8: Creating Dharmas

We have always created our dharmas and told ourselves we found them. What becomes possible, and what becomes dangerous, once we admit it and start building on purpose?

A woman sits on a cushion in a small flat in a large city. A handbook of Stoic sayings lies on the floor beside her, a meditation app glows on her phone, and she is trying to assemble, out of those scraps and a few others, something she can actually live by. We met her in Part 7. She is doing the oldest human work there is, alone, mostly without help.

Elsewhere, a team at an AI company sits down to write what it openly calls a constitution. Not a constitution for a country: a list of values for a machine, principles governing how an artificial intelligence answers when a stranger asks it how to live. The company is Anthropic; the method, Constitutional AI, is what matters here. Rather than grade the model’s answers one by one with thousands of human raters, Anthropic trains it to critique and revise its own responses against an explicit, plain-language list of principles, so its values sit in a document anyone can read, argue with, and rewrite. In October 2023, with the nonprofit Collective Intelligence Project, it put that last verb to the test. In an experiment called Collective Constitutional AI, around a thousand American adults proposed and voted on principles through Polis, an open-source deliberation platform: 1,127 statements, 38,252 votes. Anthropic then trained one model on the public constitution, another on its in-house version, and compared the results. The two constitutions overlapped about half the time, with the public leaning harder toward objectivity and accessibility; the publicly trained model matched the in-house version in capability and scored lower on bias across all nine dimensions of a standard benchmark. A thousand American adults are not humanity, and the whose-values question never fully closes. Still, ordinary people, asked directly, set out to write a dharma on purpose, and the thing they wrote held its own.

Between those two pictures sits the whole question. The seeker on the cushion and the team at their screens are doing the same thing at wildly different scales: constructing, by hand and on purpose, an answer to the question of how to live. We have been heading here from the start: from the oldest root, that which holds, through the teak forests of Java, to the moment the old foundations gave way. For all of history, the question a person faced was which of the inherited dharmas was true, and how to submit to it? The seeker on the cushion cannot ask that anymore. Her question, and now ours, is one earlier ages rarely asked so nakedly: not which dharma do I receive, but can we build a better one, knowing exactly what we are doing, on purpose?

The Thing We Were Always Doing

Building dharmas on purpose is not some unprecedented act of hubris. It is what our species has always done. Every dharma we have met was made: assembled by particular people, in particular places, out of the materials to hand. The Axial sages made theirs. Samin Surosentiko made his, in plain low Javanese, one conversation at a time. What is new is not the making. What is new is that we can no longer pretend we are doing anything else.

For almost all of history, the work was hidden from the workers. People built their dharmas and then experienced them as found: handed down from a mountaintop, woven into the order of the cosmos. That concealment was not a lie so much as a feature; a rule feels far more binding when it looks like the grain of reality than when it looks like last year’s committee decision. The secular age has stripped that concealment away. We can see the workshop now. We can see that the values we live by were authored, which means that, with a lurch, they could have been authored differently and could be authored again.

So the honest question is not whether to make dharmas. We will make them whether we admit it or not; the only choice is whether to do it asleep or awake. And the conscious construction falls due just as two things of unprecedented scale bear down on us: a planet in ecological crisis that may need some shared ethic to survive at all, and machines that may soon think, into which someone is already pouring an ethic. The biologist EO Wilson, in a 2009 conversation at Harvard, put our predicament in one devastating line: we have “Paleolithic emotions, medieval institutions, and god-like technology.” Conscious dharma-building is the attempt to grow up the emotions and the institutions before the technology finishes the sentence.

But the moment it is said out loud that dharmas can be built, the terrible follow-up questions arrive in a rush: built by whom, for whom, toward what end, and what gets crushed in the building? Those questions name a single recurring danger, and it will stalk every frontier ahead. Call it the flattening: the temptation, once the power to author an ethic exists, to author just oneand to crush the glorious plurality of human dharmas into a single official version. What remains is four tests of whether we can build without it.

A Dharma for the Whole Earth?

