Loose Mind, Tight Society: The Fault Line in India’s AI Ambition

(This article was originally published on Substack)

Culture, institutions, and why so much of India’s most original talent does its best work abroad

A puzzle in two questions

Few countries have exported as much original talent as India, and few have retained so little of it. India-born founders lead 96 American billion-dollar startups, and Indians have co-founded 156 unicorns outside India, against 61 within it. When researchers traced the careers of the authors of papers accepted at NeurIPS, the most selective conference in artificial intelligence, they found that 12% of the world’s top AI researchers had completed their undergraduate education in India, and that almost none of them were working there. Indian-origin executives run several of the world’s largest technology companies, and two of the eight authors of the 2017 paper that introduced the transformer, the architecture underlying modern language models, studied engineering in India before writing it in California.

The pattern provokes a pair of questions that are rarely asked together. If Indian culture were holding Indians back, why do Indians build so much, so boldly, once they leave? And if Indian culture were propelling them, why do so few of them do their boldest work at home? The questions matter more now than at any time in the past half-century. Artificial intelligence is the first general-purpose technology in which the decisive inputs, after capital and computing power, are a few thousand exceptional researchers and engineers, and the country that has supplied a remarkable share of them has captured a strikingly small share of the resulting work.

Public debate in India tends to answer the two questions separately, and badly. One camp points to Panini’s grammar, the decimal place-value system, or the logical traditions of Nyaya and declares an inherited genius for abstraction that merely awaits its moment. A second points to rote learning, steep hierarchy, and the timidity of Indian organizations and concludes that something in the culture suppresses originality. A third, surveying Silicon Valley, decides that Indians flourish everywhere except in India and leaves the observation there, as though it explained itself. Each view holds a part of the truth, and each mistakes its part for the whole.

This article argues for a more demanding account. It makes five claims. First, cultures do shape habits of thought, and cultural psychology has produced serious evidence that they do; the differences are learned, not innate, and they are styles with trade-offs rather than rankings of ability. Second, the evidence about India specifically is much thinner than either its admirers or its critics assume, but what exists points to a distinctive and paradoxical profile: a high tolerance for ambiguity, contradiction, and plurality in the realm of ideas, combined with tight enforcement of norms in the realm of social conduct, family, and hierarchy. The shorthand used here for that profile is “loose mind, tight society”. Third, when the traits commonly attributed to Indians are sorted honestly, a few look like genuine advantages, several are double-edged, and the most damaging are not features of how Indians think at all, but of how Indian institutions allocate authority, money, and the cost of failure. Fourth, the behaviour of Indians abroad, the return of Indian companies to India, the experience of East Asia, and the performance of scientists who return home all suggest that institutions explain much more of the pattern than culture does. Fifth, because institutions can change far faster than cultures, the pattern is open to deliberate policy, and the present moment, in which the United States has made itself markedly less welcoming to Indian students and professionals, offers an unusual opportunity to change it.

Two cautions run through what follows. People who emigrate are a selected group, more educated, more ambitious, and more tolerant of risk than those who stay, so any comparison between Indians abroad and Indians at home overstates what institutions alone can explain. And a large part of the advantage of working abroad is simply money and scale, which no institutional reform can replicate overnight. A third caution precedes both. India, with its hundreds of languages, its regional histories, its castes and religious traditions, is less a single culture than a civilization of many. Any claim about “Indian” thinking must first survive that objection, and the argument here tries to respect it.

What cultural psychology has established

For most of its history, academic psychology assumed that a mind is a mind. A landmark 2010 review exposed how narrow that assumption’s foundations were: between 2003 and 2007, nearly all the people studied in leading psychology journals, 96%, came from countries home to only 12% of humanity, and almost 70% came from the United States. The authors, Joseph Henrich, Steven Heine, and Ara Norenzayan, called these subjects WEIRD, for Western, educated, industrialized, rich, and democratic, and concluded that they were about as unrepresentative a sample as could be found for drawing conclusions about humanity. Much of what had passed for human nature was, on inspection, the psychology of American undergraduates.

When researchers looked beyond that population, they found systematic differences. The psychologist Richard Nisbett and his collaborators documented what they called analytic and holistic styles of thought. In one well-known line of experiments, Western participants fixed on the central object in a scene, whereas East Asian participants took in more of its setting and surroundings. Holistic thinkers tend to group things by their relationships, analytic thinkers by categories and rules; holistic thinkers explain behaviour by situations, analytic thinkers by dispositions. Critics have rightly warned against overgeneralization, since variation within any culture is large and the distributions of different societies overlap heavily, but the core findings have been replicated many times.

The most ingenious attempt to explain where such differences come from is the “rice theory” of Thomas Talhelm and his colleagues. Testing 1,162 people across China, they found that people from China’s rice-farming south thought more interdependently and holistically than those from the wheat-farming north, and that the difference held even between adjacent counties straddling the rice-wheat line. Rice paddies demand intense cooperation in labour and the shared management of irrigation; wheat can be farmed more independently. Two further findings in Talhelm’s doctoral work bear directly on India. The rice-wheat difference in thinking style was also found in India, along its own rice-wheat divide, and that holistic thinkers in fact scored better on logic problems, suggesting that the difference is one of style rather than ability. The same work found that rice-growing Chinese provinces produced fewer patents, after controlling for income, consistent with Henrich’s argument that analytic, individualistic habits of mind go together with invention.

A third body of work, led by Michele Gelfand, asks not how people think but how tightly societies enforce their social norms. In a study of 33 nations, tight cultures were defined as having numerous strong norms and little tolerance for those who break them. India was among the tightest, alongside Pakistan and Malaysia, while the United States sat toward the loose end. Tightness tends to arise in societies that have faced disasters, epidemics, crowding, scarcity, and war, as a means of coordinating survival under threat. Gelfand’s later work found that tight cultures are strong on order and self-discipline but weaker on creativity, though other research suggests that tight cultures can be creative by drawing on deep expertise in a field, by perfecting within a tradition rather than breaking from it.

