Lula's Rhetoric Is the Biggest Obstacle to Social Media Regulation, USP Professor Warns

By Tom Whitfield

Brazilian President Luiz Inácio Lula da Silva's own rhetoric now ranks among the largest obstacles to regulating social media, according to a University of São Paulo (USP) professor of public policy management, who argues that the president has placed the issue at the top of the political agenda in the worst possible way: as a response to what he considers digital platforms' interference in favor of the right.

The professor's column, published under the headline "Social media regulation must be non-partisan," takes direct aim at Lula's victory speech and a subsequent Sunday address in which the president said the platforms interfered in the electoral process and must therefore be regulated. "Besides affirming something about which he has no certainty, he is preparing a scenario similar to the one that led to the shelving of Bill 2.630," the professor writes, referring to the so-called Fake News Bill (PL das Redes Sociais), which stalled after the right suspected the legislation would be used against it. The country needs regulation, the column argues, but must build it in a balanced, non-partisan way.

What are the two myths muddying the debate?

The assumption that platforms favor the right rests on two myths the left has adopted, the professor argues: the belief that Cambridge Analytica manipulated the 2016 US elections, and the belief that platform algorithms systematically amplify hyperpartisan and disinformation-heavy posts.

Cambridge Analytica worked for Donald Trump's 2016 presidential campaign, selling the promise of messages tailored to each personality type. It sent pro-gun content in distinct forms to "neurotic" and "conscientious" voters versus "agreeable" ones. For the first group, guns were framed as an "insurance policy" against violence. For the second, a tradition passed down "from father to son." The firm claimed it could classify all American voters across five personality dimensions by matching personality traits with Facebook page-like patterns.

Much of the left attributed Trump's victory to this manipulation. No independent academic evaluation proved the approach worked. The strong academic consensus, the professor writes, is that it did not. Detailed analyses appear in David Sumpter's book Outnumbered (Bloomsbury) and in Network Propaganda (Oxford University Press) by Yochai Benkler and collaborators. In short: the model used to predict personality from liking patterns was highly imprecise and barely distinguishable from chance.

The identification problems were so severe that researchers believe the model was likely never used in the campaign. Even if it had been, its effect would have been negligible. Using the best-performing personality-targeted ad indices in the literature, Benkler and collaborators calculate the method could have yielded at most 3,000 additional votes in Pennsylvania — a state Trump won by 44,000. The margin dwarfs the supposed manipulation.

Do algorithms really favor hyperpartisan content?

The second distortion is the claim that platforms favor hyperpartisan and disinforming content, especially from the right. Here, too, the professor cites substantial contrary evidence.

Analyzing Meta data totaling billions of views, Andrew Guess and collaborators, in the study "Cracking open the news feed," showed that news from serious journalistic outlets is shared five times more than content from low-reliability sites — and viewed seven times more. Journalism reaches more people than its share of sharing would predict.

Fabio Giglietto and Giada Marino went further in "Beyond the Share Button": among links with the same number of shares, those from journalistic sources reach more people, while those shared by partisan audiences — left or right — reach fewer. Giglietto and Marino conclude the data suggest "a dual and deliberate strategy: suppress the reach of highly partisan content and, at the same time, amplify that of professional journalistic sources."

What would workable regulation look like?

None of this argues against regulating social media, the professor writes. Platforms play a central role in mediating public debate. Their content-distribution algorithms must be subject to audit and supervision, and their moderation practices must follow clear rules accepted by all. That interest is shared by society as a whole and by every political force.

The distrust runs in both directions. The left suspects the algorithms privilege the right. The right suspects moderation silences its voice. A regulatory model built on transparency, independent audits and rules accepted by all could reduce suspicions on both sides.

That requires treating regulation as a guarantee for everyone, not as a weapon against political adversaries. By presenting it as a response to platform interference favoring the right, Lula compromises the possibility of building the consensus needed to pass it, the professor concludes. Until the framing changes, Brazil's regulation agenda — and any marketer's planning around platform rules in one of the world's largest digital ad markets — stays stuck where Bill 2.630 left it.