The largest scale first, where the pressure to build a single shared ethic is most intense.

The Samin repay one more visit. At the centre of their hand-made dharma sat a flat refusal: the forest, the water, the land are not anyone’s to own or sell, but a commons held by everyone and no one. Stretch that conviction over the entire planet and it starts sounding like the only sane description of our situation. The atmosphere is a commons. The oceans are a commons. The stable climate that every economy quietly assumes is a commons, and we are burning through it as a village burns through a shared forest when no one’s dharma tells them to stop.

For the first time, the pressure comes with an instrument panel. Earth-system scientists led by Johan Rockström have mapped nine “planetary boundaries,” the limits within which humanity can operate safely, and the most recent health check finds seven already crossed. The latest to fall was ocean acidification, the very commons the Samin’s conviction was just stretched over. That is an is, not an ought: the science can tell us where the edges are, but it cannot, by itself, generate the obligation to stay inside them.

Our situation exposes a cruelty specific to planetary problems. Human cooperation runs on a tribal engine: we bond fiercely in here in part by defining a them out there. That engine built the cooperation every dharma runs on. But climate change is the rare collective problem that excludes nobody. There is no out-group, no enemy tribe whose threat could rally the species, because the species is the unit that has to cooperate. The hard conclusion follows from the evolutionary logic Athena Aktipis draws out: humanity has never had the shared history that would allow cooperation to evolve at the level of the entire species. A planetary dharma cannot be waited for; there is no competing planet to select it into being. It has to be built consciously, or it will not exist at all.

What should we build? Not, I am convinced, a single thick global ethic that everyone must adopt. That way lies the flattening in civilisational form. The answer the evidence keeps pointing to is a thin shared layer over a thick plurality: a small set of commitments narrow enough to be affirmed from within many different traditions. The political philosopher John Rawls called this an “overlapping consensus”: agreement reached from within very different deep worldviews, no one surrendering theirs. Kwame Anthony Appiah calls the spirit of it “partial cosmopolitanism,” universal concern for every human being held in tension with genuine respect for legitimate difference.

And it has already happened in the open. The Earth Charter, drafted around 2000, is a thoroughly secular declaration of planetary responsibility whose authority rests on the inclusiveness of the process that wrote it; fifteen years later Pope Francis’s encyclical Laudato Si’ insisted that “intergenerational solidarity is not optional, but rather a basic question of justice.” Two frameworks that agree on almost nothing about the ultimate nature of reality reach for nearly the same planetary commitments. That convergence, not a single imposed creed, is the shape a planetary dharma can actually take. The political scientist Elinor Ostrom, who spent a career showing how real communities govern shared resources without privatisation or a distant boss, prescribed the same shape for the global commons: in place of one world rule, “polycentric” governance, many local experiments coordinated toward a shared good. She was honest about her toolkit’s limits: her design principles, she warned, do not scale easily to a global commons, where there is no room for major error because the experiment runs on a single, fragile planet. Polycentricity was her hedge, not her triumph.

And right beside this hopeful picture stands the warning. There already exists a thick, singular, state-authored planetary ethic, and it is instructive because it works. China’s doctrine of “ecological civilization” was written into the constitution by a 2018 amendment, and the programme it crowns has real results to show, including one of the largest reforestations in history. It is also, in the judgement of Mette Halskov Hansen, Hongtao Li and Rune Svarverud, who traced its history, a model of authoritarian environmentalism: an ethic driven from the top by technocratic elites, enforced on people who do not get a say. The singular planetary dharma, in other words, is not science fiction; it can be effective; and it is the flattening itself, at work. The same evolutionary logic supplies the brake: a body does many things to regulate its own cells that no one would call ethical done to a person. We can study the body to learn how to scale cooperation. We must not govern a planet the way an immune system culls defectors. A planetary dharma worth having coordinates dharmas. It does not dissolve them into one.

The Ethics We Hand to Machines

From the planet, the focus narrows to the machine, where the same drama plays out faster, and the temptation of the single answer comes wearing the friendly mask of a technical solution.

