Merit and Meritocracy in the Age of AI
Merit is the quality of being good or worthy. Gyume Khensur (Lobsang Tenzin) Rinpoche, in a 2006 teaching, describes merit as, generally, “everything that is virtuous”. Virtue meaning behavior that shows high moral standards.1 That is where my grip on the word first slips, because moral standards vary from person to person.
I cannot claim to understand what “accumulating merit” means inside Buddhist philosophy. But filtered through my own lens, from what I heard from monks, it is roughly this: you accumulate merit when you do things that are good for yourself while generating positive externalities for others, consciously or not. If you exist as yourself and act in line with your values, merit follows. Teaching sits high in this scheme and I think, so does showing up with an authentic desire to learn.
Meritocracy sounds like it ought to mean the government of the meritorious: the good, the virtuous. It doesn’t. The Oxford definition is “government or the holding of power by people selected on the basis of their ability.” Somewhere between merit and meritocracy, virtue quietly became ability. The word was coined in 1958 by Michael Young, and coined as a warning, not a compliment.2
I didn’t know that when I first met the idea. In an undergraduate economics class I was assigned Daniel Markovits’s The Meritocracy Trap, and it upended a belief I had held without ever examining it: that people who failed simply hadn’t worked hard enough. Markovits argues that elite advantage is no longer transmitted mainly through inherited property but through education, a top preschool, a private secondary school, an elite university, graduate school which converts into an elite job. He calculates, something on the order of $10 million in lifetime earnings. Education has become the way the advantaged pass advantage to their children, and it pays better than any inheritance.
The cruelest part is what this does to how we explain failure. Meritocracy, Markovits writes, “frames disadvantage in terms of individual defects of skill and effort.” If the system is fair, then not getting the job, not making the rent, must be your fault. He calls the result a politics of humiliation.
Everything I have said so far is American: Markovits’s world and, until recently, the furniture of my own. I have been reading Sienna Craig’s The Ends of Kinship, which follows people from Mustang, in the Nepali Himalaya, to New York. This is a route I have some stake in, since I was born in New York and my parents in the Himalaya.3 I expected a book about a faraway place. I kept finding the same machine.
Take credentials. In Mustang, a farmer tells Craig that education is good, but “if they don’t have ideas, capacity, and motivation but they have a [certificate], BA, Master, then this means nothing” (Craig 2020, 79). A subsistence farmer in the Himalaya and a Yale law professor, arriving at the same complaint from opposite ends of the earth. What the farmer is naming is the oldest question in the economics of education: whether schooling builds capacity (human capital) or merely sorts and certifies it, a signal.4 His worry is that the credential has floated free of the capacity it was supposed to stand for. Or take the way the system quietly reproduces itself. Craig visits a village that has built its own school, and the man running it explains who actually attends: “people with money and power still send their kids away. We take care of the ones with less. So, it is same same, different different” (Craig 2020, 86). The wealthy buy their children out to become educated: Kathmandu, to Pokhara, eventually to New York; everyone else takes what is local. It is the Ivy-League-versus-public-school split at twelve thousand feet, and no one had to import it.
That is what convinced me the trap is structural, not national. It appears wherever we decide that ability, measured by schooling, is what makes a person deserving.
Which brings me to the thing that made me want to write this down. Every version of the story so far: Markovits’s, Craig’s, mine, rests on one assumption: that cognitive ability is scarce, and that credentials are how we ration and reward it. Artificial intelligence puts that assumption under pressure. A machine can now produce, on demand and at almost no cost, a great deal of the credentialed cognitive output that an elite education used to be the most reliable source of. If that holds (and it is a conjecture about the labor market, and a contested one) then the scarcity rents on a certain kind of thinking begin to erode, and Markovits’s $10 million was, underneath, a bet on that scarcity. So what happens to a hierarchy built on rationing knowledge when knowledge becomes cheap?
The farmer in Mustang, it turns out, was ahead of all of us. A degree “means nothing,” he said, without ideas, capacity, and motivation. For as long as the degree reliably signaled those things, we could pretend the distinction didn’t matter. AI makes it hard to keep pretending, because it mostly cheapens the signal, not the capacity underneath. When the credentialed output is cheap, the scarce thing left is the part the farmer named and the part Rinpoche named. Judgment. Care. The authentic desire to learn. Virtue, if you like.
It would be tidy to stop there, with ability automated and virtue restored to the throne. I don’t believe it for a second. Nothing about cheap machine cognition makes the world start rewarding care over credentials; it is at least as likely to hand a sharper lever to whoever already holds the advantage.
For someone with a surface understanding of Buddhism, it is easy to look at all of this and think: that is their karma. The people born into hard circumstances who overcome every obstacle… didn’t they simply work hard enough? And if you were born with resources and information that someone just as capable was denied, there must be some reason you deserve more… because if there isn’t, do you deserve what you have at all? Do you deserve your place in the hierarchy?
