What arrangement of society counts as just? Liberty, equality, rights, and the common good pull against one another — these four books are the sharpest answers in a two-hundred-year argument, and they are quarreling with each other.
2026 · Book Recommendations · No. 52
Everyone invokes "justice," yet it falls apart the moment you press on it. These four books aren't four isolated opinions — they're a chain of mutual rebuttal. Mill first draws the line society may not cross around the individual (the harm principle); Rawls rebuilds liberal justice with a "veil of ignorance," moving the weight onto fair distribution; Nozick counterattacks from the right — justice lies not in the outcome but in the process of acquisition; Sandel then attacks both from the communitarian side — justice can never stay neutral about "what counts as a good life." You come away not with four positions to memorize, but with a set of coordinates for locating any real-world argument.
| Book | Author | Year | The one thing it makes clear |
|---|---|---|---|
| On Liberty | John Stuart Mill | 1859 | The only reason society may rightfully coerce an individual is to prevent harm to others — over the part that concerns only himself, his sovereignty is absolute |
| A Theory of Justice | John Rawls | 1971 | To ask which rules are fair, imagine you don't know who in society you'll turn out to be — from that, two principles of justice follow |
| Anarchy, State, and Utopia | Robert Nozick | 1974 | Justice looks at whether acquisition was legitimate, not at the final pattern; maintaining any ideal pattern requires ceaseless interference with liberty |
| Justice | Michael J. Sandel | 2009 | Justice can't sidestep "what a good life is" — the self that freely chooses everything, unencumbered by any identity, is a fiction |
Mill is after a threat subtler than any tyrant: the tyranny of the majority. When power passes from a king to "the people," oppression doesn't vanish — it only changes form, into opinion, custom, and social ostracism, pressing everyone into the same mold. He wants a line around individual liberty that even the majority may not cross.
That line is the book's mechanism — the harm principle: the sole end for which a civilized society may rightfully use force against a member's will is to prevent harm to others. "For his own good" is not a legitimate reason. Conduct splits in two: the other-regarding part society may govern; over the self-regarding part (belief, taste, ways of life, choices that harm only oneself), individual sovereignty is absolute — others may counsel, never coerce.
From this comes the most powerful defense of free speech ever written, by a counterintuitive route: to silence an opinion, whether it is right or wrong, is to rob the whole human race. If the silenced opinion is right, we lose the chance to correct our error; if it is wrong, we lose something more precious — truth kept alive by collision with falsehood. A truth never challenged decays into a "dead dogma" that no one truly understands and everyone merely recites. Hence "he who knows only his own side of the case knows little of that."
The third piece is individuality: diverse "experiments in living" are the wellspring of social progress. A person is not a machine built to a blueprint but a tree that must grow along its own nature. What Mill fears is a gentle suffocation — not violence but everyone converging by choice, treating conformity as a virtue until they no longer dare even to imagine another way to live.
The line between self- and other-regarding is nearly impossible to draw in practice — almost every act spills over; addiction, self-ruin, and neglecting one's own education all touch family and society. Mill also confines the harm principle to "civilized" societies, granting exceptions for peoples he deems "backward" — a clear mark of his era and its colonial cast.
Mill's two mechanisms land squarely on raising a school-age child. Use the harm principle to zone discipline: which matters are the child's "self-regarding sovereignty" — taste, interests, harmless mistakes, testing things at her own pace — where a parent may counsel but not coerce; and which are truly "other-regarding" (hurting someone, damaging long-term safety), where authority is warranted. To try next week: pick one thing you reflexively decide for her but which really affects only her (what to wear, which homework first, how to spend a free half-day), and hand it back. The other piece: don't let your values become a "dead dogma" in her head — deliberately let her hear and restate views you disagree with, so her convictions grow roots through collision rather than recitation.
Rawls wants an answer to "what is justice" that rests neither on intuition nor on power. His stroke of genius is to turn the question into a choice procedure: imagine people about to set the basic rules of society, but all placed behind a "veil of ignorance" — none knows whether he'll be rich or poor, brilliant or ordinary, of what race, sex, or religion. The veil blocks exactly the information you'd use to favor yourself.
