Episode description:

Welcome to In Reality, the podcast about truth, disinformation, and the media with Eric Schurenberg, longtime journalist and media executive, now the founder of the Alliance for Trust in Media.

If you want to understand how we got to this moment, a moment in which most Americans say they can’t tell what’s true online anymore, when loud minorities can manufacture the appearance of mass consensus, when we have retreated from a shared sense of facts to our own bespoke realities, let me introduce you to Renée DiResta. Renée has been a warrior for truth in these polarized times longer than almost anyone.

She mapped anti-vaccine networks on Facebook in 2013. At the Stanford Internet Observatory, she exposed state-sponsored influence operations and for her trouble, she has been slandered before Congress, subpoenaed and doxed and harassed. Her book, Invisible Rulers, is an encyclopedic account of what happened to us all when the ancient human rumor mill met the modern machinery of networked propaganda. She’s now a professor at Georgetown’s McCourt School of Public Policy and she’s still very much in the fight.

Renée and Eric talk about how ordinary people get pulled into false narratives, how it feels to be targeted by powerful people living out fake and cynical narratives, and how AI is rewriting the rules of influence. And why, despite everything, Renée hasn’t given up the fight. And now here’s Renée de Resta.

 

Transcript

Eric (00:01.698)
Renée DiResta, welcome to In Reality. It is really great to have you here. Now, I first became aware of you when you were at the Stanford Internet Observatory, which was an amazing organization that existed to pull back the curtain on all kinds of online manipulation. So you have really been one of the OGs in this field of information integrity.

Renee DiResta (00:04.777)
Thank you for having me on.

 

Eric (00:29.582)
After years of very important work, the ISO was shuttered, partly in response to harassment from powerful people like Congressman Jim Jordan. You’re now at Georgetown. So what’s your mandate there?

Renee DiResta (00:40.243)
Yeah. Also, I’m a professor in the McCourt School of Public Policy. So I have a faculty appointment now, which is great, but a little bit strange. And I work on a variety of things. So a lot of it is, how do we fix some of the problems? I used to work a lot on detection. I wanted to shift a little bit upstream and work on design, platform design, tool design. Right around the time that SIO was

being dismantled, I got very involved in protocol based platforms. Blue Sky was becoming popular. You recall after, for example, Elon bought Twitter, there was this Exodus. All of sudden there was kind of critical mass on protocol based platforms. So I started doing a lot of work on what that meant, what decentralization looked like. I work on things like, where’s the intersection of digital ID and privacy and agentic AI? How do we think about credentialing agents?

something I’d written about a lot in the past from a theoretical standpoint. Now, again, agentic AI is quite popular. You know, I use it almost every day at this point. So how do we think about credentialing agents as they move into the operating environment? How do we think about who is human online? What does it mean to be a person online? And how do we signal that while still protecting privacy? So I work a little bit on that. know, variety of these issues that are

Renee DiResta (01:58.461)
at the intersection of tech and policy where we can both build technology and write policy, which was always SIO’s spot also, but now again, more on the solution side, I think than the detection side. But I was working on an investigation last night, so that’s still where I, if I’m doing something for fun on my own, that’s still what it is.

Eric (02:15.854)
I’m sure it sounds like fun. Anyway, but I’m glad you’re still doing that kind of work. So let’s talk about policy and fixes and things like that also about agentic AI later on. But first, I would like to sort of pick your brain, your very experienced brain on.

how people can cope with the information environment right now. So starting with your beginning in this, where you kind of got interested in the topic as a new mom, getting hit with anti-vaccine ads on Facebook and finding that there were a number of normal people, your neighbors and people very much like you who had gone down a rabbit hole. What did that experience teach you?

about how ordinary people, not political junkies or inveterate conspiracy believers, get pulled into bespoke realities built around false narratives.

Renee DiResta (03:18.357)
So I think it’s two things. One, it is a lot about community and identity. That was the number one thing that I felt. My understanding of the anti-vaccine movement very early on, again, beginning in around 2014, 15, was around that. I was recommended, not ads, they were still pushing ads at that point, but I was recommended groups. It was when Facebook was trying very hard to use its group product to bring people,

new friends. This was around the time when they kind of tapped out your existing friends that you had come with. And so they were really pushing this groups feature. And I was very interested in how the recommender system worked. Funny enough, I’m building a recommender system on Blue Sky right now. So this question of how do you do that ethically? How do you ethically recommend content? And the interesting question there for me was, why was I getting recommended this stuff? And a lot of it is, we think you are like this. And I lived in California.

I cook, I made my own baby food. I was in a baby food group. I was in a baby sleep group. So I think it decided I was crunchy. And then it decided that it was gonna show me things that other crunchy people liked, ergo the vaccine stuff. And I actually created a whole other Facebook profile just so I could join some of these groups, because I wanted to see what happened in them. And then that profile got recommended literally Flat Earth groups, chemtrails groups. Eventually it got recommended Pizza Gate groups. About a year later, it was QAnon groups. By 2018, it was QAnon groups.

So you started to see how Facebook had built this kind correlation matrix around conspiracy theories that pushed people who had maybe some inclination into communities that were very, very active, right? These groups were very active, hundreds of thousands of people in some of them. And so they were constantly posting, constantly talking, constantly encouraging people to be activists. And that was one of the things that was really central to it, was that once you were part of these communities, if you got involved,

you became an evangelist for the cause. And there was nothing like that on the other side. Nobody was asking parents or pushing parents into pro-vaccine groups and saying, hey, you should be advocating for vaccination. You should be advocating for your community. Everybody just kind of did it and went on about their day. So there was this asymmetry there. So I think that was the, for me, the number one thing. How did the recommender system inadvertently kind of shove people with these nudges?

Renee DiResta (05:38.759)
into these groups, but then once they were in the groups, it was that they found community. They found other people who they saw as being like them that they considered trustworthy, more trustworthy than the health authorities. And again, the health authorities were not in the conversation at all. So that was the other big takeaway, that there was just this massive asymmetry where you were going to get your information from influencers, from community, not from authority figures and institutional voices. And that was what I started writing a lot about, that this division had already happened.

And then if you had the power to make something trend, you had the power to make it true for a lot of people that this was increasingly going to be a matter of capturing attention. And that was what I wanted people to understand.