Why does artificial intelligence force a dharmic question at all? The reason runs deeper than the technology. In the philosopher Nick Bostrom’s terms, it is the “orthogonality thesis”: intelligence and goals are independent axes. A system can become arbitrarily more capable without becoming one inch wiser or kinder, because no amount of raw capability tells it what to value. This is the oldest wall in moral philosophy, the one between is and ought that has run beneath everything here, rebuilt in silicon. You cannot compute your way from a fact to a value. Goodness has to be put in by hand, which means somebody has to decide which goodness.

And here the flattening arrives, disguised. The standard slogan of the field is “align AI to human values,” and that phrase smuggles in the singular. There is no the human values to align to, any more than there is a single the Dharma. The philosopher Iason Gabriel names the real problem: not simply what values AI should align with, but whose, chosen by what fair process. That is a political and dharmic question, not an engineering one, and pretending otherwise is how a single company’s ethics, or a single state’s, could quietly become the encoded conscience of the species.

The honest answers being developed run against the single version, and they look like the planetary solution scaled down. The computer scientist Stuart Russell’s is the most disarming: the safest machine, he argues, is one built to be uncertain about what humans value, a machine that defers, keeps asking, and allows itself to be switched off. Anthropic’s public-constitution experiment is another: a way to elicit and reconcile many value systems rather than impose one. A third comes from Oliver Klingefjord, Ryan Lowe and Joe Edelman, whose “moral graph” defines wisdom as knowing “which values apply in which contexts” rather than as having the right answers. And, fittingly for a stance that refuses to rank traditions, Thomas Doctor, Michael Levin and colleagues propose “care as the driver of intelligence,” a design principle drawn from the Bodhisattva vow, the ancient promise to seek awakening for the sake of all sentient beings. That plurality of approaches looks to me like a safeguard, not a defect.

But machines do not only need an ethic. They threaten to dismantle one we already live inside. For generations, the West has run on a dharma so pervasive that most people never notice it is a dharma at all: the conviction that your claim to a place in the world is finding something to do with your time that other people will pay you for. Work has been our quiet creed of merit, status and belonging. Artificial intelligence threatens it in a new way: where the old machines came first for muscle, this one comes first for the educated professions. The economists Carl Frey and Michael Osborne saw it coming a decade ago, and the grim joke is that it comes for the celebrated technologist before it comes for your electrician. If AI severs work from survival, that dharma loses the ground it stood on, and a new ethic, with a new economics under it, will have to be built where it stood.

The most discussed candidate for the economics is a universal basic income, a regular payment to everyone that breaks the link between earning and surviving. The idea is older than the machines: Thomas Paine argued in 1797 for a citizen’s dividend as compensation for the enclosed common inheritance, which makes basic income less a novelty than the Samin’s commons paid out in money. The oldest objection, that free money breeds idleness, has been tested and largely failed: Finland’s national experiment found employment unaffected and well-being improved, Stockton’s SEED pilot found cash moved people into full-time work, and GiveDirectly’s Kenyan trials point in the same direction. But the pilots expose a deeper problem, dharmic rather than economic. Money can replace a wage. It cannot replace what the wage was secretly carrying: the sense of contributing, of mattering to others. Contributive justice is Michael Sandel’s name for this: the human need not merely to consume but to be needed. And David Graeber devoted a whole book to the misery of people trapped in jobs they know to be pointless; the wound, then, was never only about the scarcity of work. Sam Harris, pressed on exactly this, concedes the point even while waving it off: we will find status and meaning in other ways, and our need for purposeful effort is just one of the “stories we tell ourselves.” He is more right than he means to be. A story we tell ourselves about why our lives matter is one of the better short definitions of a dharma we have met. AI does not abolish our need for one. It amputates the work-shaped scaffolding that has carried that need, and dares us to build a replacement.

