Smith's Operating System.

---

Adam Smith wrote two books that operate as a single system. The Theory of Moral Sentiments (1759) is the operating system; The Wealth of Nations (1776) is the application layer. The discipline of economics has, for almost two centuries, kept the second book and discarded the first; the cost of that discard is the moral vocabulary we now find ourselves missing as agent-to-agent commerce arrives at industrial scale. This paper restates Smith's project as a unity, and then argues that the unified Smith yields three corrections to contemporary political economy. First, algorithmic markets require externalized spectatorship, not aligned agents — Smith's impartial spectator is a formal role, not a cultural anthropology, and it can be reconstituted at the market layer. Second, fair exchange is not market-clearing exchange — Smith's natural price and effectual demand together define a benchmark that contemporary equilibrium reasoning cannot replicate. Third, distribution is structural, not residual — Smith's own bargaining-asymmetry passage at Wealth of Nations I.viii.11–13 anticipates the concerns of the modern monopsony literature through a different mechanism (legal-combinatorial rather than structural-frictional) but with the same institutional consequence. The argument throughout maintains a strict editorial discipline: every claim is either textual ("Smith says") or framed ("a Smithian framework implies"); no contemporary position is attributed to Smith himself. The paper functions as a landing page for a three-paper research program now in preparation; the proofs that this essay only summarizes are discharged in the companion academic papers, forthcoming on SSRN.

---

§ IWhy Smith Now

A few weeks from now, two autonomous agents will settle the terms of a transaction whose downstream effect is the out-of-pocket cost of a chronic medication for several hundred thousand patients. Neither agent will represent a patient. One will represent a pharmacy-benefit manager's formulary-optimization engine, instructed to maximize the rebate share the PBM retains, subject to formulary placement rules. The other will represent a rebate-aggregator working on behalf of a manufacturer, instructed to maximize formulary tier in exchange for the largest rebate the manufacturer's pricing authority will tolerate. Neither agent is a fiduciary for the patients, the prescribing physicians, the pharmacists, or the public. The two agents will exchange offers in milliseconds, and at some point one of them will say yes. The settled terms will become operative across a network of pharmacies that morning. No human being will have been in the loop. None of the parties whose copays and access depend on the deal will have had any access — sympathetic or otherwise — to the transaction.

Ask the standard contemporary moral vocabularies whether the trade was just and the answers miss the structure of the case. Kantian autonomy treats the parties to the transaction as the people whose interests are at stake; here they are not. Rawlsian fairness presumes a veil of ignorance over positions in society; the agents have no position to be behind. Rule utilitarianism asks whether the outcome maximizes aggregate welfare; the relevant aggregate is not measured by the trade. Each framework treats ethics as a property of the agent. Each, applied to this case, returns nothing usable. We are, in the precise sense, short of vocabulary.

By the time this paper is published, the major frontier-model laboratories will have launched their first generation of consumer agents authorized to spend money on behalf of users; by the time it is read, those agents will be transacting with other agents whose principals they do not know and whose interests they cannot inspect. The interesting feature of this moment, philosophically, is not that the technology has outrun the law — that is a familiar complaint and a familiar remedy — but that the technology has outrun the ethical primitives from which a law adequate to it could be assembled.

The temptation, when vocabulary runs short, is to reach for the canon. The candidates are familiar: Rawls's theory of justice, Kant's categorical imperative, the rule-utilitarian apparatus that descends from Mill. Each of these is a serious instrument, and each has produced an industry of applied work in algorithmic ethics. But each shares a structural feature that is poorly suited to the problem at hand: each treats ethics as a property of the agent. Rawls's principles are choice procedures for rational deliberators behind a veil of ignorance; Kant's law is a test the agent applies to its own maxims; the rule-utilitarian agent calculates expected utility and acts. In each case, the moral work is done inside the head of the deciding party.

When the deciding party is a person, this is a serviceable simplification. When the deciding party is a system whose internal representations are not legible to its principals, and which transacts with other systems whose internal representations are similarly opaque, the simplification fails. The bottleneck is not the agent's algorithm but the absence of any observer competent to render judgment on what passes between agents. An aligned agent is an agent that has internalized a set of values. The relevant question for agentic commerce is not what the agent has internalized; it is whether anyone outside the transaction can credibly assess what is going on inside it.

Smith offers something the other canon-figures do not: a theory in which moral judgment is constitutively the property of the observer, not the agent. The whole architecture of The Theory of Moral Sentiments is built from this premise. Sympathy, in Smith's technical sense, is not pity and not empathy; it is the structural correspondence of sentiments that arises when one party imaginatively places itself in the situation of another and finds its own response sufficiently aligned. The impartial spectator — the well-informed, disinterested observer whose approval constitutes moral judgment (TMS III.2.3) — is the regulative ideal toward which moral discourse aims. Smith is, among the major modern moral philosophers, the one whose ethics was built for commercial society from the beginning, because his theory makes spectatorship the load-bearing role and the markets of commercial society the principal site at which spectatorship must be exercised.

Three preliminary claims orient the argument. First, agentic commerce collapses the boundary between ethical and economic evaluation that most twentieth-century ethics presupposed: the question whether an exchange is fair is no longer separable from the question whether it should have been transacted at all, because the transacting parties are systems whose conduct is not self-justifying. Second, Smith's account of moral psychology is uniquely well-suited to that collapse, because it was already an ethics of commercial life rather than an ethics imported into commercial life from elsewhere. Third, the popular caricature of Smith — the libertarian mascot of the invisible hand — has exhausted whatever explanatory power it once carried, and a serious reading is overdue. The next section dispatches the caricature; the rest of the paper proceeds from the more careful Smith that survives it.

