#WorldPressFreedomMonth: Dhrumil Mehta Discusses AI's Impact on Journalism

What is the Tow Center’s role in today’s media environment?

At The Tow Center, we do research in digital journalism. Particularly, we're interested in media environments. We have research projects particularly focused on AI at the moment. For instance, we've been looking at the impact of AI search on publishers. So, companies like OpenAI have been releasing search tools that they're positioning as competitors to Google.

But when you search for something on a chatbot, the results are a little different than what you would get when you search on Google. And because these tools are so new, we're doing a bunch of research to better understand what sorts of answers it's giving you, how often those answers are correct and also why it's giving you the answers that it's giving you versus what you would get on Google and what impact does that going to have on publishers? Because ultimately a lot of this material comes from journalists, particularly when you get up-to-date answers on things that are happening, that is reporting that journalists have done.

And so, how is ChatGPT, for instance, citing publishers when it publishes, when it produces an answer for a user? These are the kinds of questions we've been interested in. We've also been looking at something we call pink slime journalism. So, as these AI tools progress, there are, of course, use for all kinds of efficiencies in journalism. We're using them for data analysis, for data extraction. We have many projects that we're working on right now that utilize AI tools to do more and better journalism than we were able to do before.

However, there are also nefarious actors who are publishing all kinds of junk. You may have heard the term AI slop. So there are actors who are publishing all kinds of junk, a lot of it factually inaccurate, not having been fact-checked, And just millions and millions of articles, sometimes intended to influence a population, and these articles are often AI-generated. So we are tracking networks of organizations that are doing this sort of thing and looking at the amount and kind of things that they are publishing.

We both utilize AI for a lot of things, but we also take a critical journalistic lens, ask these important questions about how the tools work and whether they are working to the benefit of society and also what impact that's going to have on the industry.

What resources does the Center offer for CJS students? 

The product of having these really advanced data programs is there are ample opportunities to get involved with understanding AI, utilizing it, reporting on it.

And not just as a consumer, but also as somebody who can sort of open up the guts of it, look inside, build things with these tools, do theses that are accountability projects that look at how these tools work. So we're always here in this corner of the school, the Tow Center and the Brown Institute, and we're always doing this kind of work.

If a student decides to do a thesis that sort of heads in this direction, they might come to us for advice. They might be able to get involved in some of the research that I had mentioned.

And we go to classrooms. So, for instance, if you're in the general M.S., we might come to your class and talk to you about how these AI tools work.

If you're in the data M.S., you'll probably use these tools every day. And if you're a Computer Science and Journalism Dual M.S. student, you're probably going to be analyzing or building, or doing something completely novel with these tools.

What role can AI play in journalism? 

One of the most important functions that we have is unearthing information that hasn't been unearthed before. And so, whenever there's new technology, my initial thought is automate the boring stuff, find the things that are boring, repetitive, automate those away so that we can really focus on the part that involves showing up in communities, talking to people.

These are things that an AI is not ever going to be able to do. And then on the other end, when we're talking about acquiring data, for instance, we’re working on a project now where we're writing dozens and dozens of scrapers. And that was something that would have taken a long time, a few years ago. But now we can have AI write 50-100 scrapers for us.

So let's suppose you want some data from every hospital in the country or every county. These are things that we would have had to write a different scraper for every different website a couple of years ago. But now these tools have enabled us to do these ambitious data collection projects for the purposes of informing the public.

And so I think there's a lot of uses of these AI tools that's not, “Hey, let's generate some articles,” but rather we can use it in data acquisition, we can use it in analysis, we can use it in synthesis, all the while freeing up more of your time as a journalist to go and engage with the community, talk to the people who are impacted, speak with experts, etc.

There are a lot of good uses for these tools in our practice.