Eric (06:18.574)
You wrote a lot in your book, Invisible Rulers, about the asymmetry of passion, kind of algorithmic and also kind of very human factor that allows a tiny minority, but a loud minority, to manufacture the appearance of mass consensus. I suspect that most people who were not part of the anti-vax minority who got pulled into it suspected that

Renee DiResta (06:24.351)
Yeah.

Renee DiResta (06:37.781)
Mm-hmm.

Eric (06:48.782)
where there was smoke, there was fire. And I just wonder if there are telltale signs, you know, that you are being manipulated by a loud minority, that there is a, what looks like a groundswell is actually being artificially pumped up. What would you look for to see that?

Renee DiResta (07:08.821)
So some of it is where is opinion polling, right? Which is an interesting question. At that time, this was again pre-COVID, this was in the 2015 timeframe, 2015 to 2018. What you would see was that, you know, people, even in California, vaccination rates were still, generally speaking, upwards of 85 % or so, right? Which is not high enough, just to be clear. That’s about the threshold where the MMR vaccine starts to lose some of the herd immunity protection, right? But at the same time, it’s at least up there.

Whereas if you were on social media, the conversation was 90 % negative. What you were seeing was overwhelmingly the anti-vaccine perspective. So almost a complete reversal of where the polling numbers actually were. So this is a thing that you start to see this reflected actually in quite a lot of issues. The phenomenon of people not wanting to push back against vocal communities is very real. The anti-vaccine movement in particular, I got very involved. I did start a pro-vaccine group and I got doxxed, I got harassed. And that question of,

is this worth it? Is it worth it to fight with these people when there’s a high personal cost makes other bystanders see what happens and say, like, is this going to happen to me if I speak up, if I push back? So that’s another dynamic with social media that the sort of worst behaved communities actually have the power to push other people out of the conversation and exacerbate that tension even further. But I think that question of, can you tell if it’s a small minority? Really, you have to try to get out of your bubble.

and actually kind of go do some research. And I use Pew a lot of the time, where is polling, where what do most people think about this, where is public perception? Because a lot of the time you also see your opponents are caricatured, right? This is a very common phenomenon on social media. Also, the most noisy people get the most attention. Often those are the people who are willing to be the most extreme to take a very extreme viewpoint.

And so the people who come to be seen as almost avatars of the right or the left are actually these very fringe perspectives, right? They are the most extreme voices in that ecosystem. They are not the moderate, quiet person in the center who doesn’t want to speak up because they don’t want to get attacked. And so you increasingly see this trend towards the extremes that also reinforces that dynamic. So going and saying, do most of the people who are not like me actually believe this?

Renee DiResta (09:32.465)
is a good question to ask in general. Do most of my enemies, so to speak, actually think that thing? I do go looking at polling just to get a sense of where are people at. And it’s not to say polling is foolproof, it’s not, but it’s at least a better representation than, this is trending on Twitter, ergo people must care about it.

Eric (09:51.017)
Uh-huh. All right, good. So a divergence between popular opinion as reported in places like Pew and what you’re seeing online is one red flag.

Renee DiResta (09:56.138)
Mm-hmm.

Renee DiResta (10:00.861)
Yeah, and this happens on the right and the left. I mean, you see, I remember getting involved in some education policy debates in San Francisco, and it was the same thing at a very micro scale. They eliminated eighth grade algebra. It was a whole thing. And the people who saw that as an equity issue were actually virulently passionate about it in the beginning.

Eric (10:21.442)
Yeah.

Renee DiResta (10:21.629)
And it made a lot of other people not want to push back, not want to say, actually, eighth grade algebra is good. We should have it. Which, in my opinion, is the most normal viewpoint in the world. I wrote this in the book, too, actually. Eighth grade algebra is good. Our kids should have access to it. But this was a thing where if you said it, people would come for you. And so the question became, how do you grow a community that is willing to push back against that, that is willing to be attacked, that is willing to.

Eric (10:48.524)
Yes.

Renee DiResta (10:48.773)
say the thing loudly and publicly so that other people see that they can say it too. And that’s where you start to see the pushback really begin to happen. And this is very common in highly polarized political debates. You actually do want people willing to state the other perspective just to balance that out, even though it does come with a cost of speaking up.

Eric (11:07.63)
I’d like to talk about that later because that takes guts. And you, among many other people who are disproportionately women in this environment, suffer a lot from boxing and harassment. And it really does take courage. I want to talk about content moderation, which is something that you have talked about a lot. Obviously, to quote the

the quote that’s attributed to Mark Twain about how a lie goes around the world while the truth is still putting on its shoes or whatever. The correction to a false belief rarely catches up. And you and I were at a conference recently where you were challenged on stage by an audience member about the alleged censorship on Twitter during COVID, which is a conspiracy that I thought would have been discredited a long time ago.

Renee DiResta (11:47.701)
Yeah.

Eric (12:06.392)
Can you walk us through the false narrative about the Twitter files and Murthy versus Missouri and explain why false beliefs like that persist?

Renee DiResta (12:16.341)
I was actually writing, I have an article coming out in Lawfare probably later this week, maybe early next week about Murphy because of this consent decree that just came down. They just got a settlement and this speaks to the asymmetry of passion. So to give a little bit of background, Murphy v. Missouri was a court case that the attorney generals of Missouri and Louisiana brought. Now the attorney general of Missouri, just to be clear, sued the Biden administration about

Eric (12:21.294)
I’m

Renee DiResta (12:41.365)
22 times, if I’m not mistaken, over the course of a year. So was just sort of a very frequent filer. He’s now the senator from Missouri, Senator Eric Schmidt. I wrote a review of his book. I would encourage people to read it. It really tells the story of sort of the mindset of this man. And it’s worth understanding it from his point of view. He sees lawsuits as a way to tell a story. And that is a very important thing to understand, because you and I might see lawsuits as a way to right a wrong. But in the world we live in today, filing a case oftentimes

is the important point, right? Because then you have a narrative. And oftentimes, the people who you filed against, the people who are defending, shut up and say nothing, right? They go quiet because they have to defend themselves now. So really, the plaintiff has kind of favored in this case, particularly if it’s a powerful voice like the attorney general of Missouri. So what you see happen is when this case is filed, right when media begins to write the story as if it’s already been proven, right? As if the case has already been decided. What winds up happening

is that he forms shops for a district judge that gives him his first injunction saying, the government clearly was too aggressive in speaking to the tech platforms. Funny enough, he says during the election, right? So this is about the Biden censorship regime. The Trump administration ran the government during the 2020 election, just to be clear. So there’s that. But then what winds up happening is he says, OK, well, then they move it to COVID. OK, well, what we’re really complaining about is COVID.