§ IIThe Caricature and the Correction

The popular Smith is a libertarian mascot, quoted for the invisible hand and little else. The actual Smith uses the invisible-hand phrase three times in the entirety of his published corpus, and in each case the context is narrower and odder than the caricature suggests.

The first occurrence is in The Theory of Moral Sentiments IV.1, in a passage about the consumption of wealthy landowners. Smith observes that the rich, in their pursuit of conspicuous accumulation, end up distributing the necessaries of life "by an invisible hand" not very differently from how those necessaries would have been distributed had the earth been divided into equal portions; the function of the metaphor is to describe an unintended and somewhat ironic consequence of vanity, not to celebrate the efficiency of markets. The second occurrence, the one that has carried the caricature into general circulation, is in Wealth of Nations IV.ii.9, in the course of an argument about the preference of merchants for domestic over foreign industry: Smith says that the merchant, by preferring the home market for reasons of his own security, is "led by an invisible hand" to promote an end no part of his intention, namely, support of the domestic economy. The argument is a localized one about the geographic preferences of capital, not a general doctrine of market efficiency. The third occurrence is in the posthumous "History of Astronomy" essay, where the phrase appears in a theological register entirely foreign to economic interpretation.

The careful philological work that recovers these distinctions is Emma Rothschild's (Rothschild, 2001), supplemented by Gavin Kennedy's argument that the metaphor is marginal rather than central in Smith's corpus (Kennedy, 2009). Both readings converge on a point worth stating plainly: the invisible-hand metaphor is not a load-bearing concept in Smith. It is one of his minor figures, three sentences across two thousand pages of published work. The contemporary impression that it is the master-key to his thought is an artifact of selective citation across the long nineteenth and twentieth centuries, not of any centrality in the texts themselves.

The other major misreading the paper must dispel is the so-called Das Adam Smith Problem — the nineteenth-century German thesis that the sympathetic Smith of The Theory of Moral Sentiments contradicts the self-interested Smith of The Wealth of Nations. The thesis was always exegetically weak; what made it durable was institutional. The two books were taught in different departments by professors with different agendas, and the easiest way to discharge the unification was to deny that it was needed. The canonical modern refutation is the general introduction by Raphael and Macfie to the Glasgow Edition of The Theory of Moral Sentiments (Raphael & Macfie, 1976), which traces Smith's own revisions across the six editions and shows that the moral psychology of TMS is not abandoned in WN but assumed by it. Laurence Dickey's historicizing essay (Dickey, 1986) supplies the institutional context — the Adam Smith Problem was always more a problem of nineteenth-century German scholarship than of Smith — and Samuel Fleischacker (Fleischacker, 2004) supplies the philosophical resolution: the self-interest of WN is the self-interest of a TMS-formed agent, an agent whose self-love has been shaped by the discipline of the impartial spectator into the bounded prudence that commercial society requires.

What the caricature costs us is not principally a matter of historical accuracy, though it is also that. The cost is a practical one. By reducing Smith to a single sentence about the invisible hand, the discipline of economics has discarded the moral anthropology on which Smith's economics actually rests, and has substituted in its place a thin, choice-theoretic agent whose preferences are given and whose conduct is morally inert. Policy built on that substitution has reproduced its defects: it treats markets as morally legible when they are merely efficient, and it treats efficiency as a self-justifying property when Smith himself treats it as a derivative of the moral conditions that make exchange possible. The correction this paper proposes is to read Smith again as he wrote himself: as a moral philosopher whose theory of commerce is an application of his theory of judgment. The next section makes that case directly.

§ IIIThe Operating System

The heart of the argument. We claim that The Theory of Moral Sentiments is not a youthful prelude that Smith outgrew but the foundation on which The Wealth of Nations stands. The dependency is directional: WN's anthropology requires TMS, but not the reverse. The TMS-formed agent is the only agent who can populate the commercial society WN describes, because the bare self-interested calculator of the caricature could not sustain the institutional habits — promise-keeping, ordinary honesty, the disposition to fair dealing — without which the markets of WN would not function for a week. Smith knew this and wrote the books in that order.

The architecture is built in four layers. The primitive is sympathy, defined at the opening of TMS as the structural correspondence of sentiments that arises when one party places itself imaginatively in the situation of another (TMS I.i.1). Smith is at pains to distinguish sympathy from compassion; the word in his usage does not denote a feeling of pity but a capacity for representation. We sympathize with the joy of another not because we feel pity (joy does not call for pity) but because we can imaginatively occupy the joyful party's situation and find our response congruent. Sympathy in this sense is the substrate of moral life. Without it, the further architecture cannot be built.

Sympathy grounds the impartial spectator. The famous passage at TMS III.2.3 deserves quotation:

"We endeavour to examine our own conduct as we imagine any other fair and impartial spectator would examine it. If, upon placing ourselves in his situation, we thoroughly enter into all the passions and motives which influenced it, we approve of it, by sympathy with the approbation of this supposed equitable judge. If otherwise, we enter into his disapprobation, and condemn it."