Behind these findings lies a conceptual question that public debate usually skips: what exactly is the relationship between culture and institutions? The economists Alberto Alesina and Paola Giuliano, in the most comprehensive review of the question, distinguish the two cleanly, treating culture as a society’s shared values and beliefs, its unwritten rules, and institutions as its written ones of law, regulation, and organization. Their survey finds causation running in both directions: cultures shape the institutions societies build, and institutions, once built, reshape the cultures that live under them. Alberto Bisin and Thierry Verdier have shown formally that the two can reinforce each other, so that their joint effect exceeds what either could achieve alone, or stand in for each other, so that it is smaller. And the evidence does not support the common assumption that culture is the slow variable and institutions the fast one in every case. Attitudes toward redistribution in East Germany, for example, shifted rapidly after reunification brought a new institutional order. Some cultural habits are deep and durable; others turn out to be the residue of incentives, and change when the incentives do.

Three lessons follow for any inquiry into India. The first is that cultural differences in thought are real, measurable, and rooted in ecology and history: in how people farmed, what threats they faced, and how they organized cooperation. They are learned, not innate. The second is that these differences are styles with trade-offs, not rankings of ability. A holistic, context-sensitive mind may be better at some tasks and worse at others; a tight culture may execute brilliantly and invent reluctantly. Any claim that a culture is simply better or worse at thinking misreads the evidence. The third is that culture and institutions cannot be understood apart from each other. To ask whether a pattern is “cultural” or “institutional” is often to ask the wrong question. The better question is which part of the pattern sits in durable values, which part sits in rules that could be changed, and how each is sustaining the other.

What is actually known about India

India was largely absent from the studies that built cultural psychology. Even large cross-national projects typically sampled one city or region per country, a serious limitation when the country in question holds 1.4 billion people and more internal diversity than most continents. Much of what is said about “the Indian mind” therefore rests on classical texts, anthropology, and a modest body of Indian psychological research rather than on large experimental studies. What does exist is suggestive, and it points in a surprising direction.

The headline indicators appear contradictory. On Geert Hofstede’s widely used dimensions, India scores high on power distance, at 77, but only 40 on uncertainty avoidance, which indicates a moderate comfort with ambiguity and a tolerance of the unplanned. Yet in Gelfand’s study, India appears among the tightest cultures in the world, with strong norms and little tolerance for deviation. A society that is relaxed about uncertainty but strict about conduct does not sit easily on any single scale. The contradiction is the most interesting fact in the evidence, and it deserves to be explained rather than averaged away.

Indian psychologists have long argued that the explanation lies in a capacity to shift between modes depending on context. Over five decades of research, J. B. P. Sinha concluded that Indians move between collectivist and individualist ways of thinking and acting as circumstances require, attentive to the place, the moment, and the people involved, and that Indian culture tends to accept paradoxes where Western cultures try to resolve them. Sinha was frank about the costs: this coexistence of opposites can breed hesitation and half-hearted action, yet also makes Indians inventive through constant adjustment to their circumstances. A review of studies on cultural diversity within India found some widely shared preferences, including deference to status and power and the central place of personal relationships, but also large differences between regions, which its authors linked to different levels of development.

The poet and scholar A. K. Ramanujan gave the most influential account of the pattern. In a 1989 essay, borrowing terms from linguistics, he proposed that Indian thought is “context-sensitive” rather than “context-free”. Where much Western ethics seeks rules that apply to everyone, classical Indian codes made what is appropriate depend on a person’s birth, occupation, and stage of life; the right action for a warrior differed from the right action for a priest, and the duties of youth from those of old age. Ramanujan insisted that no culture is purely one or the other. He also observed that India’s great counter-movements, the pursuit of liberation, renunciation, and forms of devotion that ignored caste, were efforts to escape context and reach toward the universal. The context-free impulse was never absent from India; it was the dissenting voice within a context-sensitive mainstream.

Amartya Sen added a further correction. He insisted on India’s long tradition of public argument and heterodoxy, and rejected the habit of seeing Indian traditions as uniformly religious, hostile to science, wholly hierarchical, or credulous. The traditions usually gathered under the word “Indian” include Nyaya logicians and Buddhist sceptics, Jain thinkers who built a theory of the partiality of every viewpoint and materialist Charvakas who denied any world beyond this one, bhakti poets who mocked caste and Brahmin ritualists who enforced it, Mughal courts that staged interfaith debates and Sikh gurus who rejected both priesthood and renunciation. Any claim about the Indian mind must survive the fact that India has argued with itself, at length and in many languages, about almost every question of mind.

The evidence also undermines some flattering popular claims. India is often described as a nation of multilingual people, yet according to the 2011 census only 26% of Indians are bilingual and 7% trilingual. What is extraordinary is not the multilingualism of the average Indian but the linguistic diversity of the country: two Indians chosen at random will speak different native languages more than nine times in ten. India’s linguistic advantage is a feature of its society, not necessarily of the individual mind. The same distinction, between the properties of a society and the properties of the people in it, turns out to matter throughout this argument.

Loose mind, tight society

Taken together, the accounts of Sinha, Ramanujan, and Sen suggest a hypothesis that reconciles the contradiction in the survey data. In the realm of ideas and belief, Indian culture is unusually loose: tolerant of ambiguity, comfortable with contradiction, accustomed to holding several frameworks at once, and possessed of a long tradition of argument in which no position is beyond dispute. In the realm of social conduct, family, and hierarchy, it is unusually tight: strict about who may do what, deferential to age and rank, attentive to the opinions of kin and community, and severe in its judgment of visible failure. A loose mind lives inside a tight society. The hypothesis would explain why the same civilization that produced a logic of many-sidedness also produced one of the most rigid social orders the world has known, and why its people can be intellectually adventurous and socially cautious at the same time. It remains a hypothesis. No large study has tested it directly, and it should be held with the tentativeness that implies.

If the hypothesis is to be more than a slogan, it must specify how a tight society suppresses bold work, and it must generate predictions that could be wrong. Five mechanisms are plausible. The first is the cost of deviance and failure: when visible failure carries heavy social, financial, or legal penalties, rational people avoid the risks that original work requires, however original their ideas. The second is deference to seniority: when resources, credit, and decisions flow by rank rather than by the merit of ideas, junior researchers, who produce a disproportionate share of breakthroughs in fast-moving fields, cannot pursue their own agendas. The third is permission before action: systems that require approval in advance, rather than holding people accountable afterwards, slow every experiment and favour those with the connections to obtain permission. The fourth is narrow, high-stakes evaluation: examination-centred selection and metric-driven assessment reward getting known answers right rather than asking new questions. The fifth is family and community expectation, which steers talented young people toward safe and prestigious careers rather than uncertain ones, and which can make a failed venture a matter of family shame rather than individual learning.