Beside all this stands the sharpest version of the flattening danger yet: what the philosopher William MacAskill calls “value lock-in.” Whatever ethic happens to get encoded into the systems that run everything could freeze, foreclosing future moral progress. Pour one dharma into the machines at the wrong moment, with enough power behind it, and the plural is not merely flattened for now. It may be flattened for keeps.

Building from Old Stone

Step back from planets and machines to a quieter question that underlies both: out of what materials are we supposed to build these new dharmas? We build, as humans always have, from what we inherit, and the honest name for a dharma built deliberately from inherited materials is post-traditional.

“Post-traditional” is easy to misread as “anti-traditional,” as though the project were to bulldoze the old wisdom and start clean. The better gloss is post-conventional: a stance reached by absorbing a tradition deeply and then stepping past its unquestioned authority, not by skipping it. The former monk Stephen Batchelor gives the move its cleanest statement. He separates two things we lump together as “religious”: ultimate concern (a phrase he borrows from the theologian Paul Tillich), the wish to come honestly to terms with our own living and dying; and the institutional apparatus (texts, priesthood, supernatural claims) by which a tradition enacts that concern. A post-traditional dharma keeps the first and loosens the second. A person can be gripped by ultimate concern, Batchelor notes, without any supernatural belief at all, which is why “devout atheists” are, in his phrase, “not entirely joking.”

A deeper irony takes the romance out of the question. We picture the task as the careful fusion of two pristine, separate things, ancient wisdom and modern science. But the wisdom traditions actually available to us were already remade by modernity. The Buddhist-studies scholar David McMahan has shown how the “Buddhism” that reached the modern West was reshaped, long ago, to privilege reason and personal experience over ritual and authority, to align itself with the prestige of science, and to be retold in the language of Western psychology. The fusion we imagine ourselves performing happened before we were born. “Post-traditional” is just the honest name for a process underway for over a century, in what the philosopher Charles Taylor called the modern “nova” of meaning.

What does a well-built post-traditional dharma actually keep? A few features recur. It is fallibilist: it holds that its claims are revisable rather than fixed revelation, its one honest answer to the grounding problem. It is naturalist without being scientistic: it can find genuine reverence in the natural world, as the biologist Ursula Goodenough does, without pretending science can hand us our values. The Dalai Lama’s minimal version is hard to improve on: “our own brain, our own heart is our temple; the philosophy is kindness.” It prizes practice over doctrine, keeping the meditation, the ritual, and the ethics that change a person, while loosening its grip on metaphysics. It stays reflexive, aware of its own failure modes, because anything this powerful can curdle. And it counts its losses honestly: the secular substitutes so far, the Sunday Assemblies and humanist associations, have not come close to matching the pull of the communities they set out to replace, while loneliness exacts a health toll researchers rank alongside smoking and obesity. On its hardest metric, community, the movement is losing, and a well-built dharma has to say so.

That reflexiveness matters most because building from old stone has two opposite ways of betraying the plural. The shallow betrayal is the one the sociologist Robert Bellah named decades ago in a woman who described her private faith as “Sheilaism,” following her “own little voice,” a personal blend of bits of this and that with no community and no one to answer to. That is the spiritual supermarket: maximal personalisation, minimal substance, McMindfulness wearing a different outfit. And before we smile at Sheila: the woman on the cushion from Part 7 is one lonely year away from her; the difference has nothing to do with sincerity. It is whether the scraps ever find a community and someone to answer to. The deep betrayal is subtler and more seductive, because it looks like the height of respect: the claim that all traditions are really surface expressions of one underlying truth, so we may as well boil them down to their shared essence. This is perennialism, and the philosopher Steven Katz showed why it is both false and flattening. There are no raw, unmediated mystical experiences underneath the traditions; each tradition constructs the experiences it produces, so a Christian’s union with God and a Buddhist’s realisation of emptiness are, on Katz’s account, not two meetings with one reality but two different experiences all the way down. Honouring traditions means honouring them as genuinely, stubbornly different, not melting them into one ecumenical paste.