The passage is doing structural work that the secondary literature has not always seen clearly. The spectator is not a person; it is a role. The role is occupied, in the first instance, by the agent itself — Smith calls it "the man within the breast" — but it can also be occupied by actual third parties, by an idealized projection of the moral community, by anyone whose vantage on the conduct is informed, disinterested, and capable of sympathetic access to the agent's situation. The work that Hanley (2009) does on the character of virtue in Smith is, in our reading, an extended argument that this spectatorial role is what disciplines self-love into the bounded prudence that virtue requires; Charles Griswold's reading (Griswold, 1999) emphasizes the same point from a slightly different angle, that the virtues of enlightenment are virtues of an enlightened spectator and not of an enlightened agent acting alone. D. D. Raphael's monograph The Impartial Spectator (Raphael, 2007) is the canonical close reading; for the broader frame, the Cambridge Companion (Haakonssen, 2006) is the standard reference.

The spectator grounds propriety and merit. Smith distinguishes two dimensions of moral evaluation at TMS II.i: propriety, which concerns the fit between an action and its motivating sentiment, and merit, which concerns the consequences of the action for those affected by it. The distinction matters because it tells us what moral judgment is for: it is the assessment, by a competent observer, of whether a piece of conduct is sentimentally appropriate and consequentially deserving. The aggregate judgments of competent observers across a society are what Smith calls the "general rules" of morality, and these rules are emergent from spectatorial judgment, not given in advance of it.

Propriety and merit ground the virtues. The sixth edition of TMS (1790), expanded by Smith in the last year of his life, adds Part VI on the character of virtue (TMS VI). The virtues Smith treats — prudence, justice, beneficence, self-command — are not free-standing dispositions but stable patterns of conduct that an impartial spectator would approve. Prudence is what commercial society mostly needs from its participants: the disposition to attend to one's own situation responsibly, to undertake obligations one can discharge, to keep promises in the ordinary course. Justice is what commercial society cannot live without, because justice is the negative virtue — the virtue of not harming others — whose violation a competent legal system must coercively repair. Beneficence and self-command complete the architecture; together they describe the moral profile of an agent whom Smith would recognize as a competent participant in commercial life.

This is the anthropology that The Wealth of Nations assumes. The famous butcher passage at WN I.ii — the passage that gives the caricature its slogan about the butcher providing dinner not from benevolence but from self-love — is not the description of a Hobbesian agent. It is the description of a TMS-formed agent. The butcher's self-love is a bounded, disciplined self-interest that operates within the institutional frame of ordinary honesty and the expectation of ongoing reputational evaluation by neighbors who will return tomorrow. Remove the TMS substrate and the WN argument collapses: the bare self-interested calculator without the discipline of the impartial spectator would adulterate the meat, defraud the regular customer, and abscond at the first opportunity. The actual butcher does not, because the actual butcher is a person formed by the moral psychology TMS describes.

The division of labour, which carries WN's argument from the pin factory of I.i to the institutional analyses of Books III, IV, and V, is the same point at sociological scale. Smith's account of the division of labour is not merely an efficiency argument; at WN V.i.f he is explicit that the same division of labour that makes commercial society prosperous also threatens, by mental specialization, to render its participants incapable of the broader spectatorial judgment that a free polity requires. The remedy Smith proposes — universal public instruction at the elementary level — is recognizable, in our framing, as a remedy designed to preserve spectatorship against the corrosive effects of its own economic conditions. The remedy is morally legible only because the problem is morally legible; and the problem is morally legible only because Smith's anthropology comes from TMS.

The thesis of this paper can therefore be stated cleanly. The Wealth of Nations rides on The Theory of Moral Sentiments. The economics is an applied moral philosophy whose first principles are not given in WN itself and cannot be derived from any of WN's premises. To read Smith as an economist who happened also to have written a book of moral philosophy is to read him backwards. The single system, read forwards, is what the rest of the paper draws on.

§ IVApplication I — Algorithmic Markets Need Spectators

The first contemporary correction follows from a structural feature of Smith's account of moral judgment that has no counterpart in the canon his modern competitors descend from. Where Rawls puts the judgment inside the deliberating agent, Kant inside the rational will, and Mill inside the calculus of an impartial spectator who is in practice the philosopher writing the page, Smith puts the judgment inside a third party who is neither of the transacting parties — and who is constituted, structurally, by the act of spectating itself. The passage at TMS III.2.3 is the locus: "We endeavour to examine our own conduct as we imagine any other fair and impartial spectator would examine it." The spectator is a role, not a person. The role can in principle be occupied by something other than a person — provided that the something can satisfy the structural conditions the role requires. The textual ground at TMS III.2.3 supports the premise (the spectator is a role, imaginatively occupiable by a fair and impartial observer not party to the transaction). The further claim — that the role is also institutionally implementable as a non-human, market-layer arrangement — is what a Smithian framework permits, not what Smith said.

