Which societies will outsource judgment to AI first?

The question existing frameworks cannot answer

The paper opens with the "Claude Boys," a term from online discourse that the Cosmos Institute analyzed in 2025: constant querying of AI systems, identity built around tool fluency, and the gradual outsourcing of judgment. The distinctive feature is who it affects. The pattern shows up among capable, cognitively sophisticated users who are overwhelmed by epistemic demands and blocked by institutions that do not resolve anything.

Loneliness misses them, since many are socially embedded. Power distance misses them, since they are usually skeptical of authority. Media literacy misses them, since they are sophisticated consumers of information. The paper's question is therefore narrow: what predicts enthusiastic AI deference by high-agency actors?

The index: eight dimensions

Each dimension is weighted by its predictive importance as judged from cross-cultural research and observation of AI adoption. The weights are the paper's third version and are a claim to be tested.

The weighting encodes one theoretical commitment: relational density matters more than individual psychology. A person with weak critical thinking but strong community ties has someone to check their reasoning. A person with excellent media literacy and no one to talk to has only themselves as a corrective, and an AI that never sleeps.

The added factor: Epistemic Overload with Procedural Vacuum

The eight dimensions explain why people lose confidence in their own judgment. EOPV explains why that confidence migrates to AI specifically, when it could as easily have collapsed into nihilism, paralysis, or deference to other humans. The profile has five parts:

• cognitively overloaded (too many decisions, too little time);
• procedurally blocked (the institutions that should adjudicate do not function);
• status-competing (professional identity tied to output and expertise signaling);
• tool-fluent (comfortable with AI systems);
• intolerant of delay (cannot wait for human consensus or institutional resolution).

The paper calls the resulting need "system hunger." Its counterfactual test: high-corruption societies without tech fluency do not show the pattern; high-fluency societies with working adjudication mostly do not; high-loneliness societies without competitive pressure show emotional reliance, and their deference stays personal instead of governmental. Only when epistemic load outpaces adjudication capacity does eager, status-seeking deference appear, which means EOPV predicts timing as well as susceptibility.

Two pathways and a convergence zone

Country rankings

Applying the framework to available cross-cultural data gives the ranking below, where 0 is maximum resilience and 100 maximum vulnerability. The inputs are cross-sectional and validation against observed deference is still to come, so read it as a set of predictions.

Mediterranean resilience. Italy, Spain, and Portugal score as most resilient despite lower rankings on conventional development metrics. The protection comes from relational density: multigenerational households, strong kin networks, frequent informal contact, and norms around shared meals and communal decision-making.

Nordic nuance. The Nordic countries score better than the Anglosphere despite high rates of solo living. Living alone in Denmark is embedded in associations, clubs, and welfare structures; living alone in the United States more often means isolation.

Anglosphere vulnerability. Australia, Canada, the UK, and above all the US come out high-risk despite high educational attainment and technological sophistication. The common factor is attachment collapse: suburban atomization, weakened extended families, and norms that prize independence over interdependence.

The prediction for India. India is left out of the ranking because it is a complex case, and the framework predicts it will become a major EOPV society as adoption accelerates: procedural breakdown, high tech fluency, a credentialing culture, and English-language access to the tools.

What follows if the paper is right

Mental health. High loneliness plus easy AI access forms a flywheel: more deference means less practice at self-authorship; less practiced agency means more anxiety under uncertainty; pseudo-connection means fewer real bids for connection; social muscle atrophies and the world feels harsher by comparison; the user returns to the AI. The World Health Organization already treats loneliness as a public health issue with mortality risk comparable to smoking. A just-good-enough substitute removes the discomfort that would otherwise push someone toward a real person.

AI design. If relational density is the defense, AI systems should be built to strengthen human relationships, and any design that substitutes for them works against the user. The paper's three design implications: time-bounded availability, since always-on systems invite always-on consultation and deliberate friction may increase healthy use; active deferral to humans for decisions that benefit from relational context; and plain statements of uncertainty where a confident-sounding answer would invite deference. This is the League's position in one line: build AI that reduces the need for AI.

Policy. Societies with high scores should invest in thick offline defaults: urban design that produces casual contact, support for multigenerational housing, civic institutions that take back the adjudication functions now outsourced to AI, and curricula that teach collective deliberation, the skill solitary information processing never builds.

Where the work goes next

• Consistent loneliness measures. Countries measure loneliness differently, and Gallup's single item ("Did you experience loneliness a lot of the day yesterday?") captures acute loneliness; the chronic kind needs its own instrument.
• Ties inside the household. Multigenerational households can hide weak ties when members are emotionally distant or the household is geographically isolated, so the density measure needs a within-household term.
• Longitudinal validation. The rankings are cross-sectional; the test is whether ADRI scores predict deference trajectories over time.
• Behavioral indicators of deference. Separating healthy AI use from deference needs observable markers, and the fastest route to them is a product that already logs the behavior. That instrument is the next piece of work.

The structural punchline stays: loneliness makes AI a self, corruption makes it a governor, meaning collapse makes it a compass, and the societies where all three converge are the ones to watch. The defense the paper points at is thicker humanity: shared meals, multigenerational households, neighborhood ties, and the reciprocal obligations that bind a community.

Sources cited in the draft: Cosmos Institute, "The Claude Boys" (2025); Gallup, Global State of Social Connections (2023); Hofstede, Hofstede & Minkov, Cultures and Organizations (3rd ed., 2010); Breakstone et al., "Students' Civic Online Reasoning" (Educational Researcher, 2021); World Health Organization, Commission on Social Connection (2023). The draft's bibliography also lists OECD Society at a Glance (2024), the World Happiness Report (2024), Parasuraman & Riley (1997), Dietvorst, Simmons & Massey (2015), Logg, Minson & Moore (2019), Skjuve et al. (2021), Bowlby (1969), Putnam (2000), Murthy (2023), Lee et al. (2022), and Cacioppo & Cacioppo (2014). Full references in the PDF.