Here is the mechanism: ignorance removes bias. Because you might turn out to be anyone, the rational choice is to think from the worst-off position — what if, when the veil lifts, you are the one at the bottom? From this follow two principles, in strict order: first, each person has an equal set of basic liberties (speech, conscience, political participation), which may not be traded away for welfare; second, social and economic inequalities are permitted only if opportunities are open to all on fair terms and they also make the worst-off better off. The second half is the famous difference principle.
The difference principle's edge is that it reconceives "desert." Your talents, birth, even your disposition to work hard, are largely the outcome of a natural lottery — morally arbitrary. Why should winning that lottery entitle you to keep all the returns? Rawls doesn't argue for leveling but for this: the gifted may gain more, on condition that their gains also lift those at the bottom. Talent becomes a kind of common asset, and inequality is treated as a lever to raise everyone rather than a warrant for privilege.
This book almost single-handedly revived political philosophy in the English-speaking world, and forced every discussion since onto the same table — you either argue along with it or, like the next two authors, rise up expressly to oppose it.
Whether people behind the veil would really choose the difference principle is contested — why must they be extremely risk-averse and reason from the worst case, rather than gamble for a higher expected outcome? The theory is highly abstract, compressing messy real politics into a rational-choice puzzle; critics add that it leans too heavily on Western liberal-democratic intuitions to be universal.
The veil of ignorance is a scalpel for institutional design, aimed right at anyone leading a technical team. When you set a new rule — how AI tools get allocated, how performance is scored, who owns the work being displaced by AI — run the Rawls check: if you didn't know which person on the team you'd be once the veil lifts (the one who best commands AI, or the one whose role is being eroded and is most likely to be "optimized away"), would you still set this rule? To try next week: take a division-of-labor or evaluation scheme you're drafting and ask, line by line, "is this fair to the person at the bottom?" Any clause that feels fair only if you happen to be the winner is bias that slipped in. The difference principle gives AI's dividends a hard test: does the new efficiency gain also lift the weakest link, or does it only make the strong stronger?
Nozick was Rawls's colleague and neighbor at the same university, and wrote this book to rebut him directly. It opens with an uncompromising claim of individual rights: individuals have rights, and there are things no person or group may do to them — the individual is an end, not a means to be sacrificed for "aggregate welfare" or an "ideal distribution."
From this he moves justice entirely from "outcome" to "process," calling it the entitlement theory: whether a holding is just depends not on the pattern the distribution shows but on its history — whether it came about through just acquisition (rightfully taking something unowned) and just transfer (voluntary exchange or gift). Justly acquired + justly transferred = justly held. Distributive justice is not a design to be assembled but the natural sediment of countless legitimate transactions.
His signature argument — the Wilt Chamberlain example — is the sharpest. Start from any distribution you deem ideal (call it D1). A basketball star arranges, outside his contract, for each person who freely buys a ticket to drop him an extra twenty-five cents. Over a season a million people freely do so, and he ends up far richer than anyone else; the distribution is now D2. If D1 was just, and people freely spent money they legitimately owned, on what grounds is D2 unjust? Hence the most damaging conclusion: no "end-state" or patterned principle of justice can be maintained without continuously interfering with people's lives — because free exchange keeps breaking the pattern, and to preserve it you must keep confiscating, correcting, redistributing. Liberty and pattern cannot both be had.
Following this logic, the only legitimate state is the "minimal state": functions narrowed to protection against force, theft, fraud, and the enforcement of contracts. Any state more extensive — above all one that taxes for redistribution — violates individual rights. He draws a jarring corollary: taxation for the sake of redistribution is, in kind, on a par with forced labor.
The whole chain hangs on the starting point of "just acquisition," yet real-world original property is almost never free of conquest, plunder, and slavery — if the origin is unclean, every later "just transfer" is passing along stolen goods, and Nozick offers only a vague rectification clause for this. The theory is nearly indifferent to vast existing inequality, and struggles to answer "is starving to death in a free market also justice?"