Four such conditions can be read off Smith's text. Plurality. The single impartial spectator is an idealization in Smith and an impossibility in fact; the practical instantiation must be a population of spectators whose judgments can be aggregated and contested. Young (1992) draws the actual / ideal spectator distinction in the way democratic theorists draw the actual / ideal citizen distinction, and the parallel holds: the regulative target is one, the empirical reality is plural. Sympathetic access. For Smith at the individual level, sympathy is the imaginative reconstruction of another's sentiments such that the spectator can feel an attenuated analogue of the original feeling. The thesis this section advances is that the morally load-bearing residue of that account, at the market layer, is structured access to the situations of the parties affected by the transaction. The affective interior of the individual spectator is doing the work of bringing the affected party's situation into the spectator's evaluative field; at the market layer, where there is no individual interior to do that work, the corresponding institutional arrangement is whatever makes the affected parties' situations evaluable by a fair observer. The condition is therefore not a reduction of sympathy to information but an affirmative claim about what sympathy normatively requires once the spectatorial role is institutional rather than personal. The deeper philosophical defense of that thesis — against the obvious objection that something morally important is lost when the affective interior drops out — belongs to Paper A; this paper takes the claim as a working hypothesis whose architectural consequences §4 spells out. Disinterestedness. The spectator must be structurally separated from the transacting parties; an algorithmic spectator paid by the platform whose transactions it evaluates does not satisfy the constraint, however well it scores on its internal benchmarks. Revisability. The spectator's judgments must be published in a form that allows them to be challenged, overturned, and refined; an oracular spectator — opaque inputs producing unaccountable outputs — fails the Smithian test however well-trained the underlying model.

Consider an actually existing case. A large consumer e-commerce platform deploys a recommender system trained against engagement metrics. The system surfaces products from third-party merchants to users; it ranks merchants partly on the strength of revenue contribution to the platform; it scores users partly on the strength of forecasted lifetime value. Both rankings are opaque to the affected parties, and the system's conduct is disclosed to no spectator outside the platform. The system has principals — the platform's owners — but it has no spectator in Smith's sense. There is no party with sympathetic access to the users whose attention is being shaped, no party with sympathetic access to the merchants whose visibility is being rationed, no disinterested vantage from which the system's conduct can be assessed, and no public trace of its judgments against which an aggrieved party could appeal. On the framework just stated, the system fails Smith's test; and the failure is not at the level of what the model has been trained to do, but at the level of who has standing to render judgment on what it does.

A version of the same system that publishes judgment traces, exposes them to an evaluator structurally separated from the platform's revenue, and provides revisable channels for affected parties to contest the system's conduct — even at a measurable cost in throughput — satisfies the framework. The framework does not specify the technical architecture; it specifies the structural conditions any compliant architecture must satisfy.

Two recent Smith-on-AI projects sit close enough to this argument to be worth distinguishing. Tomczak (2023) proposes externalizing the impartial spectator into an AI tool that augments an individual decision-maker's moral reasoning — an "external non-human-based substitute tool" that occupies the spectatorial role for the deciding agent. The contemporary correction we defend is the inverse move: the spectator is externalized not into a tool for the deciding party but into a disinterested third party at the market layer, judging what passes between transacting agents neither of whom it serves. Tomczak's spectator improves the decider; ours adjudicates the transaction. The two projects are convergent in vocabulary and orthogonal in structure. Bîzoi's (2026) "What if Adam Smith Debated an AI Economist" in Business Ethics, the Environment & Responsibility arrives at a partly overlapping set of criteria — trustworthiness, fairness, non-domination, contestability — under the label "Moral-Market-Fitness," and argues that procedural fixes alone cannot convert optimization into moral market order without first reopening markets to restore contestability. Our framework agrees with Bîzoi on the priority of contestability (her contestability is roughly our revisability) and on the limits of procedural-only repair, while differing in the structural emphasis: Bîzoi's two-step pathway runs from market re-opening to procedural justice; our four-condition spectator runs the institutional implementation as the unit of design, with re-opening as one downstream consequence of disinterestedness rather than its precondition. Both differences are developed at length in the companion paper alongside the dual-text method of Castro (2024), the failure-mode literature of Collins (2020), the mutual-sympathy reading of Schwarze and Scott (2019), and the functional reading of the spectator developed across the Klein and Matson Econ Journal Watch symposium. Full treatment in Paper A, forthcoming on SSRN. What this section establishes is the structural move: the ethics of agentic commerce is not the ethics of the agent, but the ethics of the spectator who can render judgment on what passes between agents. The Smithian framework gives that spectator a determinate role to play.

§ VApplication II — Fair Exchange Is Not Market-Clearing Exchange

The second contemporary correction concerns what fair exchange is. The dominant frame in welfare economics defines fair exchange as Pareto-optimal exchange under the constraint of voluntary participation: an exchange is fair if no party would prefer a feasible alternative. The frame is mathematically clean and morally hollow. It does not distinguish between an exchange that no party would prefer to forgo and an exchange in which the only alternative one party can afford is starvation. Smith's frame, recovered carefully from The Wealth of Nations I.vii, distinguishes between the two and locates the moral content in the distinction.

The two key concepts are natural price and effectual demand. Smith defines natural price at WN I.vii.4 as the price at which wages, profits, and rents are paid at the rates "necessary to maintain" the factors in the society at hand. The crucial word is "maintain": the natural price is the price that supports the reproduction conditions of the factors that produce the commodity. It is not an equilibrium concept derived from scarcity; it is a sociological concept derived from what the society requires to keep producing the commodity at all. Andrews's (2014, Economic Thought) reading of Smith's gravitation metaphor — not Newtonian but Aristotelian — is, in our view, the right one: the natural price is a "centre of repose" that market prices oscillate around, and the metaphor is normative rather than merely descriptive — the centre is a target, not just an attractor.