So then they’re suing the CDC, the office of the Surgeon General. They suck my team into this because we worked with the Surgeon General’s office during COVID in the form of providing them briefings. They decided that what that really was was a secret censorship cabal whereby the Surgeon General was telling us to tell platforms to take things down. So we get subpoenaed and by Jim Jordan, not by the courts, but by Jim Jordan. And so we turn over all of our materials and when it becomes clear,

that none of this, that no request from the Surgeon General or the CDC or anybody in this supposed cabal actually asked us to do anything. The CDC never came to us with any moderation requests whatsoever. They don’t change their narrative. They don’t say, whoops, sorry, we got this wrong. They just kind of pivot it. And they say, oh, well, maybe we just can’t find it, but this definitely had to have been happening anyway. So you can never exonerate yourself, you understand, because the narrative is already written. The point of the court case is just to support the narrative.

Eric (14:54.606)
Thank you.

Renee DiResta (14:57.673)
This winds up going to the Fifth Circuit Court of Appeals, which removes us from the injunction and says, actually, there’s no indication that, no reason why Stanford should be enjoined in this injunction. They’re not the government. That’s a violation of their First Amendment rights. And also, it looks like some of these other entities shouldn’t be enjoined either. They appeal this. goes to the Biden administration appeals. This goes to the Supreme Court then. And the Supreme Court actually looks at the evidence. So for the first time, you have a much more neutral

court, not the forum shop district judge, not the rather conservative Fifth Circuit Court of Appeals. But now all of a sudden you have the Supreme Court looking at it. And what Amy Coney Barrett writes is that there are clearly erroneous findings in the lower courts. When they actually look at the evidence, they begin to realize that there is no indication that the government defendants ever mentioned the plaintiffs.

So there’s no way that even all the discovery, all the depositions, there’s no evidence that any of this actually happened. So they toss it back. They toss it for standing. And they say, you have no right to sue. There’s no evidence that you’ve been harmed. There’s no evidence that the government caused that harm. So this gets tossed. So about a year goes by. And then the Trump administration becomes the government again, right? So the administration rolls. And then DOJ chooses to settle the case. So.

They go back to the remaining plaintiffs. Some of the plaintiffs now run the government, right? Jay Bhattacharya is now the head of the NIH. But the remaining plaintiffs who are still on the case receive a settlement. And it’s only for them, but it says that the CDC, the Surgeon General, and CISA are no longer gonna demand, or not gonna demand the censorship of these people, which they were never doing to begin with because it was tossed. That’s the entire point.

But they write it into the settlement, the DOJ, which again is now incentivized to talk about its prior administration as if it had been a bunch of villains.

Renee DiResta (16:55.527)
writes this consent decree as if it had happened, even though there is no evidence that it did. So just gives them the language that they need. There’s a sentence in there that says, this is not an admission of wrongdoing, this is not an admission of, this is not indication of evidence that any of this happened. So that sentence is in the consent decree, but it doesn’t matter because now they have a document, right? They have a document that says that the DOJ chose to settle with them. And so the attorney general of Missouri, now the Senator of Missouri,

trumpets this all over right wing media, all over his Twitter account, everything else, as if he’s won, as if he’s been vindicated, as if the theory of censorship that he advanced has been proven because now he has this settlement document in hand. Now again, it was not proven. This was not adjudicated on the merits. This was a incentivized administration choosing to settle with its ideological allies. But that is how the sausage is made. Now he has this document. And this is what you’re starting to see happen with some of these cases. So again,

In this particular instance, right-wing media, right-wing influencers, and even the DOJ’s press release tell the story about the Biden censorship regime being destroyed by the brave attorney general of Missouri. And mainstream media doesn’t even cover this because it’s a consent decree. It’s a settlement. There’s no there there. It’s like a participation trophy. So there’s that asymmetry of passion again. So you have the people who are most incentivized to spin this are doing that spinning while

the people who are most incentivized to talk about the facts, choose not to engage at all. This really wasn’t covered anywhere. This was, think Reuters wrote about it as descriptively, if I’m not mistaken, and Bloomberg, I think, wrote about it descriptively. But they don’t actually explain the context that I’ve just given you because it takes a long time. That’s the problem, right? When you’re explaining you’re losing, who wants to understand this? It’s really just people who have this deep, obscure domain knowledge and understand how this fits in.

to that broader censorship regime narrative and it is a huge priority for them to keep it up because it allows them, this is the other point, these are election deniers who brought this case, right? These are election deniers who have chosen to attack content moderation. These are election deniers who have chosen to sue and subpoena people. Jim Jordan, another election denier. So all of this was done in preparation to weaken and delegitimize moderation and relationships between stakeholders.

Renee DiResta (19:14.613)
ahead of 2024, 2026 and 2028. That is in fact the point of it.

Eric (19:18.702)
It’s outrageous. It’s also, like many truths, convoluted. And it takes a lot of explanation to understand it. For people who are encountering the press releases that these people would be putting out, say, well, we’ve gotten this decree. It proves that we were right all along. What are the questions that you should

Renee DiResta (19:29.523)
Yes, absolutely. Yeah.

Eric (19:48.053)
ask yourself, figure a normal consumer of news before sharing something, some kind of mental speed bump to raise your level of skepticism.

Renee DiResta (19:55.582)
You know

Renee DiResta (20:00.221)
In that particular case, most of the outlets that cover it are highly ideological already. This is not showing up in the bulwark. It’s not showing up in the dispatch. I actually really wish it was, candidly, because then maybe it sits in a place that sits in the semi-legitimate area. I would actually like to see media across the spectrum write about this, because then it would at least tell the story in outlets that

people trust, but if this is showing up in the free beacon or the daily wire or whatever, that’s already a partisan media property. It is very hard to persuade those people who have been hearing for three years now that this censorship complex existed. Meanwhile, on the left, they’re not gonna encounter this at all. It’s not even gonna be on the radar, right? It’s just not gonna hit their feed. Or if it does, they’re gonna be like, that nut job media outlet wrote about it. Whatever, I can dismiss it.

