New research on Facebook shows the algorithm isn't entirely to blame for political polarization

For each the blasted Facebook has received for fostering utmost governmental polarization connected its ubiquitous apps, caller probe suggests that the occupation whitethorn not strictly beryllium a relation of the algorithm.

In 4 studies published Thursday successful the world publications Science and Nature, researchers from respective institutions, including Princeton University, Dartmouth College and the University of Texas, collaborated with Meta to probe the impact of societal media connected ideology and the 2020 statesmanlike election.

The authors, who received nonstop entree to definite Facebook and Instagram information for their research, overgarment a representation of a immense societal web comprised of users who often question quality and accusation that conforms to their existing beliefs. Thus, radical who privation to unrecorded successful alleged echo chambers tin easy bash so, but that's arsenic overmuch astir the stories and posts they're trying to find arsenic it is the company's proposal algorithms.

In 1 of the studies successful Science, the researchers showed what happens erstwhile Facebook and Instagram users spot contented via a chronological provender alternatively than an algorithm-powered feed.

Doing truthful during the three-month period, "did not importantly change levels of contented polarization, affective polarization, governmental knowledge, oregon different cardinal attitudes," the authors wrote.

In the different Science article, researchers wrote that "Facebook, arsenic a societal and informational setting, is substantially segregated ideologically — acold much than erstwhile probe connected net quality depletion based connected browsing behaviour has found."

In each of the caller studies, the authors said that Meta was progressive with the probe but the institution didn't wage them for their enactment and they had state to people their findings without interference.

One survey published successful Nature analyzed the conception of echo chambers connected societal media, and was based connected a subset of implicit 20,000 big Facebook users successful the U.S. who opted into the probe implicit a three-month play starring up to and aft the 2020 statesmanlike election.

The authors learned that the mean Facebook idiosyncratic gets astir fractional of the contented they spot from people, pages oregon groups that stock their beliefs. When altering the benignant of contented these Facebook users were receiving to presumably marque it much diverse, they recovered that the alteration didn't modify users' views.

"These results are not accordant with the worst fears astir echo chambers," they wrote. "However, the information intelligibly bespeak that Facebook users are overmuch much apt to spot contented from like-minded sources than they are to spot contented from cross-cutting sources."

The polarization occupation exists connected Facebook, the researchers each agree, but the question is whether the algorithm is intensifying the matter.

One of the Science papers recovered that erstwhile it comes to news, "both algorithmic and societal amplification play a part" successful driving a wedge betwixt conservatives and liberals, starring to "increasing ideological segregation."

"Sources favored by blimpish audiences were much prevalent connected Facebook's quality ecosystem than those favored by liberals," the authors wrote, adding that "most sources of misinformation are favored by blimpish audiences."

Holden Thorp, Science's editor-in-chief, said successful an accompanying editorial that information from the studies amusement that "the quality fed to liberals by the engagement algorithms was precise antithetic from that fixed to conservatives, which was much politically homogeneous."

In turn, "Facebook whitethorn person already done specified an effectual occupation of getting users addicted to feeds that fulfill their desires that they are already segregated beyond alteration," Thorp added.

Meta tried to rotation the results favorably aft enduring years of attacks for actively spreading misinformation during past U.S. elections.

Nick Clegg, Meta's president of planetary affairs, said successful a blog post that the studies "shed caller airy connected the assertion that the mode contented is surfaced connected societal media — and by Meta's algorithms specifically — keeps radical divided."

"Although questions astir societal media's interaction connected cardinal governmental attitudes, beliefs, and behaviors are not afloat settled, the experimental findings adhd to a increasing assemblage of probe showing determination is small grounds that cardinal features of Meta's platforms unsocial origin harmful 'affective' polarization oregon person meaningful effects connected these outcomes," Clegg wrote.

Several authors progressive with the studies conceded successful their papers that further probe is indispensable to survey the proposal algorithms of Facebook and Instagram and their effects connected society. The studies were based connected information from a abbreviated clip play coinciding with the 2020 statesmanlike election, and further probe could unearth much details.

Stephan Lewandowsky, a University of Bristol psychologist, was not progressive with the studies but was shown the findings and fixed the accidental to respond to Science arsenic portion of the publication's package. He described the probe arsenic "huge experiments," that shows  "that you tin alteration people's accusation diet, but you're not going to instantly determination the needle connected these different things."

Still, the information that the Meta participated successful the probe could power however radical construe the findings, helium said.

"What they did with these papers is not implicit independence," Lewandowsky said. "I deliberation we tin each hold connected that."

WATCH: CNBC's afloat interrogation with Meta main fiscal serviceman Susan Li