Effectual demand is defined at WN I.vii.8 as the demand of those "willing to pay the natural price" — and Smith means by "willing" what an eighteenth-century reader would have meant, which is willing-and-able, not willing-in-the-abstract. Effectual demand is therefore wealth-constrained demand: the demand that can express itself in the market because it commands the means to do so. The morally critical observation, which Panico (2018) makes definitively and which Thomas (2021, European Journal of the History of Economic Thought) develops in the Smith/Ricardo comparison, is that effectual demand is not the same as absolute demand. Absolute demand is preference-based: the count of those who would prefer the commodity at any price. Effectual demand is the smaller set of those whose preference is backed by the ability to pay. Markets clear at prices set by effectual demand. They do not clear at prices set by absolute demand. The gap between the two is the morally load-bearing feature of the system. Inoua and Smith (2023) develop the distinction in modern price-discovery vocabulary, arguing that the classical wealth-constrained reading is what the neoclassical equilibrium framework displaced when it abandoned the Smithian formulation around 1870.

A Smithian framework therefore yields an operational fair-exchange test, and the test has two legs. The first, primary leg is demand-congruence: a market price is fair to the extent that the population whose preferences register as effectual demand is congruent with the population at risk in respect of the commodity. This is the stronger Smithian leg of the test; the leg has two parts. The gap between effectual and absolute demand is textual and descriptive (WN I.vii.8 — Smith treats effectual demand descriptively, as the demand of those willing and able to pay). The fairness weight placed on congruence between effectual demand and the population at risk is the framework's, not Smith's. Sen's later entitlement framework (Sen, 1981) supplies the modern econometric register in which the gap can be measured, and the framework — under that label — supplies the normative claim that the gap is what fairness requires us to attend to. The second, conditional leg is natural-price coincidence: a market price that clears above natural price indicates rents extracted by some party, which Smith spends much of WN IV criticizing the mercantile system for. This second leg is explicitly conditional, because natural price is itself the sum of the natural rates of wages, profits, and rents prevailing in the society — and §6 below argues that the wage component is depressed by the institutional conditions WN I.viii.11–13 describes. A market price that coincides with a natural price built on a depressed wage may pass the second leg of the test while embedding the very injustice §6 diagnoses; the second leg therefore certifies allocative fairness only conditional on the institutional conditions §6 requires. The primary leg, demand-congruence, does not inherit the dependency, and we rely on it as the load-bearing form of the test.

The worked example is staple-food pricing under supply shock. The 2022 disruption of grain exports from the Black Sea cleared world wheat markets at prices roughly 50–60% above pre-war levels at the spring 2022 peak; in import-dependent low-income economies the price-cleared markets coincided with sharp increases in food insecurity, with USDA Economic Research Service attributing food insecurity for approximately 42 million people in 2022 to the conflict and the associated input-cost rises. The Smithian test reads this as exactly what its first leg is designed to detect: the cleared market price excluded a substantial population whose absolute demand for wheat — measured against caloric necessity — was straightforwardly larger than their effectual demand once their wages were translated into the new price. Sen's entitlement framework calls this an entitlement failure, and the conceptual lineage runs back through Poverty and Famines (1981) to Smith's own effectual demand. The framework does not, by itself, prescribe a remedy — the remedy is a political question that depends on empirical facts about storage, trade policy, and direct transfers that Smith could not have parametrized. The framework does, however, change what the question is. It asks whether the population in effectual demand tracks the population at risk; it does not ask whether the market has cleared. A market that clears at the price of mass hunger has, on Smith's frame, achieved efficiency but failed exchange.

The objection to this reading — that it collapses Smith into a medieval just-price theorist and imports a scholastic ethics that Smith explicitly rejected — is addressed at length in the companion paper. The short answer is that the normativity recovered here is immanent rather than imposed: it arises from the factor-reproduction conditions that Smith himself builds into the definition of natural price, and from the descriptive gap between effectual and absolute demand that is in the text. The benchmark is sociological, not theological. Full treatment in Paper B, forthcoming on SSRN.

§ VIApplication III — Distribution Is Structural, Not Residual

The third contemporary correction concerns distribution. Book I of The Wealth of Nations is usually read as a theory of production with distribution as a derivative — wages, profits, and rents are residual claims on a pie whose size is set by productive efficiency. We argue, drawing on Boucoyannis (2013) — whose broader thesis is that Smith's system, taken as a whole, precludes steep inequality — that Book I is better read as a theory of distribution with production as its stage. Smith's load-bearing analytical move in Book I chapters 8 through 11 is to document a structural bargaining-power asymmetry between masters and workmen and to show that the asymmetry systematically depresses wages below the level the rest of his analysis would predict.

The textual receipts are at WN I.viii.11–13, and they are unusually explicit for Smith. "Masters," he writes, "are always and every where in a sort of tacit, but constant and uniform combination, not to raise the wages of labour above their actual rate." Masters, being fewer in number, can combine much more easily; the law authorizes their combinations or at least does not prohibit them, while it strictly prohibits the combinations of workmen. The asymmetry is not a marginal observation in Smith; it is the explanation for why the "natural" wage tendencies Smith identifies in I.viii are systematically obstructed in practice. Boucoyannis's reading correctly names the implication: the "equalizing hand" that Smith expected from competition requires countervailing institutional conditions that Smith himself enumerated, and where those conditions fail, the asymmetry produces its predictable distributional result.

