Are Algorithms Really Driving Us Apart?
New research is complicating the popular narrative that social media algorithms are the primary engine of political polarization. A widely discussed Science study found that even simple platform mechanics — posting and following, with no sophisticated recommendation engine at all — can generate polarization in simulations, suggesting the dynamic may be baked into social media's basic structure rather than solely a product of engagement-optimizing code [1].
That said, the same body of research shows algorithms aren't innocent bystanders. A companion experiment found that reranking feeds to reduce partisan animosity measurably lowered affective polarization — more than two points on standard "feeling thermometer" scales — indicating that feed design still matters a great deal, even if it isn't the sole culprit [2]. Meanwhile, Pew survey data shows that publics increasingly perceive the internet and social media as divisive forces, with 82% of Germans, 79% of Britons, and 64% of Japanese respondents agreeing the technology has made people more divided [3].
Defenders of current platform design argue algorithms largely reflect what users already want to see and click, putting responsibility back on human psychology rather than code. Critics respond that this sidesteps the platforms' power to shape what gets amplified in the first place, and are pushing for solutions like greater transparency, user-controlled feed settings, or open protocols — such as AT Protocol — that let people or their own AI agents curate discovery rather than leaving it entirely to opaque corporate algorithms.
The Case Against Binary Thinking on AI
As AI reshapes industries, a growing chorus of commentators is pushing back against the tendency to frame every AI debate as a two-sided fight. Commentary in Forbes and elsewhere argues that framing AI purely as "job destroyer vs. job enhancer" or "existential risk vs. overhyped distraction" flattens a far messier reality, one where AI reconfigures work rather than simply eliminating or multiplying it [1].
Academic voices make a similar case: a widely circulated paper warns against "slipping into binary thinking" about AI, noting that most informed observers actually occupy a spectrum of views on risk, ethics, and capability — yet public discourse rewards the loudest extremes [2]. Analysts describe this as a trap that forces people into camps—existential-risk doomers versus capabilities accelerationists, ethics-first versus deployment-first—when a "silent majority" may hold more hybrid, context-dependent positions [3].
Those less persuaded by calls for nuance argue that some questions genuinely do have clearer lines — safety thresholds for autonomous systems, or measurable economic displacement in specific sectors — and that false even-handedness can blur legitimate urgency. The strongest version of the nuance argument isn't that all positions are equally valid, but that tribal hostility between camps actively hinders the collaborative governance work AI's complexity demands.
The Bigger Picture
Today's stories share a common thread: the gap between how issues are debated and how they actually work. The Anthropic-Nippan dispute isn't simply "AI companies vs. publishers" — it's a genuine legal ambiguity about what first-sale doctrine means in an age of machine learning, where both sides have coherent, defensible positions rooted in different values: creative ownership versus established property rights.
The polarization research cuts even deeper, suggesting that some of our instinct to blame algorithms for division may itself be a symptom of binary thinking — "tech bad" versus "tech neutral" — when the evidence points to a more tangled truth involving both platform design and basic human social dynamics. That same instinct to seek simple villains and heroes is precisely what the third story's AI-ethics commentators are warning against on a larger scale, across the entire conversation about artificial intelligence's role in society.
What connects all three is a case for epistemic humility: holding two true things at once, resisting the pull toward tribal certainty, and recognizing that the most productive disagreements start by taking the strongest version of the other side's argument seriously.
Key takeaway: The hardest part of today's AI and tech debates isn't finding the right answer — it's resisting the urge to pick a side before understanding the full shape of the problem.
Sources
- https://www.nippon.com/en/news/yjj2026100700782/
- https://english.kyodonews.net/articles/-/87406
- https://www.snopes.com/fact-check/ai-companies-destroying-rare-books/
- https://www.science.org/content/article/don-t-blame-algorithm-polarization-may-be-inherent-social-media
- https://www.science.org/doi/10.1126/science.adu5584
- https://www.pewresearch.org/chart/growing-shares-across-countries-say-the-internet-and-social-media-have-made-people-more-divided/
- https://www.forbes.com/sites/michaeledmondson/2026/04/28/why-binary-thinking-falls-short-in-the-age-of-artificial-intelligence/
- https://arxiv.org/html/2312.14230v1
- https://www.scien.cx/2024/07/16/the-trap-of-ai-binary-thinking/