And one last warning hangs over this frontier, aimed straight at those of us who feel safest here, on the secular side: a secular dharma is not automatically more rational or less dangerous than a religious one. David Sloan Wilson has a name for this: a “stealth religion,” any belief system, godless ones included, that sacrifices honesty about trade-offs for the comfort of a tidy story, and his prime exhibit is not a church but Ayn Rand’s militantly atheist Objectivism. Secular does not mean undogmatic. Building from old stone, we can lay the new bricks just as crooked as the old ones, and congratulate ourselves the whole time on having no religion.

The Mirror

One more dharma is under construction that I have not yet named honestly, and it is this one. This series is not a neutral survey. It is the second stage of a project whose stated goal is to gather traceably sourced evidence about dharma so that future artificial intelligences can be given, in the project’s own words, “a solid grounding in human ethics and values, and how these are formed.” The project is itself an instance of the thing under discussion: an attempt to deliberately build a dharma and pour it into machines. It would be dishonest to warn about the flattening at this length and exempt the warning’s author. So the project belongs under its own microscope.

The flattening, it turns out, is not a remote risk for an encoded ethic. It is the default, and three mechanisms drive it there. In 2021, the Allen Institute for AI built a system called Delphi, distilling 1.7 million human moral judgements into a single model that would rule on right and wrong. It managed the easy cases, then broke in revealing ways; the deepest criticism, pressed by Zeerak Talat and colleagues, was that it had quietly served up descriptions of what people happen to approve as prescriptions about what is good, marching straight off the is/ought cliff. A second mechanism is colder and more mechanical. When an AI is trained on data that includes the output of earlier AIs, Ilia Shumailov and colleagues showed in Nature in 2024, the model can suffer “collapse,” and the first thing to vanish is the tails of the distribution: the minority dharmas disappear first, no ill intent required. The cure was to keep the original human data, plural and messy, in the mix. Provenance is not just an ethic. It is a survival trait.

And that points to the third and most fundamental mechanism, the one that makes this danger intrinsic rather than incidental. A machine like this does not store definitions of dharma. It learns meaning by statistics, from the company each word keeps, as the linguist JR Firth meant when he said: “you shall know a word by the company it keeps.” It then stores that meaning as a position in a vast space, fixed entirely by its relations to everything else, which is Ferdinand de Saussure’s old insight (that in language “there are only differences”) rebuilt in arithmetic. A word’s encoded meaning is essentially the average of its uses, so a rare or dissenting sense, a minority dharma, is by construction a faint outlier, forever tugged toward the majority meaning. The machine does not argue against an unusual dharma. It does something quieter and more total: it means it away. The flattening is not a bug to be patched with good intentions. It is the grain of the medium.

An older charge waits here too, one we met in Part 6, and here it applies literally. When a Western, secular, English-language corpus catalogues the world’s dharmas and prepares to pour them into machines, it repeats, at scale, the move Edward Said anatomised over a career: assembling knowledge of other people’s ways of living, from one side of an old imbalance of power, in containers the catalogued never chose. The research notes behind this series press that critique by name, and they are right to. Good intentions do not answer it. The only answer I have found: keep the provenance, keep the dissent, keep the many voices audibly many, and say plainly whose shelf the collection sits on. Whether that genuinely escapes the trap or merely acknowledges it cannot be judged from within. The history is not reassuring: Western interpreters have reliably found in the Dharma whatever they brought, Kantians finding Kant, positivists finding Russell; so it must be asked whether we found “holds without crushing” in the traditions, or brought it along.

So can the project’s own goal even be coherent? Only, I think, if it gives up the usual sense of “distil.” The right thing to hand a future machine is not a resolved ethic, a single clean answer boiled down from the mess, but a preserved one: the plurality, the sources, the disagreements, and above all the reasoning of human dharmas, kept intact. That is what the method was built for: every claim in the research notes carries its source, every note flags its own biases, and the gathering of evidence was deliberately separated from the writing, so the tensions would be preserved rather than smoothed away. The best thing a corpus like this can teach a future mind is not what to believe but how human beings have come to believe in, fought over, and live their many dharmas.