And so you have reality splintering ever further. Whereas the people in the middle, these are the ones that maybe you kind of want to reach them. The way I think about that now, and the reason I just wrote 2000 words on this last night is, how do you write for the LLMs? How do you write for the entities? So when they search for it and they get that synthesis at the top of the page, it’s not a one-sided synthesis. do you help the machines understand what has happened here? it is, as you know,

Eric (21:16.946)
huh.

Renee DiResta (21:28.605)
I have this, this, this essay is in drafts. I don’t want it to be 2000 words. I would love to cut 500 words from it. I have gone back and forth over and over again, trying to figure out what 500 words I can cut. But the fact is the details of the cases, because there were actually two consent decrees. There’s an even more convoluted case with the State Department is the other one. I write the facts out almost at this point so that other machines synthesize this.

for the people who are gonna take their takeaway from that little box when they type in, what is this consent decree? It’s a wild thing to have to do, but I agree about it in those terms. Or what is the article that you write so that the people who maintain the Wikipedia page on this topic have material to work with where they can understand what’s happening? Because again, that’s where people are gonna go. They’re gonna go to the Wikipedia article about Murphy v. Missouri. And you actually see this reflected because there are diehard ideologues who are in there editing that page.

The minute that consent decree came down, seven different Twitter files related pages were immediately updated with this finding as if it had been a verdict. They actually use the word verdict in some cases. And so you see the immediacy, the recognition that synthesizing information for the public is the ballgame. And so they go and do it. They are passionate about this topic. They go and do it. But again, there’s that asymmetry of passion.

Eric (22:37.549)
Mm-hmm.

Renee DiResta (22:51.261)
I can’t go edit it, I’m conflicted, right? So the best I can do is write the long form about it, hoping that somebody understands and try to shave off 500 words today to try to make it more human readable.

Eric (23:04.217)
What you’ve just outlined and the reason you wrote a 2000 word story about something that was decided a year ago, a year and a half ago. Oh, okay. Okay. But Amy Coney Barrett’s rule is sort of the, what do they call it? Answer engine optimization. Basically, what writing

Renee DiResta (23:13.343)
Well, no, the consent decrees were two weeks ago. The consent decrees were at the start of this month. Yeah.

Her decision was a year ago, yeah.

Renee DiResta (23:31.381)
Yeah.

Eric (23:32.723)
So that the artificial intelligence, which is becoming a source of wisdom and history for so many people, it’s really interesting. How else is AI changing the landscape? I know from media that the way we used to try to prioritize and drive people to our newsrooms was search engine optimization.

Renee DiResta (23:56.649)
Yeah. Yep.

Eric (23:58.893)
That’s now obsolete and it is now trying to anticipate what large language models will decide is true.

Renee DiResta (24:07.369)
Yeah, I mean, it’s a very interesting question. I don’t think anybody really has a lock on this yet. My job has always been looking, well, so one of the things we did at Stanford Internet Observatory, had this, we had what we called buckets to understand and to kind of segment out adversarial abuse. So information integrity was one where we looked at state actor or propaganda campaigns, spam and scams, trust and safety, mostly related to children.

Eric (24:09.346)
You

Renee DiResta (24:33.685)
And then emerging technology was bucket four, and that one kind of intersected with the other three, which was how does emerging technology change trust and safety? Well, it turns out, as we saw with Grok, extraordinary things happen in the realm of child safety once you have AI that can generate images. have propaganda is transformed when you have the ability to generate cheap text to make personas, right? I’ve been a user of large language models since about

Eric (24:38.669)
Mm-hmm.

Renee DiResta (25:00.981)
2020 was when I think we got access to GPT-3. We had research access. Before it was released to the public, certain research teams were given access to kind of play with it. And we actually wrote some papers with OpenAI, again, before it was even released, talking about how was this going to influence operations? How would this change capacity for manipulation? And it was a little bit, we were thinking more in terms of bots and

automation of manipulation. How would spammers and scammers, you know, I some of the early papers on like what came to be called AI slop, right? How would this transform the ecosystem? This for legitimate uses, I think it’s a hard competition because there is now this ability to generate content so cheaply that I was actually reading a thread by a content creator last night who was saying that there’s been such a massive hit

in traffic to blogs and websites, right? That that’s become, know, a recipe blogger was kind of weighing in saying like, you know, it’s increasingly hard to get people there. There are content creators saying, I’m competing with Slop for, you know, art, right? For art, even for video now, for music videos. So it’s really been, I think, transformative for so many industries. The question of how do LLMs interact with people, though? People do find them

persuasive, people do find them trustworthy. I was doing this investigation I was working on last night, was looking at a propaganda site, state-run propaganda site, and I was actually finding as I was kind of crawling Twitter, I was finding people asking, at Grok, is this true? In response to the tweets from the Twitter account associated with the propaganda site. And so it’s very interesting, I found maybe 25 or so at Grok, is this true?

queries in response to a state propaganda page. And Grok is in there, in their replies, trying to break down, well, this fact is true, but this characterization of it may not be. Which is, mean, to its credit, know, Grok is not actually all bad, right? It does actually get it right sometimes. The terminal Grok is actually quite good, I think. But so watching this dynamic take shape of the AI being the most immediate fact check response, and on X in particular.

Renee DiResta (27:23.891)
trustworthy response, more trustworthy than a human fact check, more immediate than a human fact check, you realize that making sure that these systems have access to your content is something that I think, for me anyway, has become a priority because I work in such an adversarial contested space. While on the flip side for creators, they don’t want it to, right? It’s the opposite problem. actually…

want to keep their stuff out, they don’t want it being returned. They want to maintain that human creativity component. So there’s very real tension there, I think, in how do you negotiate the future with a machine, so to speak.

Eric (28:05.026)
Wow, that grok, is it true, is going to be a cliche of our time.

