalist.”