One self-criticism cannot be engineered away, and honesty requires ending on it. The stance running through all of this, that all dharmas sit on the same shelf and none is privileged in advance, is not a view from nowhere. The phrase is the philosopher Thomas Nagel’s: there is no neutral standpoint above the human situation from which to rank everything else. “Treat all dharmas as equal” is itself a substantive, recognisably modern, secular and liberal commitment, a thick dharma of its own. The project cannot climb out of its own skin; it cannot encode human ethics from no point of view, because it has one. What it can do, and what I have tried to do, is name that point of view and refuse to install it as if it were simply the truth. Jorge Luis Borges imagined a map so perfect it was the same size as the territory, and therefore useless. Any compression of human ethics small enough to be useful is a choice about what to leave out, and every such choice encodes a dharma. The only sin is to make that choice while pretending it is not being made.

Plural, but Not Anything Goes

Which brings us, at last, to the question Part 1 raised and every part since has sharpened. If dharmas are genuinely plural and none is privileged in advance, then are some of them better than others, or is it all just a matter of preference, with nothing to choose between kindness and cruelty but taste? It deserves a straight answer: no, it is not all preference.

Plurality does not mean relativism. The whole answer turns on that distinction. Isaiah Berlin insisted throughout his life that value pluralism is neither relativism nor absolutism. Plural values are real, grounded in a shared human nature, not arbitrary inventions; there are simply many of them, they genuinely conflict, and no single system can maximise them all at once. “Anything goes” is a different and much lazier claim, and it collapses under the first test: it cannot condemn an atrocity, and it cannot make sense of reform, because the person who stands up to say their own society’s practice is wrong (the abolitionist, the suffragist) needs a standard above current custom to appeal to, which is what “anything goes” denies them.

So what is the standard? Not a single thick code that dictates how everyone must live. The structure the evidence keeps converging on has appeared at every scale: a thin floor, with thick plurality flourishing above it. The floor is a small set of near-universal constraints: limits on cruelty and violence within the group and the protection of basic human capabilities. Nor is it wishful thinking: when Oliver Scott Curry tested seven rules of cooperation against the ethnographic record of sixty societies, every one was held as morally good wherever it appeared; morality, on that evidence, is variation on a theme, exactly the shape a thin floor under thick plurality predicts. The philosopher Martha Nussbaum gave the floor real force with her list of central human capabilities, fundamental goods any defensible way of life must let people actually achieve, honoured in a thousand cultural forms: life, bodily integrity, practical reason, relationship, play. The economist Amartya Sen, who built the capabilities approach alongside her, deliberately refused to fix any such list and kept the standard open; the floor’s exact placement is unfinished business. Above the floor, real diversity stops being a problem to solve and becomes a richness to protect. Below it lies not a different valid dharma but a failed one.

The whole standard extends the phrase we began with. A dharma is to be judged by how well it holds without crushing. The holding is the floor’s positive side: does it actually gather a human life and a human community, give them shape, meaning and the strength to endure? The not-crushing is the floor’s protective side: does it gather without breaking the people inside it or the strangers outside it? A dharma that holds but crushes is the tyranny-and-purity spiral of Part 6. A dharma that refuses to crush but cannot hold is the thin gruel of Part 7, the patchwork that never quite binds. The good ones do both. They hold firmly enough to give a life its spine, and loosely enough to leave room for other lives, and other dharmas, to do the same. That is the answer to the question of whether some are better. Yes. The better ones hold without crushing, stay honest enough to correct themselves, and leave the door open for their neighbours. Berlin would add a caution: the standard identifies the failures, but it does not rank the successes because, above the floor, the goods genuinely conflict. Everything up there is still to argue about, and WB Gallie explained why the arguing never stops: some concepts (justice, democracy, and surely dharma) are essentially contested, disputed from the inside as part of how they work. That endless argument is not the failure of the project. It is the project.