Two features of Indian social organization show with particular clarity that tightness is not merely a mood but an institution with economic consequences, and that its effects cut both ways. The first is the caste and community network. The economist Kaivan Munshi, in a comprehensive review, concludes that caste shapes economic life at every stage, from schooling and university through the labour market to old age. With Mark Rosenzweig, he showed that sub-caste networks provide their members with mutual insurance, and that families who marry outside the group or move away forfeit that protection, which holds mobility back when alternatives of comparable quality are unavailable. Their earlier work found that caste-based labour networks had confined whole communities to a narrow range of occupations across generations. Yet the same networks have also been engines of enterprise, supplying the trust, credit, and information that carried whole communities into new industries, from diamond trading to real estate. The community network is a tight institution in its purest form: it rewards loyalty, punishes exit, and makes risk-taking possible only within its boundaries. It is a superb mechanism for coordinated advance along a known path and a poor one for the lone deviant with an unproven idea.

The second is the treatment of women, and it is the most costly single expression of a tight society for India’s capacity to innovate. India has one of the highest shares of women among science and engineering graduates in the world, 43% at the bachelor’s level and nearly 50% at the master’s and doctoral levels. Yet women make up only about 18% of its research and development workforce, and only 8% of the faculty at the Indian Institute of Science and 11-13% at the IITs. The pattern is unusual. In most countries the pipeline leaks early, because girls are steered away from science at school; in India it leaks late, at the transition from doctorate to career, where the completion of a PhD often coincides with pressure to prioritize marriage, children, and household duties. Here the mechanism is not cognitive at all. India educates its women scientists and then loses most of them to norms about what a woman’s life should contain. No account of India’s talent problem that ignores this is complete, and no reform that ignores it will capture half of the talent available.

Each mechanism generates predictions that distinguish it from the main alternative explanation, which is simply resources. On the resource explanation, India’s best people leave because laboratories abroad have more money, more computing power, and better peers, and giving Indian institutions more money should therefore keep them. On the tightness explanation, money alone will not suffice: well-funded institutions that remain hierarchical and failure-averse will still lose their boldest people, while modestly funded institutions that grant autonomy and tolerate failure will punch above their weight. The hypothesis therefore predicts four things. The same people should behave more boldly when they move into looser institutions, whether abroad or into looser pockets inside India such as startups and certain research groups. Loosening a specific institution, such as bankruptcy law or a funding agency’s rules, should produce a measurable increase in risk-taking. Regions and organizations within India that are less tight should show more original activity, holding resources constant. And because institutions can change faster than deep cultural values, the gap between Indians at home and abroad should narrow when institutions change, even if attitudes have not yet moved.

The honest position is that both explanations are partly right. Resources matter enormously, and India lags far behind the leading countries in capital, computing power, and research spending. But the evidence reviewed in the sections that follow suggests that institutional design explains a substantial part of the pattern that resources alone do not, and, crucially, that it is the part most within India’s power to change quickly.

Sorting the traits

If there are Indian cognitive and social styles relevant to artificial intelligence, they will not be found in a single trait but in a profile. It helps to be clear, too, that “AI work” is not one activity but several: frontier research that invents new methods, large-scale engineering that builds and runs systems, the adaptation and evaluation of models for particular languages and settings, and the deployment of systems into businesses and governments. A trait can help in one and hinder in another. Sorted by the strength of the evidence that they exist and by their likely effect on these different kinds of work, the traits commonly attributed to Indians fall into three groups.

The first group contains genuine advantages, and almost all of them belong to the society rather than to the individual. India’s linguistic diversity, the density of its social and regional contexts, and the sheer scale and variety of its population are features of the country that matter greatly for building and testing systems that must work across languages, scripts, dialects, and cultures. Context-sensitivity, Ramanujan’s characteristic of classical Indian thought, plausibly helps in designing how models should behave for very different users, and in evaluating whether they do. A long intellectual habit of building on existing frameworks, of extending and commenting on a foundational text rather than discarding it, suits the work of adapting general-purpose models to particular purposes. Adaptation under constraint, the ability to make things work with scarce resources, is often claimed and largely anecdotal, but it is plausible, and it is mostly a product of environment rather than temperament. What none of this establishes is that any individual Indian thinks better. Even the evidence on multilingual individuals is weaker than often supposed: preregistered studies have found no creative advantage for bilinguals. What India has is not a better mind but a richer environment in which to build technologies that must work for everyone.

The second group contains traits that are double-edged, and they share a common shape. Examination-centred schooling, family-defined ambition, and tight social norms all favour executing well within a defined framework over questioning the framework itself. Each has real virtues. Examination culture produces formidable fundamentals in mathematics and engineering, and research on East Asian learners shows that memorization can go together with genuine understanding, so the stereotype of the rote learner who understands nothing is unfair. Family-defined ambition has produced one of the largest technical talent pipelines in human history, sustained by enormous household sacrifice. Tight norms produce reliable execution, coordination, and discipline. But a profile tuned for execution within a framework suits building and deploying technology better than inventing it, and it is the inventing that commands the largest returns at the frontier. The double edge is visible in India’s technology industry itself, which became superb at delivering what clients specified and slow to specify anything of its own.

The third group contains the headwinds, and the most important finding of this exercise is that they are institutional rather than cognitive. Steep hierarchy is strongly evidenced in Indian organizations, and its effects are measurable: a meta-analysis found that hierarchy modestly reduces team performance, chiefly by generating conflict, and a separate study found that power distance harms the generation of ideas while helping their implementation. That is precisely the profile of a culture that executes well and invents reluctantly. Fear of failure is equally well documented: India ranks second of 47 economies in the share of adults who say that fear of failure would stop them from starting a business. One trait, holistic thinking, belongs in none of the three groups with confidence. The Indian evidence on it is emerging and regional, and while the Chinese evidence links holistic thinking to fewer patents, the Indian evidence suggests holistic thinkers reason at least as well. Its net effect is simply unknown.