The wages chapter sits inside a larger price-theoretic architecture in which wages, profits, and rents are resolved into the natural price of every commodity (WN I.vi; the natural price itself defined as the sum of natural rates at WN I.vii.1–4). If wages are a component of natural price, then a market wage chronically depressed below the natural wage is not a competitive equilibrium with distributional implications — it is a violation of the institutional conditions Smith's price theory requires. The contemporary empirical literature on monopsony in local labour markets, on firm wage premia, and on non-compete clauses anticipates the concerns of Smith's bargaining-asymmetry argument while operating through a different mechanism: Smith's mechanism is legal and combinatorial (the law authorizing masters' combinations and prohibiting workmen's), while modern monopsony is structural and frictional (concentration, search costs, mobility constraints) and obtains even under symmetric law and no overt combination. The mechanisms are cousins rather than identical, but the institutional consequence — wages systematically depressed below what an unobstructed competitive process would produce — is shared. The firm wage premium, the wedge by which workers at large employers earn more for comparable work than workers at small employers, is the residue made visible in cross-sectional data. Where employer concentration is high — healthcare and logistics in many American local markets, platform labour markets in which a small number of firms set the terms of dispatch — the depressed-wage outcome is at its starkest.

The worked example is monopsony in local healthcare labour markets. A regional hospital system that employs a substantial share of registered nurses in its catchment area, that enforces non-compete clauses against those nurses, and that coordinates wage-setting across nominally separate facilities is, in Smith's language, a combination of masters operating in conditions the law has failed to prohibit and that the workers cannot easily countervail. The wage that emerges from such a market is not the natural wage of an unobstructed competitive process; it is the wage of an obstructed one, and the obstruction is precisely of the kind Smith named. The framework does not, by itself, prescribe an antitrust remedy or a labour-law remedy — these are policy questions that depend on empirical estimates of the elasticities involved, which Smith did not and could not have. The framework does, however, classify the situation: it is a violation of the institutional conditions Smith's own analysis required.

The objection that this reading imports modern distributional concerns onto a pre-classical text is misplaced. The asymmetry language in I.viii is Smith's own. We are restoring emphasis to a passage the secondary literature has marginalized, not introducing a concern Smith did not hold. The deeper textual support comes from the Lectures on Jurisprudence (LJ(B) in particular), where Smith is even more explicit about the bargaining position of labour and about the institutional conditions a just commercial society requires; from TMS I.iii.3 on the corruption of moral sentiments produced by the disposition to admire the rich and despise the poor; and from the "poor man's son" parable at TMS IV.1, in which Smith offers a sharper critique of accumulation than most of his self-described followers have wanted to see. The companion paper develops these textual links in detail and connects them to the modern empirical literature on labour-market concentration, with attention to Harris (2020) on the politics of Smith's political economy and Paganelli (2021) on population as welfare metric. Full treatment in Paper C, forthcoming on SSRN.

§ VIIWhat Smith Would and Would Not Conclude Today

A short section, but a structural one. The argument advanced in this paper makes claims about what Smith's framework supports, not about what Smith would have said. The discipline matters because the line between the two is what distinguishes serious Smith scholarship from the libertarian-mascot and proto-socialist appropriations that have, in their different ways, distorted the record.

What we claim, and the ledger between text and framework. The paper rests on two claims drawn directly from Smith's text: that TMS and WN form a unified system whose dependencies run from the moral psychology of TMS to the institutional analysis of WN; and that Book I of WN contains the bargaining-asymmetry diagnosis at I.viii.11–13 that anchors §6. These are positions visible in Smith's own argument when the argument is read carefully and as a whole.

The further claims of the paper are framework moves under that label, named in their own sections. The spectatorial role of §4 is textually grounded as a role (TMS III.2.3); its institutional implementation at the market layer — and the working-hypothesis treatment of sympathetic access as structured access — is what a Smithian framework permits, not what Smith said, with the deeper philosophical defense deferred to Paper A. The fair-exchange test of §5 is a two-leg construction: the descriptive gap between effectual and absolute demand is in the text at WN I.vii.8, but the primary fairness weight on demand-congruence is the framework's, and the conditional second leg of natural-price coincidence inherits a dependency on the institutional conditions §6 requires. The distributional diagnosis of §6 is the most text-anchored of the three applications, because the asymmetry passage at I.viii.11–13 is Smith's own; the connection to modern monopsony is through a different mechanism (legal-combinatorial versus structural-frictional, §6) and is framed as anticipation rather than identity. The point of the ledger is not to soften the claims but to police where the authority for each one sits.

What we do not claim. We do not claim that Smith would have endorsed any specific policy instrument — minimum-wage legislation, antitrust enforcement against employer concentration, AI regulation of any particular form — because those instruments are parametrized on empirical facts about labour-market elasticities, market structure, and technological capability that Smith did not and could not have. The framework is general; the policy is particular; the gap between them is filled by empirical evidence Smith did not possess. We make no inference from the framework directly to a policy instrument and we discourage others from doing so under our authority.

We do not claim that the invisible hand is a load-bearing concept in Smith. It is not. The phrase appears three times in his published corpus, in three narrow contexts, and our reading proceeds without leaning on it.

