Arash Nikkhah
UW-Madison assistant professor of journalism and mass communication Tomas Dodds.
New journalism prof Tomas Dodds has launched the Public Tech Media Lab.
Perhaps it’s the same for you: AI is coming up in nearly all my conversations with friends and acquaintances.
There’s the English professor who is grateful that her recent retirement means she will not have to deal with the ethics of AI usage among students. And the investment analyst who used AI for the first time to write her own code to predict market trends. And the CPA who told me she predicts AI will be used within two years to prepare tax returns.
AI is also a hot topic for discussion among those of us who work in local news. Editor & Publisher, a trade magazine covering the news industry, recently emailed a survey about AI practices to news organizations. The first question is, “Has your organization published a formal policy or guideline related to the use of artificial intelligence?” Also, “Has your organization created an internal task force, committee, or working group focused on AI?”
It plans to use the survey to better understand how “news organizations are adopting, governing and experimenting with artificial intelligence — from newsroom tools and policy decisions to audience trust and future planning.”
AAN Publishers, one of the trade groups Isthmus belongs to, is also surveying its members on what they would like to see included in a webinar series on AI. Among the options: “How to craft an AI policy for your newsroom,” “How to use different AI tools at work,” “Best practices for using AI tools,” etc.
I’m helping plan these sessions but I also filled out the survey: It was a “Yes” to all of the webinar possibilities for me.
We might have some help with these training sessions from a new assistant professor of journalism and mass communications at UW-Madison. Within weeks of arriving in Madison, Tomas Dodds has already launched an exciting lab on campus: the Public Tech Media Lab. Dodds, a native of Buenos Aires, was happily working at Leiden University in the Netherlands, where he was a research fellow at the AI, Media & Democracy Lab and the Institute for Advanced Study, when he saw a job opening at UW-Madison’s J-school.
In an interview, Dodds says he decided to apply because he admires many on the faculty here. Accepting the new position required uprooting his life, but it was “too good of an opportunity” to pass up.
According to Dodds, a main goal of the Public Tech Media Lab, which already counts faculty associates from around the globe, will be to teach journalists how to use open source technologies to create their own AI systems that align with their values and needs. The idea is to make newsrooms less dependent on big tech companies that have their own private interests.
“AI is becoming the infrastructure over which newsrooms are being built,” he explains. This is true not only for producing and distributing news, but also for the advertising side of the business. “And so newsrooms are putting themselves in a position where they’re becoming hyper dependent on these companies.”
I asked Dodds for an example of a project the Media Lab could help newsrooms take on. How about a personalized, in-house large language model of ChatGPT that would help reporters easily access all of the coverage their organization has done on a given topic or issue?
“If you can train an LLM with only your own data, you can ask it questions about specific topics,” he says, including “How have we covered this topic in the past?” That saves reporters a lot of time preparing the next story on a given topic and makes sure that they are “following the editorial line” of their newsroom, he adds.
It would also allow newsrooms to review their coverage over time for bias and sourcing. “You can actually see the entire history of your stories [on a particular topic] and be like, ‘Oh, actually, we have a tendency to only interview men.’” Dodds says newsrooms could then train their own AI systems to, for instance, include more women sources in these types of stories.
“That is really, really nice and really important, and something that you only can do if you are actually training your own models, rather than using an off-the-shelf tool like ChatGPT.”
Creating your own ChatGPT sounds like a big project but, Dodds promises, “It’s not that difficult when you have academia and industry collaborating together.”
As Dodds sees it, AI systems are a “great tool and companion for some things,” but are also “extremely dangerous when you don’t know what it is that you’re using or don’t understand how these models have been trained.” It’s the role of media organizations to teach these systems, but many don’t and few newsrooms have policies for AI. And some of those who do have simply cut and pasted a policy from another organization.
He says some of the recent stories we’ve heard about AI mishaps — including the Wisconsin State Journal article that included made-up information attributed to a violation of the paper’s AI policy — reflect more of a systems lapse than an employee misstep. “What we are seeing here is organizations’ failures in training journalists,” he says.
Now is the time for newsrooms to come together to discuss what AI tools they might use and how they jibe with their values and mission. Dodds believes it’s important to have a policy in place and to make it public. It makes your relationship with your audience stronger and more transparent.
“Even if you don’t want to be restricted [by a policy] you need to make sure everyone in the newsroom knows what you’re talking about when you’re talking about AI, what the dangers are,” he says. “Otherwise, we will keep seeing these stories and maybe worse.”
Dodds and his lab are here to assist. In fact, he will be helping Isthmus facilitate our first official conversation on AI usage in the coming months. I know some of our staff members have played around with using AI to brainstorm headlines — to my knowledge we have not ever published one verbatim but used the suggestions to spur ideas — but that’s about it. So we welcome the assistance.
“The idea of this lab,” Dodds says, “really is to work with journalists, elbow to elbow, trying to see how we can create systems that for them are useful, that they want to use, that align with their values.”
