The AI divides framework is a description, originated by Luiza Jarovsky, of three societal divides she argues generative AI is already producing: a literacy divide (access versus no access to advanced AI tools and training), an occupational divide (white-collar reorganization versus blue-collar insulation), and an ethico-philosophical divide (AI as an end in itself versus AI as a tool for human flourishing). Jarovsky frames these as durable divides that will reshape the next several decades.
Origin
The framework appears in The AI Divides Are Here (Jarovsky, April 2026) (April 2026, edition #284). It frames generative AI as already producing three durable societal divides.
The three divides
1. Literacy divide
The first divide separates people who have access to and resources for experimenting, exploring, learning, and applying the latest AI developments in their professional practice from those who do not. Jarovsky describes this divide as partially economic — a computer, fast internet, and paid AI subscriptions are prerequisites — but not only economic. It also requires time and availability for upskilling, access to relevant courses, training, and upskilling programs, a supporting professional environment in which to practice and apply skills, and informative resources and a supporting network.
In Jarovsky's account, basic literacy is not enough. AI literacy that serves as a differentiator must be applicable in a person's specific professional environment. Her analogy is that knowing how to navigate the internet in 2020 was not enough to get a competitive job, and that knowing the basic ChatGPT interface in 2026 will not be enough. She argues that most professionals today will not have access to the resources and supporting environment needed for adequate theoretical-and-practical AI literacy.
2. Occupational divide
The second divide concerns the differing pressures on white-collar and blue-collar work. White-collar workers are being told that if they do not become "AI-fluent" and obsessive AI users soon enough, they will be replaced by someone who is. Blue-collar workers (manual labor, skilled trades) are told their jobs are safe, for now.
Jarovsky cites a documented inversion as evidence the divide has already arrived. Per Washington Post January 2026 reporting on BLS data, the unemployment gap between bachelor's-degree workers and occupational-associate-degree workers (plumbers, electricians, pipe fitters) flipped in 2025 — the first such inversion since BLS began tracking the data in the 1990s, with trade workers holding a slight edge for six months out of the past year.
In her account the new techno-social reality places the two groups in opposite corners. White-collar workers depend on intensive, continuous AI automation and "AI-first transformation"; blue-collar workers depend on the absence of automation and the continued need for human labor. Jarovsky argues their incentives are diametrically opposed, with the existence of one threatening the survival of the other, and predicts "unimaginable levels of social strife" if this trajectory continues.
3. Ethico-philosophical divide
The third divide separates two views of AI's purpose. On one side are people who see AI as a goal, an end in itself, with maximum automation, acceleration, "efficiency," and "abundance" as the desired future — the more, the better. Jarovsky describes this as the mentality promoted by the majority of the AI industry, and as what many see as the only mindset for survival in the AI age, per the AI race and national security argument.
On the other side are people who see AI as a means, a tool that might be helpful sometimes but that must be constantly regulated, contained, and overseen, on the view that lack of control could lead to systemic individual and societal harm. For this camp the priority is preserving what is inherently human — biological boundaries, the need for connection, belonging, and meaning — and using AI to serve humans and to foster human rules, policies, and rights. Jarovsky places herself in this group.
Jarovsky argues that this ethico-philosophical divide is slowly taking center stage in global politics and that it has nothing to do with the old "left" and "right" polarization, with both camps containing people who would identify across the conventional spectrum.
Smartphone and social media precedent
Jarovsky's broader claim is that AI's divides should be expected to entrench similarly to those of smartphones and social media. She argues the negative impacts of those technologies — on focus, cognitive development, reading scores, mental health, well-being, and political polarization — were not visible in their early years, and that twenty years later "we might actually never overcome" them. In her account the disruption-and-divide cycle occurs in every mass-scale technological wave, and the open question is which side of each divide a person ends up on.
A candidate fourth divide: machine fluency
Imas, Lee, and Misra (December 2025) add a candidate fourth divide to Jarovsky's three: machine fluency, the ability to instruct AI agents to align effectively with one's objective function through prompt design. Their experimental marketplace (N=299, induced-values bargaining) finds that machine fluency varies systematically with cognitive sophistication, education, gender, and personality, and predicts large differences in negotiated economic surplus, despite agents not seeing principal demographics.
In their account machine fluency is distinct from Jarovsky's literacy divide because it operates after a person has access to AI tools — the gap is in how well one's prompts elicit aligned, high-performing agent behavior. They argue that as more economic decisions are delegated to agents, this gap is likely to compound the literacy and occupational divides.
The UN's Independent International Scientific Panel reframes the divide in its July 2026 report: it "is not just about access, but about capacity to influence artificial intelligence development" — a participation question rather than a distribution one. The report also records that "over a billion people now use conversational AI weekly" while "AI access and usage vary widely globally, with adoption across the global South lagging far behind," and notes that inputs and outcomes are "geographically and linguistically uneven."
Relationships
- introduced-by: The AI Divides Are Here (Jarovsky, April 2026)
- related: AI Literacy, Labor Disruption Timelines: Who Predicts What and Why, America's AI Action Plan (oppositional)
- related: Machine Fluency — candidate fourth divide
- related: Agentic Interactions — Imas, Lee, Misra (2025) — empirical evidence for the machine-fluency divide
- depends-on: Digital divide (general concept)
- instance-of: AI Policy (umbrella)