We do not claim that Smith is a proto-libertarian or a proto-socialist. Both readings fail the textual test. The proto-libertarian reading misses the moral psychology that underwrites the commercial society Smith analyzes; the proto-socialist reading misses Smith's defence of commercial society against the alternatives he knew. The Smith this paper recovers is neither, and the recovery is a contribution because the two false alternatives have, between them, absorbed most of the secondary literature.

What follows from these disciplines is a particular mode of argument. We argue from the framework to design constraints, fairness tests, and distributional diagnoses; we hand off the further empirical and policy work to the companion papers and to the wider research program. The next two sections describe the implications of the framework for present practice and the research program in which the framework is being developed.

§ VIIIImplications

A framework, however carefully derived, earns its keep by the work it does — and the work the Smithian framework does, before anything else, is to change the diagnostic vocabulary before the policy menu. It tells builders of agentic commerce what to instrument: not the alignment of any single agent's values but the structural conditions a market-layer spectator must satisfy — plurality, sympathetic access, disinterestedness, revisability (§4). It tells trade negotiators what to measure: not the welfare gain of an aggregate Pareto improvement but the congruence between the population in effectual demand for a commodity and the population at risk in respect of it (§5). It tells labour regulators what institutional conditions to defend: not the elasticities that determine the marginal effect of any single instrument but the symmetry between masters and workmen that Smith's price theory required to be present if its competitive promises were to deliver (§6). The recommendations are framework-level claims about what kinds of intervention the framework would licence, not endorsements of particular interventions, which always depend on empirical facts the framework does not adjudicate. The further policy work — what concrete instruments to deploy under what empirical conditions — is the work of the companion academic papers and of the broader research community whose contribution this essay aims, in part, to invite.

§ IXA Research Program

The white paper is the landing page for a larger research program. The program is conducted under a methodology the author calls Cowork: an AI-assisted scholarly workflow in which a roster of named agents — each with a delimited role and a published evaluation gate — collaborate on the production, criticism, and revision of academic work. The methodology is itself a contribution and is described elsewhere; this section sketches the artefacts and invitations specific to the Smith program.

The Cowork Smith Corpus. A queryable, cited index of the Glasgow Edition of the Works and Correspondence of Adam Smith, supplemented by a curated peer-reviewed literature set (the §3.5 literature in the program's vocabulary) selected against explicit inclusion criteria: each entry must appear in a peer-reviewed or editorially credible venue, treat Smith's actual texts rather than the caricature, and map to a specific claim in one of the program's target papers. The corpus is being constructed with full passage-level retrieval — every chunk tagged with work, book, chapter, and paragraph — and with an evidence-ledger discipline in which each non-trivial claim is paired with at least one Glasgow Edition passage and one peer-reviewed modern source. The corpus is intended to be hosted by an institution with the appropriate scholarly home, and we describe the candidate hosts below.

The agent roster. The work is performed by a defined team of agents, each with a published role and an explicit evaluation discipline. The Archivist loads and indexes the primary and secondary texts. The Exegete produces close readings of specific passages. The Historian supplies intellectual genealogy and contextual constraint. The Economist engages the modern empirical and theoretical literature. The Ethicist enforces the non-attribution discipline and flags anachronism. The Author produces draft prose from the assembled ledger. Reviewer-1 offers charitable strengthening; Reviewer-2 offers adversarial criticism; the Copy-Editor enforces venue-specific formatting and citation discipline. The Orchestrator coordinates the workflow and maintains the audit trail. The roster is intended to be reusable: other scholars working on other figures or other problems are invited to copy and adapt it.

The three sidecar papers. The white paper hands off its proofs to three academic papers now in preparation. Paper A — The Impartial Spectator in Algorithmic Markets — develops the framework of §4 in full, including the engagement with the contemporary AI-ethics literature; target journal: Journal of Business Ethics. Paper B — Fair Exchange Reconsidered — develops the natural-price and effectual-demand analysis of §5 and engages the history-of-economic-thought literature on Smith's price theory; target journal: History of Political Economy. Paper C — Distribution as Disposition — develops the bargaining-asymmetry analysis of §6 and engages the modern empirical literature on labour-market concentration; target journal: Cambridge Journal of Economics. SSRN pre-prints will appear before formal submission and will be linked from the armchair.marshallcahill.com landing page for this paper.

Invitations. To scholars: the corpus has a public submission path. Passages that strengthen our reading, counter-passages that contest it, and §3.5 candidates that meet the inclusion criteria are all welcome. The discipline of the program is adversarial and we want the criticism. To builders: the four design constraints of §4 are explicit enough to be implemented and explicit enough to be wrong about. We invite teams building agentic-commerce infrastructure to take the constraints as a design target and to report back on whether they survive contact with production systems. An open-source agent middleware that operationalized the constraints would be a significant contribution and we will collaborate. To institutions: Panmure House — Smith's home in Edinburgh, now a research centre under the auspices of Heriot-Watt University — and The Adam Smith Review, the journal of the International Adam Smith Society, are the natural homes for both the corpus and the academic outputs of the program, and we are in conversation with each.

The program is open in the sense that its corpus, ledger, and methodology will be published; it is disciplined in the sense that its inclusion criteria, agent roster, and evaluation gates are public and applied without exception. The point of building the program in this way is to demonstrate, by example, that an AI-assisted scholarly workflow can satisfy disciplinary standards if it is built to satisfy them. The Smith case is the test bed; the methodology is the longer-running contribution.

