251: AI Project Tackles Online Misinformation

ROEBRT JOHNSON:

This is Public Health Review Morning Edition for Monday, August 29th, 2022. I'm Robert Johnson with another special edition of the newscast recorded during our recent visit to Atlanta. It's brought to you by the Association of State and Territorial Health Officials.

Recently, hundreds of public health communicators gathered for the 2022 National Conference on Health Communication, Marketing, and Media. That's where we caught up with Connie Moon Sehat, a researcher with Hacks/Hackers, a nonprofit focused on journalism innovation. She's working on an AI project that will help organizations evaluate and respond to social media, misinformation, and disinformation.

In the context of social misinformation and disinformation the last few years of the pandemic, how big of a problem has it been? I think we all know the answer is a really big problem—but in your words, how big of a problem?

CONNIE MOON SEHAT:

It is, unfortunately, a really large problem. You know, as we know, for example, maybe this is the best example out of an MIT study.

And we know, for example, that false information or false rumors can spread six times faster than factual information. And so, if you think about how fast and how far a false rumor can spread, that's really problematic.

And so, consider then COVID-19 related information and how much has come out over the last years. It is exponential.

JOHNSON:

You are a researcher. You look at news online and elsewhere.

You're part of a project that is trying to tackle this issue. Tell us about ARTT.

SEHAT:

So, the Analysis and Response Toolkit for Trust, or ARTT, is a toolkit that's trying to build really one central software tool and that's to help people in moments when they're online to try and figure out, "Hey, I've come across a really difficult piece of information, and I want to have a conversation with somebody about it." What kinds of things can experts do to give us guidance about how to understand that information and what kinds of things do experts have to say about? What options do I have about how to best respond to that information, whether that's correcting someone or maybe it's just about listening and empathizing?

JOHNSON:

How does it work? I saw your presentation. It looks like it does a lot of the thinking for you.

SEHAT:

So, what happens is that you can take, for example, a tweet, a link to a tweet, or you could take—you can copy and paste the text of a news article, for example, you put it into the tool.

And what it does first is it analyzes it through several different lenses. So, it tells you how trust and safety professionals think about it. Like, is this post going viral? It will tell you things about how journalists think about it. Is this a high quality piece of information? Is it linked to maybe some sketchy sources? Also, it'll help you understand what psychologists think about it. So, for example, is there a manipulative language happening in the background?

And then, because we know that sometimes in these conversations you want to have something good to share, like some positive information, we want to give you a list of reliable sources but that are also local. Because I think that one of the things we're understanding about this whole problem is that not all news or information fits equally well in terms of sizes. Sometimes, you want to hear from your local community members.

And finally, it gives you nine different ways to respond. So maybe, you know, what do researchers say is the best way to correct someone if you want to do that. But then, if you want to actually kind of figure out what does it mean to empathize, here are some tips and examples.

JOHNSON:

So, it actually helps you come up with the words that you might use to engage that post.

SEHAT:

So, you'll be able to have a list of different kinds of example responses that have either worked in like real life or things that a template, for example, that you can use to copy and paste and build your own sort of response.

JOHNSON:

So, it does not replace human judgment.

SEHAT:

Oh no. So, I mean, I think actually one of the things that we wanted to really respect in all of this, right, is that we know that the people in the middle of the conversations have the best perspective about the context about the people they're talking to. So, you know, sometimes like they're going to know, you know, this is not a good time to correct. It's a better time to listen, it's a better time to ask more questions. Those are the things that we know that a local person, a person in the middle of conversation, is going to have that expertise that no tool will ever have. And especially combined with public health communications expertise, you know, that's just going to become all the more powerful.

So, we want to just help their decision making, or help your decision making, when you use the tool. How can we give you more superpowers, if you will, to kind of have a stronger conversation.

JOHNSON:

And you're looking for input from public health communicators to make the tool better.

SEHAT:

Yes, absolutely. So, right now we are in a prototype development stage, or basically this next year through basically May 2023. We're hoping to have our very first public launch.

And so, while we are building the tool, we want to have partnered with public health communicating groups to help us understand, "Is this actually working for you? Is it not? Let's go ahead and make something that actually feels like it's practical and it's going to help you in real time."

JOHNSON:

You can read more about the ARTT Project using the link in the show notes.

Also, if you liked the newscast, please take time to leave us a rating and a review. We really do appreciate the feedback.

That'll do it for today's newscast. We're back again tomorrow morning with more ASTHO news and information.

I'm Robert Johnson. You're listening to Public Health Review Morning Edition. Have a great day.