Discover more from Infonomena by Bret Swanson
Censors Target Internet Talkers With A.I. Truth Scores
Casting pods into darkness
Elon Musk’s purchase of Twitter may have capped the opening chapter in the Information Wars, where free speech won a small but crucial battle. Full spectrum combat across the digital landscape, however, will only intensify, as a new report from the Brookings Institution, a key player in the censorship industrial complex, demonstrates.
First, a review.
Reams of internal documents, known as the Twitter Files, show that social media censorship in recent years was far broader and more systematic than even we critics suspected. Worse, the files exposed deep cooperation – even operational integration – among Twitter and dozens of government agencies, including the FBI, Department of Homeland Security, DOD, CIA, Cybersecurity Infrastructure Security Agency (CISA), Department of Health and Human Services, CDC, and, of course, the White House.
Government agencies also enlisted a host of academic and non-profit organizations to do their dirty work. The Global Engagement Center, housed in the State Department, for example, was originally launched to combat international terrorism but has now been repurposed to target Americans. The U.S. State Department also funded a UK outfit called the Global Disinformation Index, which blacklists American individuals and groups and convinces advertisers and potential vendors to avoid them. Homeland Security created the Election Integrity Partnership (EIP) – including the Stanford Internet Observatory, the University of Washington’s Center for an Informed Public, and the Atlantic Council’s DFRLab – which flagged for social suppression tens of millions of messages posted by American citizens.
Even former high government U.S. officials got in on the act – appealing directly (and successfully) to Twitter to ban mischief-making truth-tellers.
With the total credibility collapse of legacy media over the last 15 years, people around the world turned to social media for news and discussion. When social media then began censoring the most pressing topics, such as Covid-19, people increasingly turned to podcasts. Physicians and analysts who’d been suppressed on Twitter, Facebook, and YouTube, and who were of course nowhere to be found in legacy media, delivered via podcasts much of the very best analysis on the broad array of pandemic science and policy.
Which brings us to the new report from Brookings, which concludes that one of the most prolific sources of ‘misinformation’ is now – you guessed it – podcasts. And further, that the under-regulation of podcasts is a grave danger.
In “Audible reckoning: How top political podcasters spread unsubstantiated and false claims,” Valerie Wirtschafter writes:
Due in large part to the say-whatever-you-want perceptions of the medium, podcasting offers a critical avenue through which unsubstantiated and false claims proliferate. As the terms are used in this report, the terms “false claims,” “misleading claims,” “unsubstantiated claims” or any combination thereof are evaluations by the research team of the underlying statements and assertions grounded in the methodology laid out below in the research design section and appendices. Such claims, evidence suggests, have played a vital role in shaping public opinion and political behavior. Despite these risks, the podcasting ecosystem and its role in political debates have received little attention for a variety of reasons, including the technical difficulties in analyzing multi-hour, audio-based content and misconceptions about the medium.
To analyze the millions of hours of audio content, Brookings used natural language processing to search for key words and phrases. It then relied on self-styled fact-checking sites Politifact and Snopes – pause for uproarious laughter…exhale – to determine the truth or falsity of these statements. Next, it deployed a ‘cosine similarity’ function to detect similar false statements in other podcasts.
The result: “conservative podcasters were 11 times more likely than liberal podcasters to share claims fact-checked as false or unsubstantiated.”
One show Brookings misclassified as “conservative” is the Dark Horse science podcast hosted by Bret Weinstein and Heather Heying. Over the past three years, they meticulously explored the complex world of Covid, delivering scintillating insights and humbly correcting their infrequent missteps. Brookings, however, determined 13.8 percent of their shows contained false information.
What would the Brookings methodology, using a different set of fact checkers, spit out if applied to CNN, the Washington Post, the FDA, CDC, or hundreds of blogs, podcasts, TV doctors, and “science communicators,” who got nearly everything wrong?
Speaking on journalist Matt Taibbi’s podcast, novelist Walter Kirn skewered the new A.I. fact-checking scheme. It pretends to turn censorship into a “mathematical, not Constitutional, concern” – or, as he calls it, “sciency, sciency, sciency bullshit.”
The daisy chain of presumptuous omniscience, selection bias, and false precision employed to arrive at these supposedly quantitative conclusions about the vast, diverse, sometimes raucous, and often enlightening world of online audio is preposterous.
And yet it is deadly serious.
The collapse of support for free speech among Western pseudo-elites is the foundation of so many other problems, from medicine to war. Misinformation is the natural state of the world. Open science and vigorous debate are the tools we deploy to become less wrong over time. Individual and collective decision-making depend on them.