Nozick's "look at process, not pattern" is the right ruler for investment and data-ethics judgment in the AI era. Looking at an AI company, don't fixate on its impressive capabilities and tidy financial pattern; ask the Nozickian question: how was its core asset "acquired"? A large model's power mostly rests on human creative work scraped without authorization — precisely the stain on "original acquisition" Nozick names, a ticking legitimacy risk. To try next week: run an "acquisition legitimacy" due diligence on an AI holding or watchlist name — where the training data came from, whether labelers were fairly paid, the consent chain on user data. The prettier the pattern and the more dubious the process, the larger the bill buried in the valuation, one that eventually comes due. Conversely, it's a reminder to walk your own products' "just transfer" chain clean from the start.
Sandel first sorts all theories of justice into three approaches: welfare (which distribution maximizes total welfare), freedom (which respects each person's choices and rights — both Rawls and Nozick belong here), and virtue (which cultivates the good and makes good citizens). The first two are the modern mainstream, and liberalism in particular holds that the state should stay neutral about "what a good life is," building only a fair framework in which people pursue their own.
Sandel's central blow targets the hidden premise of that "neutrality" — it presupposes an "unencumbered self": a pure subject that exists prior to all ends and identities and can pick its values like goods off a shelf. That self, Sandel says, is a fiction. Real people are always constituted by identity — whose child you are, which nation and tradition you belong to, which unchosen obligations you carry (debts to family, homeland, history). These aren't things a blank self selects into itself afterward; they constitute who you are.
This overturns the possibility of neutrality: many disputes over justice (surrogacy, selling organs, conscription, memorials and apologies) simply can't be settled without judging "what is worth prizing." To ask whether something should be bought and sold is really to ask whether treating it that way degrades its meaning. Hence "justice is not only about the right way to distribute things; it is also about the right way to value things."
The conclusion therefore points to a harder, braver road: a just society cannot preserve neutrality by dodging controversy, only by having citizens reason together, in public, about "the meaning of the good life." This also makes the book the best entry point of the four — it hands you no standard answer but forces you, inside concrete dilemmas, to see which approach you actually stand on.
Sandel demolishes elegantly but builds with difficulty: once politics is set to pursue "the common good," who defines it? It slides easily toward the majority imposing its values on everyone — exactly what Mill most feared. Written as course lectures, the book is highly readable, but for that reason it stops short at key points, letting depth give way to accessibility.
Sandel's "unencumbered self" pokes right at the self-image behind "the AI-empowered super-individual." Chasing that ideal, it's easy to picture yourself as an agent who can shed every attachment and scale infinitely on tools alone — the very fiction Sandel names. Your real edge comes precisely from the "constitutive" things: whose parent you are, which unfashionable fields you've gone deep in (consciousness, Buddhist thought, complexity science), which specific technical history you carry. AI amplifies capability; it can't amplify "who you are" — and that is the moat no one else can copy. To try next week: list the three things you most want to amplify with AI, and ask of each — is it amplifying a replaceable, generic capability, or a constitutive advantage that belongs only to this life of yours? Tilt resources toward the latter. Same for parenting: don't only teach a child "free choice" — pass on, with weight, the specific tradition and attachments she inherits; that isn't a shackle, it's the base material of her future self.
Run a substitution test: replace "this is for your own good" with "your doing this will harm whom, and cause what concrete, nameable damage." If you can name a real third-party harm → it's other-regarding, and intervention is legitimate; if all you can name is "he'll lose out / he'll regret it" → it's self-regarding, and you may counsel at most; coercion crosses Mill's line. Most "for your own good" fails this swap.
These two grievances call for opposite prescriptions. If it's an outcome problem (the bottom is too miserable) → redistribute and raise the floor. If it's a process problem (the money came by monopoly, fraud, or plunder) → redistribution treats the symptom; go after the legitimacy of acquisition instead. Many arguments never resolve because one side is talking about the pattern and the other about the process, each past the other. First be clear which side you're on.
The touchstone: if your defense of an arrangement is "it's his own choice, I don't judge," press one layer deeper — are you able to withhold judgment only because you've already assumed "this thing can be freely bought, sold, or disposed of without loss of meaning"? Much "neutrality" is a hidden value judgment described as no judgment at all. Sandel's assignment isn't to force you to take sides — it's to force the judgment you made on the sly out into the open, and own it.