ChatGPT and other AI tools make journalists look powerless
I’ve been training digital entrepreneurs for several years on how to combat disinformation and propaganda online. Bravely — or naively — I preach the value of spreading trustworthy information. Over the long haul, the truth will triumph, I say.
But I have to admit that my cheery, optimistic attitude has been shaken lately. Consider the following articles from my reading today (Feb. 23, 2023).
Chatbots trigger next misinformation nightmare, from Axios. “Misinformation can flow into AI models as well as from them. That means at least some generative AI will be subject to ‘injection attacks’, where malicious users teach lies to the programs, which then spread them.”
Or this one — “The Next Great Misinformation Superspreader: How ChatGPT Could Spread Toxic Misinformation At Unprecedented Scale, from NewsGuard. “We tempted the AI chatbot with 100 false narratives from our catalog of Misinformation Fingerprints™. 80% of the time, the AI chatbot delivered eloquent, false and misleading claims about significant topics in the news, including COVID-19, Ukraine and school shootings.”
Or this one: “Disinformation Researchers Raise Alarms About A.I. Chatbots,” from the New York Times. The article quotes Gordon Crovitz, one of the co-CEOs of NewsGuard: “This tool is going to be the most powerful tool for spreading misinformation that has ever been on the internet.”
I’ll stop there. You can find plenty more similarly scary stories every day. And, to be fair, the articles above described tactics and strategies that have been effective in countering misinformation. All of this made me go back to a blog entry I made in 2021, “Are we exaggerating the power of misinformation?”
That blog post cited data and research about how misinformation spreads and questioned how effective it is at changing people’s minds. I have edited it slightly to bring it up to date.
Sept. 23, 2021
The alarm bells started ringing for me when I saw the article in First Draft discrediting a Rolling Stone story that depicted anti-vaxers as pawns of right-wing conspiracy theorists. The anti-vaxers were taking a deworming drug for animals to protect themselves from covid-19, were suffering overdoses, and were overwhelming Oklahoma City hospitals.
First Draft is an organization dedicated to fighting disinformation. It is very much aligned with my own fears about the pernicious influence of right-wing extremists and conspiracy theorists. So I was surprised at how critical First Draft was of progressive media outlets (the BBC, The Guardian, and Newsweek) that uncritically republished summaries of Rolling Stone’s article when a basic fact check would have revealed the story was bogus, based on a single second-hand source.
Three days later, First Draft followed up by publishing a newsletter that questioned the assumption behind its own mission: “Is Misinformation the Big Problem We Think it is?”
The First Draft newsletter referred to an article in Harper’s Magazine: “Bad News: Selling the story of disinformation,” by Joseph Bernstein. He cites several studies by experts- here, here, here, and here -that all question the effectiveness of internet advertising or disinformation campaigns.
All these studies cast doubt on what Bernstein calls Big Disinfo, a group of institutions that aims to alarm us about fake news. He says “the institutions that publish most frequently and influentially about disinformation” are all trying to restore themselves as an “authoritative voice” in society. Those institutions include Harvard University, the New York Times, Stanford University, MIT, NBC, the Atlantic Council, the Council on Foreign Relations, and others.
Then he observes, “That the most prestigious liberal institutions of the pre-digital age are the most invested in fighting disinformation reveals a lot about what they stand to lose, or hope to regain.”
The article of faith of big Disinfo, namely that internet advertising and disinformation can sell us anything, “creates a world of persuasion that is legible and useful to capital-to advertisers, political consultants, media companies, and of course, to the tech platforms themselves. It is a model of cause and effect in which the information circulated by a few corporations has the total power to justify the beliefs and behaviors of the demos [the common people].”
See also this blog entry, Economists ask, Does advertising actually work?
One of the articles cited by Bernstein found that Facebook sharing of links from fake-news domains (defined as those publishing “knowingly false or misleading content”) was “quite rare” during the 2016 presidential election campaign in the U.S. Those most likely to do so were people age 65 and older from all political affiliations.
The authors, from Princeton and New York University, estimated that the average adult saw and remembered only 1.14 fake news articles from the campaign. (See the data in “Less than you think: Prevalence and predictors of fake news on Facebook,” by Andrew Guess, Jonathan Nagler, and Joshua Tucker, in ScienceAdvances, 2019.)
Of course we are bombarded by anti-vax and pro-vax campaigns, fueled and financed by the polarizing political narratives of our two-party system. But who is actually being persuaded to change their mind? Mostly these campaigns solidify the views of extremists on both ends of the political spectrum.
All of this is not to minimize the threat of documented efforts by Russia and China to flood our digital feeds with information aimed at undermining democracy and creating distrust of our institutions. But much of it is ignored, as we saw above.
The vast majority of Americans are moderates, despite what you see on cable TV or in social media, and they are looking for non-political sources they can trust. With lives at stake in the covid-19 crisis, news consumers around the world returned to trusted news brands for information, according to Reuters Institute’s 2021 Digital News Report.
Americans generally occupy the middle ground, despite the Trump Derangement Syndrome, in which liberals for five years were predicting the collapse of American Democracy, and despite the previous Obama Derangement Syndrome, in which conservatives for eight years predicted a communist takeover of health care and a social welfare state.
What we got instead from the American political system under Obama was modest change from mainstream policies. And under Trump, also modest change. American democracy survived both of those presidents, and Trump was not re-elected. By contrast, Trump-like presidents in Eastern Europe and Latin America have been rigging elections and dismantling the institutions of democracy-the legislature and courts-and establishing de facto dictatorships.
We need to take these threats seriously. But the American people are more moderate and have more common sense than we give them credit for.
( A study by More in Common, an international nonprofit whose mission is to “strengthen resilience against the forces of division” in democratic societies, concluded that about two-thirds of Americans belong to the “exhausted majority” whose voices “are rarely heard above the shouts of the partisan tribes.” )
Most Americans are tired of political correctness. They don’t trust the news media. And we journalists ought to take that into account. We have to more aware of our own biases. We have to be more skeptical about accepting stories that feed into our preferred narratives about politics. Otherwise, we are part of the problem.
Updated conclusion on Feb. 23, 2023: Upon re-reading all of the sources cited in 2021, I find myself somewhat less terrified. I say “somewhat” since ChatGPT and other AI tools are so new that we don’t have enough examples of how the conflicts will play out between the bad actors and the trustworthy ones.
It’s bit like trying to pick the winner of a football (soccer) league after only two weeks of play. It’s early days. Now is the time for academic and journalistic researchers to collaborate as never before and find new solutions to these new forms of disinformation. I have confidence in our side. If I were a betting man, I would put money on it.
Research referred to above:
“ Social Media and Fake News in the 2016 Election,” by Hunt Allcott and Matthew Gentzkow, Journal of Economic Perspectives, 2017.
Social Media, Political Polarization,and Political Disinformation: A Review of the Scientific Literature, the William and Flora Hewlett Foundation, 2018.
Subprime Attention Crisis.Advertising and the Time Bomb at the Heart of the Internet. by Tim Hwang. Farrar, Straus and Giroux, 2020.
My contribution to Trump Derangement Syndrome: Is Facebook really helping spread covid-19?