Renee DiResta (28:09.211)
It is, I know it’s, used to kind of be more disparaging of it than I am now because I think it, one of the reasons why it happens is that people wanna know immediately, is this true? And that’s where the problem is actually, because in an immediate sense, the AI is not going to be as up to date as you want it to be. This is why Grok gets things wrong. It’s still also limited by immediacy, by recency. So if it’s pulling from tweets,

and there’s just been a terrible event, a tragedy of some sort, natural disaster, it’s not gonna know that that happened for a small period of time, or it’s gonna use what’s called retrieval augmented generation, where it’s gonna go out and search, and whatever it has available to it in search is what it’s going to tell you, and that might be wrong, particularly if it’s searching X, right? And that’s where you see these mistakes that it can make, but on the flip side, for longer term, longer horizon things that are at least semi-decided, or there’s a big corpus of information,

it can do that synthesis. that’s where you see the, one of the frustrations I have with Grok actually is that it will treat sites like, you know, Infowars and some of these others. And you see this reflected in Grokopedia as just another website, just another normal thing, or, you know, maybe you want to see this other perspective, which is Infowars, as opposed to actually taking a little bit more of what we might call a…

how a fact checker would think about it, recognizing that some sources actually just aren’t as good as others. And that’s not a partisan statement necessarily. There’s garbage across the spectrum. But that is something that you have to reckon with when the machines are fact checking for you.

Eric (29:49.305)
Do you think Grok in particular, Grok specifically is going to go the way of Twitter or X under the rule of Elon Musk? He obviously has a very different opinion from you about our world.

Renee DiResta (30:01.373)
He does, absolutely. Right, and this is really the interesting piece, I think, is that I would hope that people would also see Grok as illustrative of how vulnerable machines are to what are called system prompts that kind of govern how they engage or to their training data, which again, when Grokopedia is pulling from Infowars, you’re gonna get a very different article about vaccines or whatever than Wikipedia is going to give you.

Renee DiResta (30:33.376)
So that phenomenon of like, look at the huge divergence here in what Grok is telling you about, and one rather notorious example was like, you know, white farmers in South Africa, right? This became a kind of a pet issue of Elan’s for a while and it still is, I imagine. But Grok all of a sudden would respond, I don’t know if you remember this with, it started adding on like,

consider the Boer basically stuff where it was like, I don’t know if I pronounced that word correctly, the telling people in response to all sorts of queries for a couple of days there, at least a day, it was just saying like, are you aware of what’s happening in South Africa? So you could ask it anything, anything is the earth flat? Well, no, it’s not. But also have you heard about what’s happening in South Africa? And it just went crazy with this appending of this completely unrelated topic in that was Elon’s pet topic.

to this issue where it really started just constantly pushing this point of view. And my hope is that when people see things like that, they realize that the machine is not some oracle, right? That it is very much a function of its programming. There is no neutral. So trying to make sure that, you know, I write about these things when they happen. I have a substack where I kind of chronicle just so that people understand to the best that I can do what is actually happening there, why that system prompt going haywire should actually

give you that sense of healthy skepticism as you realize that the machine’s answers are very inflected by what it is incentivized to or given instructions to promote in those moments. And that’s where I think this question, as we move into the next election, as everybody is on chat bots is they’re increasingly shown to be persuasive, right? People, again,

find them trustworthy, convincing, and have long discourses with them. And so research, know, David Rand’s team has done research showing that this actually does have the power to move people away from conspiracy theories, toward or away from political positions. That’s very powerful. So you’re going to start to see people really interrogating that ahead of the elections as politicians try to figure out how to leverage this power now. And it’s going to be interesting to see how that is handled and treated by the different

Eric (32:24.472)
you.

Renee DiResta (32:48.923)
the different models.

Eric (32:50.158)
Gosh, I would hope so. David Rand earlier on in reality was talking about that issue. He’s done it all right recently.

Renee DiResta (33:02.183)
Yes, sorry, there’s a fire truck that went by.

Eric (33:05.047)
OK.

Renee DiResta (33:10.043)
About David Rand, just tell the David Rand said. Okay.

Eric (33:12.994)
Yeah, David Randall was talking about how AI can be persuasive or at factual.

Renee DiResta (33:19.167)
So David Rand has done a lot of work showing how chat bots can be persuasive, both in the realm of moving people away from conspiracy theories, but also moving people towards or away from political positions, candidates. And as people see the power of that persuasion, obviously, these papers get coverage, political campaigns pay attention, and then they try to figure out how can I put this to use for myself. Advocacy organizations also are starting to ask this question. So that is going to shape, I think, what our

2026 election is going to look like, you who is using chatbots effectively.

Eric (33:52.687)
I kind of get a chill up my spine thinking about a war of large language models and the bespoke realities that you write about in your book, just moving from social media platforms to large language models. But put that aside, I often worry a lot about my old profession of professional journalism, which has one of the casualties in the

decline of trust in all institutions. Some of it I would say we can, we should blame ourselves for. Some of it was the result of deliberate campaigns to discredit the work that we do. How would you recommend a regular person think about when to trust a news organization realizing that they’re made up of humans, humans are all.

prone to confirmation bias and other kinds of brain shortcuts. When to trust him, when to be skeptical.

Renee DiResta (34:56.949)
You know, there’s this funny quote, I think it’s Michael Crichton talking about Gell-Mann amnesia, this phenomenon where Murray Gell-Mann, I believe he was a physicist maybe, his name’s on the Santa Fe Institute, chaos theory folk. He, I forget his exact discipline, but he tells a story where he says, you know, I read an article in the paper about my discipline and I realize how full of mistakes or holes or errors it is, and then I turn the page and I see an article about

Eric (35:11.096)
Mm-hmm.

Renee DiResta (35:24.917)
know, world politics and I’m absolutely convinced, right? And I know I have this feeling, you know, when I’m going through reading coverage of the stuff that I work on and I think like, you know, that sentence is wrong, this is overstated, that’s not true, you know, the editorial board got this completely backwards, you know, so. So, and you know, and as somebody who writes myself and I write commentary, I’m not a journalist, but.

as somebody who’s written commentary for a long time, there’s two things I think about. One, does the publication fact check? And if I write for the Atlantic or the New York Times or something like that, it is gonna go through a rigorous round of fact checking. It is not getting out the door until two people have interrogated every single thing I’ve written. And even if it’s not linked in the article, they have gone and looked at it and asked me for the link in the Google Doc to verify that the claim is true.

So having been on the sort of writing end of that process, I do think about is this a publication that issues corrections? That’s the other thing. I just had a wild incident with the free press. I put this up on my sub stack where this guy wrote an article accusing me of getting his book review censored because I requested a correction and the publication took the book review down. I did not request that it come down, but that was what they chose to do. I had nothing to do with it.

And yet they gave him 2000 words to write about how I’d gotten it censored. And I sent them emails. I sent them the entire email exchange, the editorial team, saying, here’s my email. Here’s the email of the reviewer telling me he pulled the review. Can you please stick an editor’s note on this story? This is ridiculous. You pushed this out to a million people as your top story of the day. It’s wrong. It was absolutely baseless. You need to do something about this. And their response to me was, we’ll give you a letter to the editor to respond.