It is worth dwelling on why the headwinds should be classed as institutional. Hierarchy and fear of failure are not facts about how an individual mind processes information. They are facts about how organizations and families allocate authority, how they distribute credit, and how heavily they punish the person whose venture fails. A young researcher who defers to a senior professor is not thereby less able to think; she is responding rationally to a system in which disagreement is costly and the senior professor controls her funding, her authorship, and her next position. A graduate who chooses a safe job over a startup is not thereby less imaginative; he is responding to a system in which failure is slow, expensive, socially visible, and potentially criminal. Change the costs, and the behaviour should change. That is a testable claim, and the rest of this article examines the evidence for it.

Where Indian ambition goes

The clearest evidence on the outflow of India’s most ambitious minds in artificial intelligence comes from the Global AI Talent Tracker, which follows the education and employment of researchers whose papers are accepted at NeurIPS. In its first edition, based on 2019 data, it found that India was the undergraduate home of 12% of the world’s top researchers but employed almost none of them. By 2022 the picture had begun to shift: whereas nearly every Indian-educated top researcher had gone abroad in 2019, about one in five was staying to work in India by 2022. China and India together supply more than half of the elite AI researchers who work outside their country of origin, and the tracker found that the great majority of foreign researchers who complete a doctorate in the United States remain there afterwards. The doctorate, in other words, appears to be the decisive point of departure.

The pipeline begins with students. India’s Ministry of External Affairs counted 1.88 million Indian students in 153 countries at the end of 2025, including those at school level, with Canada and the United States the largest destinations. The flow has begun to change sharply. Departures to study abroad fell by about 15% between 2023 and 2024, from roughly 893,000 to 759,000, with the steepest falls in Canada, the United Kingdom, and the United States, down 41%, 27%, and 13% respectively, while departures to Germany rose by almost half. Enquiries from Indian students about studying in America fell by more than 46% in a single year. Tighter visa rules, political hostility to immigration, and rising costs in the traditional destinations appear to explain most of the change.

At home, the base from which original research must come remains thin. India’s researchers per million people rose from 110 in 2000 to 255 in 2017, a doubling, but still a small fraction of the levels in China or the advanced economies. The government’s own Economic Survey observed that India falls short of what its level of development would predict in the number of research personnel, with among the lowest numbers of any of the world’s ten largest economies, and that an unusually small share of its researchers work in business. A country that sends a great many of its best minds abroad and employs relatively few researchers at home is not short of talent. It is short of places for talent to go.

Two features of this evidence matter for everything that follows. The first is that the outflow is concentrated at the top. India does not lose its average engineers; it employs millions of them in its services industry and in the capability centres of global firms. It loses a disproportionate share of those most likely to produce original research and to found new kinds of companies, the people at the tail of the distribution whose work generates the largest returns. In a field like artificial intelligence, where a single research group can redefine an industry, the composition of the outflow matters far more than its size.

The second is that the pattern is not fixed. Retention of top researchers improved between 2019 and 2022, a period in which global companies expanded research laboratories and capability centres in India and a domestic startup ecosystem matured. Students are already changing where they go. Founders, as the next section shows, have begun bringing their companies home. Flows of talent respond to incentives at both ends, and the incentives at both ends are now moving. The question is whether India’s institutions will move with them.

The evidence that institutions bind

The most powerful evidence about culture, institutions, and Indian innovation comes not from psychology laboratories but from a series of natural experiments in which the same people, or people from similar cultures, faced different institutions. Five of them bear directly on the question.

The first is the diaspora itself. If an Indian upbringing suppressed boldness, the founders and researchers who left would carry the suppression with them. Plainly they do not, or not decisively; the 96 American unicorns led by India-born founders are not the work of people who learned timidity at home and could not unlearn it. Two cautions apply. Migrants are a selected group, and immigrants in general are about 80% more likely to start firms than native-born Americans, so some of the diaspora’s dynamism reflects who leaves rather than where they go. And the diaspora’s success depends on the ecosystem it joins: abundant risk capital, forgiving bankruptcy rules, dense networks of experienced founders and investors, and a social environment in which failure is survivable and often invisible to one’s extended family. Selection and environment together explain most of the gap. Neither requires any change in how the founders think.

The second experiment is sharper, because it holds the people constant. For a decade, ambitious Indian founders incorporated their companies abroad, in Delaware or Singapore, because foreign capital, foreign listing venues, and foreign legal systems offered more than India’s did. Then the rules changed. Amendments to company law and foreign exchange rules in 2024 eased the path for startups to move their legal home to India, regulators relaxed listing requirements, and India’s public markets deepened. Within a few years, PhonePe, Groww, Zepto, Razorpay, Meesho, and Pine Labs had begun or completed moves back, PhonePe and Groww paying about ₹8,000 crore and ₹1,340 crore respectively in taxes to do so, and in 2026 Flipkart completed its redomiciliation from Singapore ahead of a planned listing. The founders did not change their culture between 2020 and 2026. Their institutional environment changed, and their decisions followed within a few years, at considerable cost to themselves. The limitation is that a reverse flip concerns where a company is registered and listed, not where its most original work is done. But as a demonstration of how quickly Indian behaviour responds to Indian rules, it is hard to improve upon.

The third experiment is East Asia. In the 1950s and 1960s, Japanese manufacturers were widely derided as makers of cheap imitations. South Korea was long described the same way, until its shift from imitation to innovation in the late 1970s and early 1980s, propelled by state policy, the conglomerates, investment in education, and deliberate strategies for exporting and acquiring technology. The Korean case contains a precise institutional lesson. The Korea Institute of Science and Technology, founded in 1966, was designed explicitly as a centre for reversing the brain drain; to bring scientists home, the government offered special incentives and guaranteed both the autonomy of their research and its financial support. Its success persuaded the government to abandon policies that restricted study abroad and to build a series of research institutes on the same model. Taiwan’s experience was similar. The returning engineers whom the economic geographer AnnaLee Saxenian called the “new Argonauts” built Taiwan’s technology industry around the Hsinchu Science Park only after the state supplied what the market had not: tax breaks and matching funds that drew venture capital into the park, and an ecosystem with Silicon Valley’s ease of entry and exit. China, dismissed as a copycat economy well into the 2000s, now educates 57% of the world’s top AI researchers. These are precisely the societies that cultural psychology describes as holistic, tight, examination-centred, and hierarchical. Their cultures did not prevent innovation once institutions and investment changed.