That Which Holds

We have come the whole way round. We started with a word we thought meant incense and inner peace, and we followed it down to a plain and sturdy root: dharma, that which holds, the beam that keeps the roof up, the thing that keeps a life, or a world, from chaos. Dharmas are what hold human beings together. There are many of them. They are made, not found. And the making has now become conscious, and urgent, and ours.

For all of history we built the things that hold us and then knelt before them as if we had received them. That age is ending, not because we grew wiser but because the concealment wore through. We can see the workshop now, and we are being handed the tools at a demanding moment: a planet that needs some shared and gentle holding to come through its crisis intact, and a new kind of mind we are about to build, into which we will pour some ethic or other, on purpose or by neglect. The temptation, now of all moments, will be to reach for the single answer, the one official dharma, the clean machine-readable verdict that ends all the messy argument. That temptation is the one thing the whole argument stands against. The plurality is not the problem to be solved on the way to the answer. The plurality is the answer, fenced by a thin floor that forbids only cruelty and demands only that the holding not become a crushing.

Part 1 quietly promised this ending. None of us gets to opt out of having a dharma; each of us is running one right now, chosen or not, and nowhere along the way have we met a human life that holds together without one. The only real choice is whether we run them asleep or awake. To practise a dharma awake is to know that it is made and not found, to know that it is one good way among other good ways, to hold it firmly enough that it actually holds us, and lightly enough that it never has to crush anyone else to do its work. The woman on the cushion, alone with her scraps, has been doing that work all along; what she lacks is not sincerity but company. We have always been choosing what holds us. Now we get to do it on purpose, with our eyes open. That is not a smaller task than the old religions set themselves. And it is, at last, openly ours.

« 7: Dharmas in a Secular Age

Sources & further reading

This essay is built from the project’s research notes, which carry the full inline citations. Listed here: the notes it draws on, and the key works behind them.

Research notes

Research notes used in this essay series are published on GitHub.

• 8.1 Planetary Dharma — the no-out-group problem; thin-over-thick; authored not evolved; polycentric governance; the totalising temptation.
• 8.2 AI & Dharma — orthogonality; “whose values”; pluralistic alignment; economic disempowerment, UBI and the meaning it cannot buy; value lock-in.
• 8.3 Post-Traditional Dharmas — post-conventional, not anti-traditional; the constructive features; the two flattening failure modes; “stealth religion”; the community deficit.
• 8.4 Encoding Dharma into Machines — the project under its own microscope; the flattening as default; encode the map and the argument, not the verdict.
• 3.4 Relativism vs. Universalism — pluralism is not relativism; the thin floor, thick variation; “holds without crushing.”
• 7.4 Measurement & Falsifiability — the is/ought wall; provenance is not truth; the project’s own evidentiary bar.
• 7.5 Orientalism — Said’s knowledge-power critique and its limits, turned on this project’s own corpus-and-machine ambition; the projection-screen history behind the Mirror.