And then they edited my letter to the editor. Like redlining entire paragraphs that made them look bad or that stated facts. And I was like, this is surreal. Like I was accused of censorship and then the editorial board is like, no, no, we’re not gonna allow you to tell the truth here in our pages. And so these are the sorts of moments where I was like, you know, I actually realized that like, this is an outlet where there’s no easy.

Eric (37:31.842)
Ha

Renee DiResta (37:43.605)
place to request a correction. You actually don’t see correction request forms anywhere. There’s not a ledger at the start of the day where they’re like, these are the things that we got wrong, which you do see in the New York Times, you do see in the Washington Post, you do see in BBC, all these major outlets do that. So I ask, do they issue corrections? It’s a huge thing for me because it indicates to me that when they get it wrong, they’re willing to say they got it wrong.

And that’s very important. That’s basic ethics, but that also tells you that you should consider what’s in the pages more trustworthy, because when new information comes out, they do go and reckon with it. So that’s how I think about it. I follow a set of outlets across the political spectrum. I have a subscription to the Journal, the Times, half a dozen Substacks. Again, just try, and I follow a lot of people on Substack for free that are…

Eric (38:23.894)
huh.

Renee DiResta (38:35.263)
content creators, writers, again, across the political spectrum, just so that I can feel like I have a handle on what are different perspectives on this issue, how are people writing about it or talking about it. I think that’s really, it’s frustrating sometimes when you definitely get articles where you’re like, okay, this is ridiculous, how could this person think this thing? But it at least shows you where, you know, where kind of different perceptions are on a topic. And that’s the other thing that I think is important.

So one, does it issue corrections and have a fact checking process? And then two, and that’s really legacy media or mainstream media. And then two is how do you diversify the ecosystem of things that you consume? And that’s where I found Substack and others to be useful.

Eric (39:18.958)
In lateral reading, diversifying the part of the ecosystem that you are engaged in, how do you distinguish between, say, influencers who don’t have the institutional…

Renee DiResta (39:31.623)
Mm-hmm.

Eric (39:35.584)
apparatus around them to do things like fact checking, or for that matter, in most cases, to do any kind of basic reporting. So it’s basically analysis of news reported elsewhere. What’s your judgment about how seriously to take people?

Renee DiResta (39:52.479)
So I think of the influencer ecosystem as commentary. It’s built on top of, most of the time, the journalistic ecosystem. It is rare that you have an influencer who is actually out there doing independent reporting. Now, sometimes you do. There are reporters who have left some publications. think the Intercept comes to mind. There’s a few guys who have left the Intercept and started Substax. And they actually maintain a network of sources. And they actually continue to break news on Substax.

Again, the inner stuff has its angles, there is still, again, roughly speaking, they will show you the documents. My source gave me this. Do you agree with their characterization? Maybe not, maybe. But they’re at least showing their work. This is the document, this is who I talk to, here’s what I found. And so there are some who are still doing reporting, but most of the time it’s commentary. And so I think about it more as persuasion. This is persuasive content.

And there’s a kind of a talk that I’ve been giving because as I talked about in the beginning of our chat, during the vaccine conversation, institutional entities were not in the room. They weren’t in the conversation. They weren’t on social media. The CDC was not putting out good content. You really saw this during COVID. don’t know who and whatever your position on COVID vaccines are, I feel like universally there should be an acknowledgement that the communication was abysmal.

Right, was writing about that in the Atlantic at the time and I’m very pro-vaccine saying just like, man, they are really doing a terrible job with responding to where the public is on this stuff. They’re just not in the conversation. Frontline physicians were doing a great job, right? Again, people who were saying like, I am an emergency room doctor, let me tell you what is happening. And that’s a non-institutional communication. And you’re actually starting to see a lot of those people who started that during COVID continuing with it. So these influencers.

these content creators who have a background in medicine, who are epidemiologists, who are creating that content. And so there is a degree of reputability there because of their, you they’re speaking in their lane on the topic that they are experts in. And I think that they provide an accessibility and a speed, right? They’re able to kind of keep up with the speed and the discourse style of social media in a way that institutions don’t. until institutions are willing to experiment more and update

Renee DiResta (42:15.967)
how they engage, those folks actually I think are doing a hugely important service. I am generally speaking like quite pro influencer. I try to distinguish in the book this little taxonomy of there’s a difference between people who are just these rage bait, aggrievement, propiteers that are making all of their money by keeping their audiences upset and confused and enraged.

versus the influencer types who are out there just trying to do explainers and know folks like Dr. Mike for example who I think are doing great work.

Eric (42:48.492)
Yes, yes, I’m a big fan of Lena Wenn in the Washington Post. Yes, your lovely phrase in the book about that ecosystem is the perpetually aggrieved.

Renee DiResta (42:50.474)
cough

Renee DiResta (43:00.053)
Yes, I mean that’s how I think about them. just you know, I don’t know how people don’t get exhausted after a while you go to X or something I go to X mostly when I have some work I have to do and And I open the I open my tab and it’s just like rage bait after rage bait after rage bait So, you know somebody who puts a red siren on the most innocuous thing and is breaking, you know and it’s just trying to get you mad and you know, but they have thousands of engagements right and so You know it works. They make money

Eric (43:27.116)
Yes. don’t know if you saw the headline on FoxNews.com yesterday, Sunday as we record this, was about the deaths of 11 scientists over some indeterminate period. The thing that connected these 11 deaths was never explained. It was just the kind of implication that there was some conspiracy or this

Renee DiResta (43:42.906)
yeah.

Eric (43:56.942)
This needs investigation.

Renee DiResta (43:57.941)
The anti-vaxxers used to do that all the time. There was this person who I’m not even gonna give her credit because it’s so ridiculous, it would be, you’d have these like, these doctors would die of heart attacks and she would write, another one died. You you’re like, yes, of a heart attack, you know? And the implication was like somebody gave them the heart attack. Like, you know, the CIA swooped in and injected them or something. you know, it’s like people have seen too many movies,

But this was huge because they would, you know what it is? You can make collages of the faces of these people as collateral, right? And when you have collateral, you can kind of post that and it’s something for people to retweet. When you have the article headline, Suspicious Deaths, it gives something for people to share. And Fox is very willing to lean in to those instincts. And I think, again, this is where,

It’s very frustrating sometimes to be in the line of work I’m in because I remember Jim Jordan’s report was like, they disproportionately targeted conservatives. And I was like, no, we were doing a study of lies and rumors about the 2020 election. The proportionality wasn’t even, it wasn’t that we targeted anybody. We looked at what happened and then we wrote our statistics after the election was over. And lo and behold, most of the rumors about the election being stolen came from one side.