The fourth experiment is the most rigorous, because it compares returnees with similar scientists who stayed abroad. A 2023 study in Science followed 339 young scientists who returned to China under the Young Thousand Talents programme. After returning, the recruits published 144% more last-authored papers than comparable peers who remained overseas, last authorship being the marker of who leads a research group, because they were far more likely to become independent investigators pursuing their own agendas. The authors traced the gain to better access to funding and to larger research teams, and noted that young scientists in the United States and Europe often struggle to win independent funding. Two further findings sharpen the lesson. The gains were concentrated in resource-intensive fields such as chemistry and the life sciences and did not appear in mathematics and theoretical physics, where North America’s draw lies in its academic freedom. And the programme, which consumed less than 0.5% of China’s academic research budget, still failed to attract the very best, who commanded far larger grants abroad. Early independence plus resources at home can make returnees more productive than they would have been abroad. Freedom and scale determine whether the very top will come.

The fifth body of evidence complicates the language of “brain drain” itself. A study of Indian patent data by Ajay Agrawal, Devesh Kapur, John McHale, and Alexander Oettl found that inventor emigration on balance reduces domestic access to knowledge, but that access through the diaspora is especially valuable for India’s most important inventions. Emigrants are both a loss and a bridge. The economists Gaurav Khanna and Nicolas Morales went further, showing that American immigration policy helped create India’s software industry. Indian students trained in computer science hoping to work in America’s booming IT industry, and since visas were rationed, many of them stayed behind, where they built the Indian IT sector; when the probability of migrating rose, India experienced a “brain gain” in the skills its own industry could use. Many of those who did go returned when their visas expired, bringing know-how and connections. Saxenian documented the same circulation in Bangalore, where returnees worked with policymakers to reshape telecommunications regulation and promote venture capital. The prospect of leaving, it turns out, has been a powerful force in shaping what Indians learn at home.

Taken together, the five experiments support the institutional account while refining it. When Indian institutions offer independence, resources, and a path to scale, Indians use them, as the returnees, the reverse flips, and the improving retention of top researchers all show. Where those institutions are absent, Indians go where they exist, and a society that tolerates their departure but makes their return difficult turns circulation into drain. None of this proves that culture plays no role, and the research on the Indian case specifically remains thinner than it should be. But it strongly suggests that the most important variables are ones that policy can reach.

Five tight spots

The mechanisms by which a tight society suppresses bold work become concrete in five places where Indian institutions are measurably tighter than those that compete for Indian talent. None of them is a matter of temperament. Each is a matter of rules, budgets, and administrative habits.

The first is research funding. Original research needs money that arrives quickly, in adequate amounts, with the freedom to change direction when an experiment fails or a better question appears. India’s main new instrument, the Anusandhan National Research Foundation, was created in 2023 with a plan for ₹50,000 crore over five years, of which ₹36,000 crore was expected to come from private and philanthropic sources. A parliamentary committee found that it had spent none of its ₹2,000 crore annual allocation in its first two years, and only about 61% of its revised outlay in 2025-26. Recent reforms have cut compliance burdens and dropped a requirement that national laboratories share half the cost of projects, and they deserve credit. But a funding system that cannot spend its budget sends an unmistakable signal to a young researcher weighing a laboratory in Bengaluru against one in Boston. From her point of view, a foundation that does not disburse is a foundation that does not exist.

The second is the cost of failure. The Insolvency and Bankruptcy Code of 2016 was meant to make corporate failure fast and orderly, with a legal limit of 330 days for resolution. By June 2026 the average resolution was taking 757 days, and about three-quarters of ongoing cases had run past 270 days. Slow, expensive failure raises the cost of every risky venture, because founders and investors must price in the possibility of years trapped in a dead company, with capital frozen and reputations suspended. Personal guarantees, routinely demanded of small-business founders, can turn a failed venture into a lifetime of debt. Silicon Valley’s celebrated tolerance of failure is not a cultural mystery. It rests in large part on the fact that a failed startup there can be wound up quickly and cheaply, and its founder can begin again within months, often with the same investors.

The third is permission and criminal liability. A 2022 study by the Observer Research Foundation found that over half of India’s 1,536 business laws, and 37.8% of its 69,233 compliance requirements, carry the possibility of imprisonment, some 26,134 clauses in all. The Jan Vishwas Act of 2023 decriminalised 183 provisions across 42 Acts, and a second bill covering 784 provisions in 79 central Acts passed the Lok Sabha in 2026. These are real reforms. But most imprisonment clauses sit in state laws, which central reform does not reach. A regime in which a paperwork error can in principle lead to jail is the purest institutional form of a tight society, and its effect on behaviour is felt not mainly through prosecutions, which are rare, but through the background fear that discourages anyone from straying beyond the familiar.

The fourth is patient capital. India attracted well under 1% of global private investment in artificial intelligence in 2025. The government has responded with a ₹10,000 crore Startup India Fund of Funds 2.0, notified in April 2026 with an emphasis on deep technology, alongside a ₹1 lakh crore research and innovation scheme. These are meaningful steps. Yet India’s large domestic pools of long-term capital, its pension funds and insurers, still invest very little in venture funds, so the country’s deep-technology founders depend heavily on foreign investors whose priorities, time horizons, and appetite for Indian risk are set elsewhere. When those investors retreat, as they periodically do, Indian ambition retreats with them.

The fifth is hierarchy and evaluation. In Indian research institutions, steep hierarchy shows up as resources and credit flowing to senior figures, few genuinely independent positions for researchers in their late twenties and early thirties, and evaluation by publication counts, impact factors, and rankings rather than by the ambition of the question asked. A young scientist who returns from a postdoctoral position abroad often finds that she must attach herself to a senior professor’s group, request permission for purchases her foreign supervisor would have approved with a signature, and wait years for the kind of independence she already had. In fast-moving fields like artificial intelligence, where breakthroughs come disproportionately from the young, this is especially costly. The transformer was invented by a team whose members were mostly in their twenties and thirties, several of them without the seniority that Indian institutions would have required before entrusting them with a research agenda of their own.

