Pew Research Context: Americans Can’t Even Agree on Facts Anymore

Survey data from polling organizations provides important context for understanding algorithmic polarization research. According to Pew Research, eight in ten American adults believe Republicans and Democrats cannot agree not only on policies and plans but on basic facts themselves. This epistemic crisis—the collapse of shared reality—creates conditions where algorithmic manipulation becomes particularly dangerous.
When citizens inhabit separate information universes, they become more vulnerable to polarizing algorithms. If people cannot even agree on what is factually true, then content that reinforces partisan interpretations of reality faces little resistance. Users already primed to distrust political opponents are more likely to accept and amplify divisive content that confirms their existing suspicions.
New research demonstrates that algorithms systematically amplify content presenting biased evaluations of politicized facts. When over 1,000 X users received feeds with slightly more such content during the 2024 presidential election, their political polarization increased dramatically. This suggests that algorithmic choices don’t just reflect existing epistemic divisions but actively deepen them.
The feedback loop is concerning. Platforms optimize for engagement, which rewards emotionally provocative content. Divisive political content generates strong reactions and high engagement. This incentivizes algorithms to amplify posts that present biased fact interpretation. Users exposed to such content become more polarized and less trusting of political opponents. This increased polarization makes them even more receptive to biased content, continuing the cycle.
Breaking this feedback loop will require interventions at multiple levels. Platforms must redesign algorithms to prioritize truth and shared understanding over pure engagement. Media literacy initiatives must help citizens develop stronger critical evaluation skills. And political leaders must work to rebuild basic epistemic common ground where productive disagreement becomes possible again.