There’s no proportionality there. So you sound like you’re somehow like some crazy biased person when you say, look, part of this ecosystem just isn’t telling you the truth. And unfortunately, it is this one. And then they’re like, of course, the, know, the libtard professor would say that. And that’s what you deal with, you know.

Eric (45:42.691)
Yeah. Let’s talk about solutions. this is a big question. Bennett sort of circulates around section 230, which gives platforms immunity from liability for their content on their sites. What is the responsibility of social media platforms to provide reliable information? what can they do without challenging the First Amendment?

Renee DiResta (46:11.199)
So the First Amendment is not what applies, right? So first of all, I think it’s important to understand that legally speaking, this always gets you, again, in trouble with the same audiences that are like, you’re splitting hairs about the legality. But this is legally speaking, the platform has the First Amendment rights, not the user. The user is on a private platform.

Eric (46:19.79)
and

Eric (46:28.846)
Good point.

Renee DiResta (46:32.295)
the platform can kick you off for like misspelling the word cat if it wants to, right? It has carte blanche to do what it wants to do because you’re there under its terms of service, which are expansive. Now, theoretically, the terms of service are you follow these rules and then you are moderated in this way, you are treated in this way, and that’s the sort of implied contract. But in reality, they can change those rules however they want to. And you see this happen as they kind of pendulum swing back and forth at times to change their policies, not because any

First Amendment jurisprudence has changed, but just because they’ve decided that, the new administration is more inclined toward this, or the Senate will investigate me if I don’t do that. And so the First Amendment concern is the platforms. You are there under their sort of freedom of association rights that are allowing you to stay. So that I think is important to understand. And so they have a lot of leeway. Where Section 230 applies is this, it gives them the right but not the obligation to moderate. And it says that they can’t be held liable

for content posted by the users. Sorry, there’s a siren, one second.

toot toot toot.

Renee DiResta (47:46.261)
Okay, there it goes.

Renee DiResta (47:51.795)
Okay, so what that means is that they are called intermediaries, right? They are not liable for the speech of their users. But that is referring to a kind of a different issue altogether. That doesn’t determine their moderation rule. So I think these things get conflated a little bit of the time, particularly because, you you see the Senate saying, well, we want you to do more about certain types of manipulative content.

And they can go and do it, but they are not legally obligated to do it. And that’s where that discrepancy kind of comes into play. So a lot of people have very different views on how expansive the platform should be, but I think it’s also important to understand that people don’t enjoy unmoderated platforms. I do debates about content moderation, particularly with conservatives, all the time. And you ask the people in the room, how many of you regularly spend time on 4chan? And almost no hands go up, zero.

Eric (48:25.934)
Hmm.

Eric (48:49.102)
Hmm.

Renee DiResta (48:49.637)
Maybe one. And the reason is because people don’t actually enjoy being in unmoderated spaces. They are not looking for that type of dialogue and discourse. They are actually quite happy with an environment that at least is nominally doing something about brigading, bullying, harassment, doxxing, none of which is illegal under the First Amendment, right? In most cases. And so that’s where…

you begin, when you sort of point this out to people, they’re like, oh, well, yeah, do want that kind of moderation. Oh, I want that kind of moderation. I want that kind of moderation too. So the question is, how do you give people, how do you maximize freedom of expression while also at the same time minimizing the worst stuff? And that is where I think around the edge cases of, is this misinformation? How should we think about health content? How should we think about political content during an election? There’s a little bit more.

divergence of opinion around that.

Eric (49:44.143)
In the, there was recently a case in Los Angeles in which civil action found met a level for harming a plaintiff by the design of his platform. It was a consumer safety issue, not a censorship or content moderation issue. That’s, guess, one of the ways that the legal system can intervene. Are there other ways that?

Renee DiResta (49:45.237)
Cough

Renee DiResta (50:00.373)
Mm-hmm.

Renee DiResta (50:11.189)
So New Mexico, no sorry, I’m going to cut you off. Yeah, that’s a really interesting case. I’m kind of curious to see how it stands on appeal. New Mexico had a similar one, I think the week before or the week after. It was right around the same time. There’s a couple things that you see there. So first, it’s a recognition that, and this is something I’ve written about for a long time, the design of the platform and again, what it chooses to recommend.

That is a, it’s a curatorial decision in terms of, you know, it is the editorial content of the platform. is making, that is a speech decision when you think about what to recommend. But the design of things like infinite scroll, right? Is that an addictive product design feature? Are there certain things specific to, one of the reasons I mentioned recommendation engine is in the New Mexico case, they set up kind of a sting operation. They called it operation metafile.

where they made teenage girl accounts and then they found Metta recommending those accounts to older men. And then in response to that dynamic starting to take shape where these older men begin messaging these sort of fake teenage girls, Metta encouraged the teenage girls to monetize their accounts. It wasn’t doing anything to minimize the harassment or keep these messages out of their inbox or not recommend teenage girls to men. Instead, it kind of went the other direction.

Eric (51:10.894)
Mm.

Renee DiResta (51:34.289)
began to exacerbate. So that’s where you start to see when I ask these questions about what is the ethical way to design a recommendation system, most people I think would find that objectionable. I certainly do, right? That should just be a no. So this is where you get at these questions of at the same time on the flip side, what is the law? Like where does the law come into play so that while we see that as a bad thing,

Eric (51:45.762)
Yeah.

Renee DiResta (52:00.029)
you don’t want to create an environment where it turns into you shouldn’t recommend this topical content. So these questions around what is behavioral and product driven versus what is topical and content driven, that’s the line. And these two cases found that there were product design flaws. So it addressed what is called conduct, not content.

It addressed the conduct of the platform in its design. And that’s where both of those verdicts went against Meta. But again, this will be appealed. And so we will see what happens next. It’s a relatively new theory. But again, I have no problem in general with looking at certain feature design in this product safety way. think you just want to recognize that.

you want to make sure that it is not abused by people who are trying to use it as a backdoor to get at the content piece.