The five tight spots reinforce one another. A young researcher who cannot obtain independent funding cannot build a group; a founder who cannot fail cheaply cannot try; a firm that fears criminal liability does not experiment; an ecosystem without domestic patient capital cannot hold its founders through a downturn; and a hierarchy that controls funding, authorship, and promotion makes disagreement with seniors a career risk. Each alone might be tolerable. Together they form a system that is very good at producing disciplined, capable people and very poor at giving them room to do something no one has done before.

Loose pockets at home

If the institutional account is right, India should already contain pockets where institutions were deliberately loosened, and those pockets should have produced work out of proportion to their resources. They exist, and they are among the proudest achievements of independent India.

The most instructive is the early Indian space programme. On 10 August 1979, the first experimental launch of India’s Satellite Launch Vehicle, SLV-3, failed when the control system in its second stage malfunctioned and the rocket fell into the sea. The project director was a young engineer named A. P. J. Abdul Kalam, who was devastated and wanted to resign. The chairman of the Indian Space Research Organisation, Satish Dhawan, persuaded him to stay and then faced the press himself, announcing that he took responsibility for the failure although Kalam had directed both the project and the mission. When the next launch succeeded, on 18 July 1980, Dhawan stayed in the background and sent Kalam to face the cameras. Kalam told the story for the rest of his life as a lesson in leadership. It is also a lesson in institutional design. In a tight society, the cost of a visible failure falls on the most junior person who can be blamed; Dhawan deliberately moved it to the most senior person who could absorb it, and in doing so lowered the price of risk for everyone beneath him. It is worth noting who did the work. Of the roughly 1,200 scientists and engineers on the SLV-3 project, hardly a handful had been educated abroad. This was not imported talent working in an imported culture. It was Indian talent working under different rules.

The second pocket is more recent: the teams that built India’s digital public infrastructure. When the Unique Identification Authority of India was created in 2009 to give every resident a biometric identity, its chairman, Nandan Nilekani, was employee number one of an organization that did not yet exist. He assembled a small team that combined experienced civil servants with technologists recruited from the private sector, including volunteers who came from Silicon Valley, kept design and strategy in-house, and rolled out the first identities within about fourteen months. Nilekani later described the identity authority and the payments body that built UPI, the National Payments Corporation of India, as practically startups owned by financial institutions. Whatever one thinks of Aadhaar’s later controversies, which were real, the speed and ambition of its construction are hard to reconcile with any theory that Indian institutions are incapable of boldness.

Other pockets point the same way: the founding generation of institutions such as the Tata Institute of Fundamental Research and the IITs, given unusual autonomy by their founders and their statutes; the software firms of Bengaluru and Hyderabad, which grew in a sector that regulation largely ignored in its early years; and the startup ecosystem of the past fifteen years, whose founders took risks their parents would have found unthinkable once venture capital made failure survivable. What these pockets share is not a distinctive culture. Their people came from the same families, schools, and communities as everyone else. What they share is a set of institutional features: a mission that gave purpose to risk, leaders who absorbed failure rather than passing it down, small teams with real authority, freedom from the ordinary rules of permission, and the presence of outsiders who brought different habits. Where those features were present, Indian talent behaved like talent anywhere. Where they were withdrawn, as the original spirit of many such institutions faded into bureaucratic routine, the talent behaved as tight institutions teach it to behave.

The existence of these pockets is the strongest domestic evidence for the argument of this article. It shows that the gap between Indians at home and Indians abroad is not a gap between two kinds of people but between two kinds of rules, and that India has already, more than once, written the second kind for itself. The problem is not that India does not know how to build loose institutions. It is that it has treated them as exceptions to be created for special missions rather than as a model for its research system as a whole.

Three traps

Any serious discussion of Indian culture and innovation must avoid three traps, each of which has caught intelligent and influential people, and each of which, in its own way, prevents the institutional reforms that the evidence calls for.

The first is chauvinism: the conversion of real heritage into imaginary capability. Its most durable example is the claim that NASA declared Sanskrit the best language for computers. The claim traces to a 1985 paper in AI Magazine by Rick Briggs, a researcher at an institute at NASA’s Ames centre, which compared the semantic analysis of the Sanskrit grammarians with the representation of knowledge in artificial intelligence. It was a serious and interesting paper. Briggs never claimed that Sanskrit was the most suitable language for computer software, yet a Union Home Minister repeated the NASA version in 2015, and an IIT Kharagpur linguist has described how a rigorous article mutated online into disinformation that could no longer be controlled. The same year, the Indian Science Congress hosted a paper claiming that aircraft invented in the Vedic age could fly between planets, although the text it relied upon had been examined and rejected by aerospace engineers decades earlier. Chauvinism does double damage. It discredits India’s genuine intellectual achievements, which are considerable, by associating them with fantasy. And it substitutes pride in the past for investment in the future, encouraging the belief that a civilization which once produced great mathematics need do nothing in particular to produce it again.

The second trap is the colonial dismissal, which treats Indian thought as mystical, derivative, or unworthy of serious study. Its classic statement is Thomas Babington Macaulay’s Minute on Education of 1835, which asserted that “a single shelf of a good European library was worth the whole native literature of India and Arabia”, a judgment pronounced by a man who could neither read nor speak any Asian language. The dismissal survives in subtler forms: in the habit, which Sen criticizes, of praising India for its spirituality while overlooking its traditions of logic, astronomy, mathematics, medicine, and political thought; and in the commercial assumption, widespread among multinational firms and Indian firms alike, that Indians are implementers rather than inventors. The psychologist Ashis Nandy argued that colonialism’s deepest effect was to colonize the minds of the colonized. An expectation that India’s role is to execute other people’s ideas can become self-fulfilling, shaping where multinationals locate their research, what Indian investors are willing to fund, what Indian universities dare to attempt, and what Indian engineers believe is possible for them.