Key works

• Y. Bai et al., “Constitutional AI: Harmlessness from AI Feedback” (2022, arXiv:2212.08073) — the machine’s values as an explicit, readable, rewritable document.
• Saffron Huang, Divya Siddarth et al., “Collective Constitutional AI: Aligning a Language Model with Public Input” (ACM FAccT 2024, DOI 10.1145/3630106.3658979) — the public-constitution experiment of the opening: ~1,000 Americans, Polis, 1,127 statements, 38,252 votes; the publicly trained model matched on capability with lower bias.
• Johan Rockström, Katherine Richardson et al., the planetary-boundaries framework (2009; Planetary Health Check, 2025) — seven of nine boundaries transgressed; ocean acidification the latest to fall.
• Elinor Ostrom, Governing the Commons (1990), and DS Wilson, E. Ostrom & ME Cox, “Generalizing the core design principles” (JEBO, 2013) — polycentric self-governance against Garrett Hardin’s “Tragedy of the Commons” (1968); her own warning that the principles do not easily scale to a single fragile planet.
• Mette Halskov Hansen, Hongtao Li & Rune Svarverud, “Ecological civilization: Interpreting the Chinese past, projecting the global future” (Global Environmental Change, 2018) — authoritarian environmentalism; with the 2018 constitutional amendment.
• Kwame Anthony Appiah, Cosmopolitanism (2006), and John Rawls, Political Liberalism (1993) — partial cosmopolitanism; the overlapping consensus.
• The Earth Charter (2000) and Pope Francis, Laudato Si’ (2015) — secular and religious frameworks converging on planetary commitments.
• EO Wilson, remarks in debate at the Harvard Museum of Natural History, 9 September 2009 — “Paleolithic emotions, medieval institutions, and god-like technology.”
• Nick Bostrom, Superintelligence (2014) — the orthogonality thesis; William MacAskill, What We Owe the Future (2022) — value lock-in.
• Stuart Russell, Human Compatible (2019) — machines built to be uncertain about what humans value.
• Iason Gabriel, “Artificial Intelligence, Values, and Alignment” (2020) — whose values, by what fair process.
• Oliver Klingefjord, Ryan Lowe & Joe Edelman, “What Are Human Values, and How Do We Align AI to Them?” (2024) — the moral graph; wisdom as knowing “which values apply in which contexts.”
• T. Doctor, M. Levin et al., “Biology, Buddhism, and AI: Care as the Driver of Intelligence” (2022) — the Bodhisattva vow as a design principle.
• Carl Frey & Michael Osborne, “The Future of Employment” (2013), and Daniel Susskind, A World Without Work (2020) — this wave of machines coming first for the educated professions; the argument that this time the automation may be structural.
• Thomas Paine, Agrarian Justice (1797) — the citizen’s dividend as compensation for the enclosed common inheritance.
• Michael Sandel, The Tyranny of Merit (2020), and David Graeber, Bullshit Jobs (2018) — contributive justice; the dignity work was carrying.
• Kela, final report of Finland’s basic-income experiment (2020); the Stockton SEED pilot (2021); GiveDirectly’s Kenya trials — the idleness objection, tested and largely failed.
• Sam Harris, interviewed by Steven Bartlett on The Diary of a CEO (2023) — the UBI case (“break this connection between work and survival”) and the “stories we tell ourselves” reply to the purpose objection.
• Stephen Batchelor, After Buddhism (2015), and David McMahan, The Making of Buddhist Modernism (2008) — ultimate concern (a phrase from Paul Tillich) vs. institutional form; tradition already remade by modernity.
• Robert Bellah et al., Habits of the Heart (1985) — “Sheilaism”; with Steven Katz on the constructed, plural nature of mystical experience against perennialism.
• David Sloan Wilson, “The New Atheism as a Stealth Religion” — secular belief systems can be dogmatic too; prosocial.world, “The World Needs a Secular Community Revolution” — the community deficit.
• L. Jiang et al., “Delphi” (2021; Nature Machine Intelligence, 2025), with the critique pressed by Zeerak Talat and colleagues — descriptions served as prescriptions; I. Shumailov et al., “AI models collapse when trained on recursively generated data,” Nature (2024) — the tails vanish first; keeping human data in the mix is the cure.
• JR Firth (1957) and Ferdinand de Saussure (1916) — meaning as the company a word keeps, and as “differences without positive terms”: why an encoding averages the plural away.
• Thomas Nagel, The View from Nowhere (1986), and Jorge Luis Borges, “On Exactitude in Science” (1946) — no neutral standpoint; the map that becomes the territory.
• Isaiah Berlin on value pluralism, and Martha Nussbaum & Amartya Sen, the capabilities approach — pluralism as the alternative to both relativism and absolutism; the floor with teeth, its exact placement left open.
• Oliver Scott Curry, “Is It Good to Cooperate?” (Current Anthropology, 2019) — seven moral rules across sixty societies; morality as variation on a theme.
• WB Gallie, “Essentially Contested Concepts” (1956) — why the argument over dharma never closes.