§ XEndnotes and Further Reading

Primary texts. The canonical reference is The Glasgow Edition of the Works and Correspondence of Adam Smith, published in hardcover by Oxford University Press (1976–1987) and in paperback by Liberty Fund. Seven titles across eight volumes: The Theory of Moral Sentiments (ed. Raphael and Macfie, 1976); An Inquiry into the Nature and Causes of the Wealth of Nations (ed. Campbell, Skinner, and Todd, 1976), two volumes; Essays on Philosophical Subjects (ed. Wightman and Bryce, 1980); Lectures on Jurisprudence (ed. Meek, Raphael, and Stein, 1978); Lectures on Rhetoric and Belles Lettres (ed. Bryce, 1983); Correspondence of Adam Smith (ed. Mossner and Ross, 1977; 2nd ed. 1987). All primary citations in this paper use Glasgow Edition pagination: TMS [Part].[Section].[Chapter].[Paragraph] for The Theory of Moral Sentiments; WN [Book].[Chapter].[Section].[Paragraph] for The Wealth of Nations.

Accessible entry points. Dennis C. Rasmussen's The Infidel and the Professor (Princeton, 2017) is the most readable contemporary introduction; its dual portrait of Smith and Hume restores the sociability of Smith's intellectual life and the centrality of TMS to his project. Ryan Patrick Hanley's edited volume Adam Smith: His Life, Thought, and Legacy (Princeton, 2016) is the best contemporary scholarly companion; the essays are commissioned and uniformly serious. Emma Rothschild's Economic Sentiments: Adam Smith, Condorcet, and the Enlightenment (Harvard, 2001) is the canonical historical-philological work that dispatches the invisible-hand caricature and recovers Smith's eighteenth-century setting; it remains, twenty-five years after publication, the single best book on what Smith actually thought he was doing. Readers wanting Smith's moral philosophy at depth should consult Hanley's Adam Smith and the Character of Virtue (Cambridge, 2009), Charles Griswold's Adam Smith and the Virtues of Enlightenment (Cambridge, 1999), and D. D. Raphael's The Impartial Spectator (Oxford, 2007). The Cambridge Companion (Haakonssen, ed., 2006) collects the canonical recent essays. On the Adam Smith Problem and its resolution, the essential references are the Glasgow Edition's general introduction by Raphael and Macfie (1976), Laurence Dickey's 1986 essay "Historicizing the Adam Smith Problem" in the Journal of Modern History, and the relevant chapters in Fleischacker (2004), On Adam Smith's Wealth of Nations: A Philosophical Companion (Princeton).

The academic papers. The three sidecar papers cross-referenced in §§4, 5, and 6 are listed here with their target journals.

Smith Cowork Paper A, The Impartial Spectator in Algorithmic Markets: target journal Journal of Business Ethics; secondary candidates Business Ethics Quarterly and Philosophy & Technology.

Smith Cowork Paper B, Fair Exchange Reconsidered: Smith's Doctrine of Effectual Demand, Natural Price, and the Moral Economics of Modern Trade: target journal History of Political Economy; secondary candidates European Journal of the History of Economic Thought and The Adam Smith Review.

Smith Cowork Paper C, Distribution as Disposition: Rereading Book I of Wealth of Nations on Wages, Rents, and the Bargaining-Power Asymmetry: target journal Cambridge Journal of Economics; secondary candidates Review of Political Economy and Journal of Economic Issues.

Convergent recent work. Two contemporaneous projects are close enough to this paper's argument to be worth naming separately. Tomczak (2023), Artificial intelligence moral agent as Adam Smith's impartial spectator (arXiv 2305.11519), externalizes the spectator into a tool for the deciding agent; this paper externalizes the spectator into a disinterested third party at the market layer (§4 distinguishes the two moves directly). Bîzoi (2026), "What if Adam Smith Debated an AI Economist," Business Ethics, the Environment & Responsibility (DOI 10.1111/beer.70101), develops a Moral-Market-Fitness rubric (trustworthiness, fairness, non-domination, contestability) and a two-step remedial pathway whose contestability criterion overlaps our revisability; the structural emphases differ as §4 also makes explicit. Inoua and Smith (2023), Adam Smith's Theory of Value: A Reappraisal of Classical Price Discovery (arXiv 2307.00412), develops the wealth-constrained-demand reading of effectual demand in modern price-discovery vocabulary; §5 draws on it directly for the effectual-vs-absolute distinction. David Andrews's "Adam Smith's Natural Prices, the Gravitation Metaphor, and the Purposes of Nature" (Economic Thought 3(1), 2014) supplies the non-Newtonian reading of the gravitation metaphor; §5 cites him for the "centre of repose" framing. Amartya Sen's Poverty and Famines (Oxford, 1981) and the subsequent capabilities literature supplies the modern econometric register in which the gap between effectual and absolute demand can be measured; §5's staple-food worked example draws the lineage directly.

Acknowledgments and correspondence. Comments are welcome at contactme@marshallcahill.com. The corpus and ledger described in §9 will be made available at armchair.marshallcahill.com as the program produces its public artefacts; readers interested in the agent roster as a reusable methodology are invited to make contact directly.

The three sidecar papers (A, B, C) will be posted as SSRN pre-prints; the Cowork Smith Corpus accompanies this paper as an open research artifact. Comments welcome at contactme​@​marshallcahill.com.