The third trap is newer and more comfortable: using culture as an excuse. When disappointing outcomes are blamed on an “Indian mindset”, the structural causes escape scrutiny. A country that spends less than 1% of its GDP on research, denies its universities meaningful autonomy, disburses research funds late or not at all, and makes failure financially, socially, and sometimes legally ruinous will not produce bold research, whatever its culture. Blaming culture allows governments, universities, and firms to avoid the harder and more specific work of changing rules. It has a further, quieter cost. It tells young Indians that the problem lies in them, in their upbringing, their families, their habits of mind, when the evidence of the diaspora, the reverse flips, the returnees, and the loose pockets at home shows the same people behaving very differently once the rules change. A diagnosis that blames the patient for the disease does not merely fail to cure it; it discourages the patient from seeking a cure.

The three traps share a premise: that culture is destiny, whether glorious, inferior, or immovable. The chauvinist believes the destiny is greatness and needs no effort; the colonial dismissal believes it is mediocrity and cannot be escaped; the excuse believes it is fixed and therefore no one’s fault. The evidence reviewed here points the other way. Culture is a set of learned habits, shaped over centuries by ecology and history, sustained by the institutions people live within, and reshaped, sometimes surprisingly quickly, when those institutions change.

The window

For half a century, the United States was the default destination for India’s most ambitious graduates. In 2025 and 2026 that default weakened. Indians received 71% of H-1B approvals in fiscal 2024, so the American decision in September 2025 to impose a $100,000 fee on new H-1B petitions fell most heavily on them. A federal district court in Massachusetts vacated the fee in June 2026, and in July the First Circuit Court of Appeals refused to reinstate it while the government’s appeal proceeds, with briefing on the merits concluding in mid-October. The administration is still appealing, and review by the Supreme Court remains a real possibility. Combined with sharp falls in student visas and rising political hostility to immigration, the effect has been to make the American path uncertain for a generation of Indian students and researchers. Uncertainty is itself a cost. A researcher deciding where to spend the next decade of her working life cannot plan around a rule that may return, and a student choosing where to begin a doctorate is choosing, on the evidence of the talent data, where he is likely to stay.

Other countries moved quickly to exploit the opening. Canada’s 2025 budget set aside funds to recruit more than 1,000 leading researchers and announced a faster route for American H-1B holders, although the route had not opened as of mid-2026. China launched a new K visa for young science and technology professionals in October 2025. Indian students, meanwhile, are already diverting to Germany, Ireland, and other European destinations. The competition is not between India and the United States alone. It is among every country that wants the talent America is turning away, and several of those countries are better prepared than India to receive it.

India’s response has been real but modest. In June 2026 the Ministry of Education opened applications for the Prime Minister Research Chair scheme, which aims to bring at least 120 researchers of Indian origin home over five years, as young fellows, senior fellows, and research chairs, with set-up grants reported at up to ₹14 crore for the most senior positions. It joins older programmes such as the Ramanujan Fellowship and the VAJRA visiting faculty scheme. The contrast with China is instructive. Its Young Thousand Talents programme funded more than 3,500 young researchers between 2011 and 2017. India’s flagship scheme targets 120 over five years. The difference is not only of scale. The Chinese programme succeeded where it offered returnees early independence and the resources to run their own groups, and failed where the institutions they returned to could not match the freedom or funding available abroad. A scheme that recruits senior figures with large set-up grants into institutions that remain hierarchical and slow to disburse risks reproducing the weaknesses of the system it is meant to strengthen.

The window will not stay open indefinitely. American courts may settle the visa question, American universities will adapt, and other countries will absorb much of the talent now in motion. More fundamentally, a returning researcher’s decision depends less on the size of a set-up grant than on what happens in the years after it is spent: whether the next grant arrives on time, whether a failed project ends a career, whether a thirty-two-year-old can run a laboratory without a senior professor’s permission, whether a spouse, who in the Indian case is often a highly educated woman, can find work that uses her training. Recruitment schemes can bring people to the door. Only the institutions behind the door determine whether they stay, and whether those who stayed abroad hear, from those who returned, that coming home was worth it.

What would change the pattern

If the binding constraint is institutional, the remedies are institutional too. Seven changes, each tied to a mechanism identified above, would do most of the work. None is costless, but none requires India to become a different civilization.

The first is to fund people early and independently. India needs several hundred research positions each year for scientists under thirty-five that come with five to seven years of assured funding, the freedom to set their own agenda, and no requirement to work under a senior professor. The European Research Council’s starting grants and China’s young talent programmes show the model, and the Chinese evidence shows that it is independence, not money alone, that produces the gain. Such positions must also be administered by people who understand that a grant which arrives a year late has, for a young scientist, partly failed. Speed of disbursement is not an administrative detail. It is the difference between a research system that exists on paper and one that exists in a laboratory.

The second is to make failure cheap. Small companies and startups need a fast, low-cost route to closure measured in months, not the two years that insolvency now takes on average, and founders need protection from personal guarantees that turn a failed venture into a lifetime of debt. Decriminalization must reach the states, which account for the great majority of the 26,134 imprisonment clauses in business law, while the Union government accounts for about 5,239. A society becomes looser, in Gelfand’s sense, partly when the penalties for deviation fall. India’s legislators can lower those penalties directly.

The third is to replace permission with accountability. Research institutions should be able to buy equipment, hire staff, travel, and change project plans without seeking approval at each step, and should be audited on what they achieved afterwards. It is a change Indian scientists frequently ask for in private and one that receives little attention in public, because it produces no announcement, no new scheme, and no ribbon to cut. It would nonetheless do more for the daily experience of research than most new funding programmes.

The fourth is to mobilize domestic patient capital. The Fund of Funds 2.0 and the ₹1 lakh crore research and innovation scheme are important, but they remain government money, allocated by government processes. Allowing India’s large pension and insurance funds to commit a small share of their assets to venture and deep-technology funds, under prudent limits and with professional managers, would create a domestic source of long-term risk capital comparable in scale to what Indian founders now seek abroad, and would make Indian ambition less hostage to the moods of foreign investors.

The fifth is to stop losing women at the doctorate. India already educates women scientists at near parity; it loses them at the point where a research career begins and family expectations intensify. The remedies are concrete: childcare on research campuses, flexible early-career timelines that do not penalize the years in which children are born, grants that support re-entry after a career break, appointments that accommodate dual-career couples, and hiring committees held accountable for the gap between the share of women who earn doctorates and the share who are hired. Government schemes for women in science exist, but they are small relative to the scale of the loss. No other single reform would enlarge India’s pool of research talent so quickly.

