Amartya Sen

In the [YOU] on AI Field Guide

The cycle that began with [YOU] on AI documents a twenty-fold productivity multiplier achieved by engineers in Trivandrum. Sen's framework identifies the question the number does not answer: did the multiplier expand or contract the capability set of the engineers who achieved it? An engineer who produces twenty times more code but works twenty times more hours has increased output without increasing well-being. An engineer who produces twenty times more code but has lost the developmental friction through which expertise is built has increased output while potentially decreasing capability. An engineer who produces twenty times more code but whose expanded productivity is captured entirely by the employer, with no corresponding expansion of the engineer's own freedom to choose how to work, has increased output while the distribution of the gains remains unchanged. These are not hypothetical concerns. The Berkeley study Segal documents found exactly this pattern.

The cycle's democratization narrative—the developer in Lagos who can now access the same coding leverage as an engineer at Google—is evaluated by Sen's framework with a precision that neither the optimists nor the pessimists have achieved. Access is not entitlement. Entitlement is not capability. The developer in Lagos may have access to the tool but lack the conversion factors—reliable electricity, broadband connectivity, financial infrastructure, legal protection, time free from subsistence labor—that would allow the access to convert into expanded substantive freedom. The tool is the food. The distribution system is the entitlement. The entitlement is what the technology industry consistently fails to build or measure.

Sen's concept of adaptive preferences—the mechanism by which preferences adjust to deprivation until the satisfied user no longer values what she has lost—is the analytical instrument the cycle most needs. The Berkeley workers who reported that AI made their work more productive and more engaging while exhibiting task seepage, boundary erosion, and accumulating fatigue were expressing genuine satisfaction. But their preferences had adapted to a contracted capability set. They valued what the tool provided and had ceased to value what the tool had eroded. User engagement metrics cannot detect this mechanism. Sen's framework was designed to.

His five instrumental freedoms—political freedoms, economic facilities, social opportunities, transparency guarantees, and protective security—constitute a checklist for evaluating whether the AI transition is expanding substantive freedom rather than merely formal access. Each category identifies a dimension of institutional infrastructure that formal tool availability does not guarantee. Each is unevenly distributed. Each represents a domain of construction that is at least as important as the technological construction the AI industry is performing with such speed and confidence.

Origin

Amartya Sen was born in 1933 in Santiniketan, the university town in West Bengal founded by Rabindranath Tagore, into a family of scholars and teachers. He was nine years old when the Bengal famine killed between two and three million people. He watched laborers appear at the doorstep of his family's home, skeletal and pleading. He watched people die in the streets. The experience marked him not with trauma alone but with the specific burning need to understand why—and the answer he eventually produced, decades later, was the demonstration that the famine was not caused by a shortage of food. The food existed. What had collapsed was entitlement—the economic and institutional machinery by which people gained access to the food that was available. Wartime inflation had destroyed the purchasing power of rural laborers. Speculative hoarding had removed rice from the market. The colonial government had prioritized military supply chains over civilian distribution. People starved surrounded by food.

This insight became the foundation of his analytical framework: the critical variable is not the existence of a resource but the institutional and social machinery that translates the resource into human welfare. Trained at Presidency College, Calcutta, and then at Cambridge, where he was deeply influenced by classical political economy and analytical philosophy, Sen built his capability approach over four decades of work—from his early studies of social choice theory and famine through the formal architecture of Development as Freedom (1999) and The Idea of Justice (2009). He received the Nobel Memorial Prize in Economic Sciences in 1998 for his contributions to welfare economics, the first Asian economist to receive it.

The capability approach was developed in dialogue with philosopher Martha Nussbaum, who provided a list of central human capabilities that Sen deliberately declined to specify, on the grounds that the selection of relevant capabilities should be left to democratic deliberation rather than expert determination. This deliberative commitment is itself a safety mechanism: it ensures that the question of which freedoms matter most cannot be settled by any single group claiming technical authority to answer a question that is irreducibly political.

Key Ideas

The capability approach. The capability approach redefines human welfare as the substantive freedom to achieve functionings one has reason to value. A functioning is a state of being or doing—being well-nourished, being educated, participating in community life, engaging in creative work. A capability is the real freedom to achieve a functioning. The distinction matters because a person can have a capability without exercising it, and the freedom to choose not to exercise a capability is itself valuable. Applied to AI: the question is not whether a person uses AI to produce creative work but whether the person has the substantive freedom to engage in creative work, with or without AI.

The evaluative space problem. The most consequential decision any society makes about a new technology is not whether to adopt it but how to evaluate its impact. The technology industry evaluates AI in the output space: parameters, benchmarks, revenue, productivity multipliers. Sen's framework evaluates it in the capability space: the full range of genuinely achievable functionings from which people are free to choose. The evaluative space determines what is visible and what is invisible. Evaluate in the output space, and you will optimize for output while the most important costs—the narrowing of capability sets, the erosion of tacit knowledge, the contraction of substantive freedom—remain undetected.

Conversion factors and the famine parallel. Conversion factors are the conditions that determine whether a resource translates into a capability. AI capability is abundant. The constellation of conditions under which AI capability translates into expanded human freedom is not: reliable infrastructure, educational preparation, financial security, institutional structures that channel productivity gains toward human development rather than pure output extraction, cultural conditions that recognize higher-order capabilities, and political conditions that give people voice in determining how the technology is deployed. Each conversion factor follows the contours of existing inequality. Each determines whether the AI revolution expands freedom or merely expands output.

Formal freedom versus substantive freedom. Substantive freedom is the real opportunity to achieve something, not merely the absence of prohibition. The technology industry's celebration of AI democratization is almost entirely a celebration of formal freedom: no one prevents the developer in Lagos from using Claude Code. The substantive freedom question is different: Can she actually build a viable product? Does she have the electricity, connectivity, financial infrastructure, legal protection, educational preparation, and market access necessary to convert tool access into a life she has reason to value? Formal freedom opens the door. Substantive freedom builds the path.

Adaptive preferences and the satisfied user. Adaptive preferences are preferences that have been shaped by deprivation: the person who has never had access to education may not value education, because the preference has adapted to the constraint. The developer who has always used AI to generate code may not value the deep architectural understanding that comes from years of manual coding, because the understanding has never been developed and its absence—masked by the tool's competence—has never been felt. User satisfaction metrics cannot detect this mechanism. They record what people prefer, not whether the preferences reflect genuine choice or constrained adaptation.