The deepest version of the question is not about schooling at all; it runs before anyone has done anything to earn or fail. Angus Deaton opens The Great Escape with the brute arithmetic of the birth lottery: whether you arrive in a place with clean water and a clinic, or with neither, is settled for you before your first breath, and it shapes your life more than almost anything you will later choose.5 And the lottery is not only material; it is informational. Craig sits with a new mother who wipes away her colostrum, the first milk, because she has been told the yellow liquid is impure and that she should wait for the white milk to come, when in fact it is among the most protective things a newborn can receive. “We live the knowledge we are given,” Craig writes (Craig 2020, 55). Some infants get antibodies in their first hour of life; others have them wiped away. None of them earned the difference.
I don’t think meritocracy ever really answered that question; it just made the question feel rude to ask. What it did was borrow the moral authority of merit (you deserve this) while doing the opposite work, sorting people by an ability that was mostly inherited to begin with. Intergenerational mobility is sticky at both ends: the children of the rich tend to stay rich, and the children of the poor, poor. The more unequal a society, the stickier it tends to be.6 AI could make that stickiness worse, or it could loosen it. That is the whole question, and it is not yet decided.
This is not idle speculation; it is roughly where a generation of randomized trials has already landed. Education technology, when it is tested carefully, tends to produce modest average gains with enormous dispersion around them: largest where the tool is adaptive and paired with a capable teacher, smallest, and sometimes negative, where it is hardware is underutilized in a school.7 In my own work evaluating a computer-lab rollout across schools in Solukhumbu, the same program raised learning in some schools and lowered it in others; the average was near zero, but the average was never the interesting part. The labs helped where teachers could already fold them into their teaching and hurt where they could not: uniform hardware, non-uniform capacity, divergent returns. The gap between advantaged and disadvantaged schools did not close. If anything, it widened.
This is the desert question answered with data instead of philosophy. The same program, meeting the same effort, went opposite ways depending on where a child happened to be born. Not karma. Not grit. Starting conditions, wearing the mask of merit. A well-meaning intervention can deepen the divide as easily as close it.
This is not an abstract question for me. It is the intervention at the center of my dissertation: DeepGyan AI, built with Himalaya AI Labs. Gyan (ज्ञान) is knowledge; I like that deep carries both depth and दीप, the small lamp you light. It is a teacher-facing, offline-first, Nepali-language tool that runs on a phone the teacher already owns: it lets them find where each young child actually is with spoken English (a diagnosis otherwise too costly to produce child by child) and hands back curriculum-aligned activities to close the specific gaps it finds. The theory of change is deliberately modest. Not a device that has to arrive at the school, but a tool that runs on one already in the teacher’s pocket; something that can be dropped into a place and used with little scaffolding, and that keeps developing in use: improved by, and generating research for, the communities using it, built with people from those places rather than delivered to them. The point of studying it, rather than simply building it, is that I do not get to assume the answer. The question the dissertation actually asks is whether a phone-delivered tool like this works as a leveler (best precisely where capacity and infrastructure are scarcest) or as a gap-widener that, like the many hardware interventions before it, needs those very inputs to work at all. Which one it turns out to be is an empirical question, and my wager is that the outcome is a design choice, not a law of nature. In the frame I began with, testing that wager honestly is itself a kind of merit-making: work that is good for me while, I hope, throwing off positive externalities for the people who were told, wrongly, that they simply hadn’t worked hard enough.
This essay was written in collaboration with an AI assistant (Claude): I brought the argument, the fieldwork, and the sources; it helped me structure and tighten the prose. The argument, the research, and any errors are my own.
Footnotes
https://fpmt.org/mandala/archives/mandala-issues-for-2007/august/ask-a-lama-what-is-merit/↩︎
https://blogs.lse.ac.uk/lsereviewofbooks/2020/03/30/book-review-the-meritocracy-trap-by-daniel-markovits/↩︎
Sienna R. Craig, The Ends of Kinship: Connecting Himalayan Lives Between Nepal and New York (University of Washington Press, 2020).↩︎
Michael Spence, “Job Market Signaling,” Quarterly Journal of Economics 87, no. 3 (1973): 355–74. For the strong-form version, Bryan Caplan, The Case Against Education (Princeton University Press, 2018).↩︎
Angus Deaton, The Great Escape: Health, Wealth, and the Origins of Inequality (Princeton University Press, 2013).↩︎
On persistence and the “Great Gatsby curve,” see Miles Corak, “Income Inequality, Equality of Opportunity, and Intergenerational Mobility,” Journal of Economic Perspectives 27, no. 3 (2013): 79–102; and Raj Chetty, Nathaniel Hendren, Patrick Kline, and Emmanuel Saez, “Where Is the Land of Opportunity? The Geography of Intergenerational Mobility in the United States,” Quarterly Journal of Economics 129, no. 4 (2014): 1553–1623.↩︎
Maya Escueta, André Joshua Nickow, Philip Oreopoulos, and Vincent Quan, “Upgrading Education with Technology: Insights from Experimental Research,” Journal of Economic Literature 58, no. 4 (2020): 897–996; Karthik Muralidharan, Abhijeet Singh, and Alejandro J. Ganimian, “Disrupting Education? Experimental Evidence on Technology-Aided Instruction in India,” American Economic Review 109, no. 4 (2019): 1426–60; and J-PAL, “Leveraging Mobile Phones for Learning,” Abdul Latif Jameel Poverty Action Lab, 2026.↩︎