The sixth is to treat the diaspora as a network, not only as a pool to be recalled. Joint appointments that let researchers divide their time between Indian and foreign institutions, co-funded projects with diaspora laboratories, and visiting positions with real research budgets would capture the bridging value that the patent evidence identifies and that Saxenian documented in Bangalore, while building the domestic base that emigration erodes. The goal is not to stop Indians from leaving, which would be neither possible nor desirable, but to ensure that leaving is the beginning of a circulation rather than the end of a connection.

The seventh is to scale and to measure. A programme to recruit 120 researchers over five years is a pilot, not a strategy. India should set a target in the thousands, fund it at a meaningful share of its research budget, and commission independent evaluation from the start, so that it can learn, as the Chinese study allowed China to learn, which recruits thrive and why. More broadly, India and its diaspora already contain the variation needed to settle many of the questions this article has raised, and several natural experiments are waiting to be read. Districts on either side of India’s own rice-wheat divide differ in thinking style but can be matched on income, urbanization, and access to universities and capital; comparing their patents, startups, and research output would show whether culture predicts innovation once structure is held constant. Indian engineers move constantly between services firms, the research laboratories of multinationals, and startups, so tracking what the same individuals produce under different hierarchies would separate the person from the institution in a way that cross-national comparisons cannot. Returning researchers offer a test of whether behaviour formed in one environment persists in another. Every major reform is an experiment: cross-country research has found that more forgiving bankruptcy laws did more to raise self-employment than economic growth did, and India’s own insolvency, funding, and decriminalization reforms can be studied in the same way. Finally, tightness, on Gelfand’s account, responds to the level of threat a society faces, and India is urbanizing, growing richer, and connecting its young to the world at extraordinary speed. Measuring whether norms are loosening across generations, especially in cities, would tell India not what its culture is, but what it is becoming.

None of these changes requires altering India’s culture. Each requires altering a rule, a budget line, a procurement procedure, or an administrative habit. That is the practical meaning of the claim that institutions, not culture, are the binding constraint. It is also, for anyone impatient for change, the hopeful meaning. Cultures change over generations and resist direction. Rules can be rewritten in a parliamentary session.

Culture as style, institutions as scale

The verdict of this inquiry can now be stated plainly. Cultures do produce different styles of thought, and those styles are learned, rooted in ecology and history, and best understood as trade-offs rather than rankings. The evidence about India specifically is thinner than it should be, but it points to a pattern worth taking seriously: a mind comfortable with plurality, paradox, and context, living inside a society that enforces its rules about conduct, hierarchy, and family tightly. Sorted honestly, the traits most relevant to frontier technology fall into three groups. The clearest advantages are societal, chiefly linguistic diversity and the density of context in which Indian products must work. Several traits are double-edged, favouring excellent execution within frameworks over the questioning of frameworks. And the main headwinds, hierarchy, the punishing cost of failure, the loss of women from research, and the restriction of risk to the boundaries of family and community, are features of how Indian institutions allocate authority and absorb failure rather than of how individual Indians think.

The natural experiments all point the same way. Indians abroad, working under looser institutions, build boldly. Indian companies came home within a few years of the rules changing. East Asian societies that cultural psychology describes as holistic, tight, and hierarchical became innovators once their institutions and investments changed, and they did it in part by bringing their emigrants home with promises of autonomy and support. Scientists who return to well-designed positions become more productive than peers who stay abroad. And India itself, in its space programme, its digital public infrastructure, and its startup ecosystem, has repeatedly shown that the same people behave very differently when a leader absorbs their failures, a team is given real authority, and the rules of permission are suspended. Culture shapes the style of what a society achieves. Institutions set its scale.

This argument has limits that should be stated as plainly as its conclusions. Not every emigrant needs to return; the diaspora’s value as a bridge to global knowledge is real, and a country of India’s size will always, and should always, send many of its best abroad. Resources matter as much as rules, and no reform will give Indian laboratories the budgets of the leading American universities or the largest technology companies within a decade. Culture is not irrelevant: the tightness of Indian society is real, it is sustained by families and communities as much as by laws, and some of it will outlast any reform. And the evidence on the Indian case specifically remains much thinner than the importance of the question deserves, which is itself an argument for the measurement agenda proposed above. A reader who concludes that institutions explain everything has misread the argument as badly as one who concludes that culture does.

For India’s place in artificial intelligence, the analysis has a precise implication. If there is a distinctive Indian strength, it is less likely to lie in inventing general-purpose systems from first principles than in fitting general systems to particular people, languages, institutions, and situations: in the layer where abstract intelligence is given local, usable form for a very large and very diverse population. That strength follows naturally from a context-sensitive intellectual tradition and from the linguistic and social diversity of the country. But it is a starting position, not a destiny. Better institutions could extend it toward invention as well, as they did for societies once dismissed as imitators. Worse institutions could confine it to the execution of other people’s ideas indefinitely.

The present moment makes the choice unusually consequential. The United States has made itself less certain as a destination for the talent India produces, and other countries are competing to receive it. India has, for the first time in decades, a realistic chance to keep and recover a much larger share of its most original minds. Whether it does so will depend less on the grandeur of its announcements than on the dull particulars of institutional design: whether grants arrive on time, whether failure is fast and survivable, whether a young researcher can lead without permission, whether a woman who earns a doctorate can build a career, and whether a society that has always tolerated many ideas can learn to tolerate the people who act on them.

The question with which this article began, why Indians build so boldly abroad and so cautiously at home, turns out to have a less mysterious answer than either the admirers or the critics of Indian culture suppose. Indians are not held back by how they think. They are held back, at home, by what it costs to act on what they think. That cost is set by rules, and rules are written by people. India has written better ones before, for missions it cared about. The task now is to write them for the research system as a whole, and to make home a place where its most questioning minds want to return, and are welcomed when they do.

Author’s Note: This article was conceptualized, researched, and outlined by the author, with generative AI tools utilized to assist in drafting and refining the prose.