Eric (52:55.534)
Yes. You personally have been sued and subpoenaed and doxed, harassed, publicly accused of being in league with the government and censorship. That’s the report reviews, yes. Thank God. any of that changed what you’re willing to say publicly or how you go up?

Renee DiResta (53:08.425)
censoring book reviews.

Renee DiResta (53:12.902)
My powers are endless, you know?

Renee DiResta (53:23.541)
think honestly it’s probably made me more willing to talk because I find it so absurd. I mean, I find at this point, I find it so irritating. I think, look, a lot of people, I’ve been getting yelled at on the internet for a little over 10 years now across the political spectrum, right? And so, you know, I at this point see it as,

This is what happens when you choose to be in the arena on hot button issues. I’m, again, for whatever reason over the years, people do listen or pay attention to what I have to say. And so some criticism is part of that. And I actually don’t mind that. Again, I really feel like debates and response, and I interact a lot online. I’m extremely online person. it’s where, for me, the realization that

the government can just harass you. I was like, oh man, we’re really, I started reading a lot, I’ve got the books on the shelf behind me like that one, 30 Years of Treason is a transcript archive of the House and American Activities Committee, just like selected transcripts. And the one next to it is Walter Goodman’s The Committee, A History of the Committee. I spent a lot of time reading what happened from 1938 to 1968 or so with the House and American Activities Committee when I got my subpoenas.

Eric (54:33.774)
Renee DiResta (54:49.013)
Because it’s a collective action problem, right? And that’s where you realize that when these subpoenas start to come down and you’re asked to turn over all of your emails to Congress, they can just ask for that, by the way. There doesn’t have to be a justification. There was no indication that we had mass censored anything. It was just like Matt Tybee denounced us in a congressional hearing and then Jim Jordan gets to ask for my emails. And I was like, this is wild.

Eric (55:03.502)
Mm-hmm.

Renee DiResta (55:13.437)
And then material that was obtained under subpoena was turned over or found its way into the hands of America First Legal, which is suing us, which is suing me in Louisiana with the same judge as the Murthy judge, right? So lo and behold, you know, the machine operates as it does. And I was like, how is it that stuff that was turned over under subpoena or depositions that were given supposedly behind closed doors, testimonies to Congress, just find their way to plaintiff’s council? Like, what even is this?

Eric (55:25.678)
Yeah.

Renee DiResta (55:41.677)
And that’s when you realize that, you know, powerful people get to do things. Jim Jordan didn’t honor his own subpoena from the January 6th committee, I might add. But here we are. I have to. And so the question is, what is the what is the appropriate response? The answer that the House and American Activities Committee targets found is that the blacklists were all voluntarily self-imposed, right? At no point did Congress say to Hollywood or to academia or anyone else, you need to put these people on a blacklist.

Eric (55:48.376)
That’s right.

Renee DiResta (56:11.005)
They did that out of self-preservation. They didn’t see it as we should collectively fight back. They saw it as we should ostracize these people in hopes that it doesn’t come for us. And that is the wrong instinct to have, but that is unfortunately the instinct that people had when the subpoenas came for us. And now I think as the administration rolled, you have seen other people realize that this is, as they came for the law firms, as they came for the universities, that this is a power game.

right, it’s not even a game, this is power play, and that you have to have some sort of collective response. People have to realize that they are going to continue to apply pressure and you’re not going to exonerate yourself with your documents the same way those people couldn’t either. It was only when they started making the House Un-American Activities Committee hearings a spectacle that served them, where media coverage highlighted the injustice being done to the targets.

Eric (57:05.131)
and

Renee DiResta (57:09.555)
that you started to see the committee’s power wane. And that I think is the, for me as I’ve, you know, spent now three years with the government harassment too, realizing that we have to be thinking about this differently. And actually that speaking up I think is far, far, far better than staying silent. What does staying silent get me? It just lets other people set the agenda. Let’s the LLMs hear what they have to say and not me, right? So no, you have to fight.

Eric (57:37.794)
Renee, do you have hope? And just from what you said, collective action, we’ve seen it on the streets of Minneapolis. We’ve seen it in juries or in Virginia, rejecting some of the indictment or refusing to indict people. Does that kind of thing give you hope?

Renee DiResta (57:58.869)
It does, yeah, I think it’s where the people have a lot of power, right? And that’s where industries do, collective groups of people do. And I think that is the only thing that we have, right? The recognition that this doesn’t have to be forever. We can demand that other things happen, but as long as people cower or quietly turn over documents and shut up and say nothing,

you’re not going to win that way. So you’re not even going to help the public understand what’s happening that way. So it really is a function of people having strength. And I mean, there’s some real heroes that have come out of the on the street fights that we’ve seen in various states. So that is very inspiring.

Eric (58:29.098)
See

Eric (58:49.73)
Well, Renee, I consider you one of the heroes of this fight as well. And so thank you so much for what you’re doing. Thank you for joining us on In Reality Today.

Renee DiResta (58:57.673)
Thank you for having me.

Eric (59:01.774)
Um, that was awesome. Um, it’s probably really kind of cold.

Renee DiResta (59:04.533)
That was fun. Sorry, have a cough. like, I do. I’ve been like, it’s been man, like three and a half weeks at this point. I think it’s just the residuals, but trying not to cough. Sorry about that.

Eric (59:17.416)
or also it’s springtime too. There’s probably a lot.

Renee DiResta (59:20.711)
Yeah, the allergies, it’s actually, it’s very, you know, I’m in DC and there’s flowers and whatnot everywhere, so.

Eric (59:26.636)
Yeah, I hate that about you. Anyway, this was fabulous, exceeded my already high expectations. You’re amazing. And I’m not kidding, you’re one of the heroes of the fight. And we’ll get this up in a couple of weeks, I think. And then I’d love to see you. I mean, I’d love to. I will love to see you and look forward seeing you at the University of Chicago class.

Renee DiResta (59:37.001)
Thank you so much for having me.

Renee DiResta (59:44.895)
Okay.

Renee DiResta (59:56.509)
Likewise, likewise.

Eric (59:56.558)
That’d be great. All right. All the best to you, Renee. Bye bye.

Renee DiResta (01:00:01.159)
You too, take care, bye.

---

Created & produced by: Podcast Partners / Published: May 1 2026

---

Share this episode: