PHIL 123 AI & Ethics

10 / 18

Duty, Consequences, and Moral Choice

Utilitarianism: Consequences, Rules, and Two-Level Reasoning

The Transcript-Storage Decision

Consider a fictional case with stipulated evidence. A community college like yours is deciding what to do about the AI tutor its Academic Affairs office piloted last term. The pilot went well by most obvious measures. Pass rates in gateway math rose. Students who never went to office hours got feedback at eleven at night. Waiting lists at the writing center shortened. The vendor’s dashboard reports fewer withdrawals and higher completion rates in developmental writing.

The harder question came from a student worker in the library. She showed a colleague what the tutor stores by default: full transcripts of every tutoring session, and, in the voice version, audio recordings of the student and any classmate audible in the background. Faculty and staff can request access to those transcripts to improve support. The vendor keeps them indefinitely so the model can be fine-tuned. Students opted into the tutor when they clicked through the terms; they did not exactly opt into being an ongoing training set for a company they did not choose.

Now the college has a specific policy question in front of it. Transcript and audio storage is currently on by default, with an opt-out buried in account settings. The committee could keep that default to support faculty review, switch to informed opt-in, or require deletion after a set window. Every option in that meeting has real defenders and produces real winners and losers.

The pilot committee needs a way to think carefully about what happens to whom, and about which of those effects should decide the question. Blanket appeals to “student success” or “student privacy” do not settle it, because both are partly at stake in both directions. Utilitarianism is one serious answer to that kind of question. It says the moral weight of a policy lies in its consequences for the well-being of everyone the policy affects. The tradition running from Jeremy Bentham and John Stuart Mill through R. M. Hare and Peter Singer has spent two centuries refining what that claim actually commits us to. By the end of this chapter, you should be able to use utilitarian reasoning in a form careful enough to survive a serious Kantian objection, and clear enough to defend in writing.

From Enlightenment Reform to Philosophical Radicalism

Utilitarianism took its classical form in Britain during the late eighteenth and nineteenth centuries. Its sources crossed national borders. The British moralists Francis Hutcheson and David Hume had already connected morality with happiness, benevolence, and social utility. Jeremy Bentham also learned from the French thinker Claude-Adrien Helvetius and from Cesare Beccaria, the Italian critic of cruel and arbitrary punishment. Bentham’s achievement was to gather several of these ideas into a secular and systematic framework for moral judgment, law, and political reform. The histories in the Stanford Encyclopedia of Philosophy and its account of Jeremy Bentham trace that mixed British and European inheritance.

The contrast with the previous frameworks in this course helps locate Bentham’s project. Aquinas placed practical reason within a created order of human goods and eternal law. Kant grounded obligation in the rational will’s capacity to give universal moral law to itself. Bentham began with the effects that laws and actions have on sentient lives. He wanted a public standard for judging punishments, prisons, legal procedures, poor relief, political representation, and other institutions that distributed suffering and advantage. His Introduction to the Principles of Morals and Legislation was written as groundwork for a penal code. Utilitarianism entered modern philosophy partly as a method for asking what law ought to accomplish.

This orientation drew strength from British empiricism. Bentham followed Bacon, Locke, and Hume in treating experience and sensation as basic sources of knowledge. Pleasure and pain gave moral and legal inquiry an observable point of contact with the lives being governed. Bentham used that standard against appeals to custom, inherited rank, natural law, and original contract when those appeals insulated existing institutions from criticism. He was especially hostile to the fictions and obscurities he found in English common law and in William Blackstone’s defense of it. The UCL Bentham Project describes the principle of utility as the starting point for a much wider critique of legal and political institutions.

Bentham’s alliance with James Mill turned this philosophy into an organized reform movement. Their circle became known as the Philosophical Radicals. It included the political economist David Ricardo, the legal thinker John Austin, the historian George Grote, and the young John Stuart Mill. The group linked Bentham’s principle of utility with political economy, representative government, education, freedom of the press, and institutional design. They used organizations and publications such as the Westminster Review to press for changes in British public life. The SEP account of James Mill shows how the movement joined Bentham’s moral standard to a program of legal, administrative, and democratic reform.

John Stuart Mill inherited this project through an unusually intense education directed by his father. He learned Greek as a child, studied logic and political economy, edited Bentham’s writings, and was prepared to lead the next generation of radical reformers. The education gave him extraordinary analytical discipline. Mill later concluded that it had left his emotions and imagination badly underdeveloped. During the mental crisis he recounts in his Autobiography, the poetry of Wordsworth helped him recover capacities that Benthamite analysis had neglected. Coleridge, Romanticism, and French social thought also changed how he understood culture, character, and historical development. Mill kept happiness as the ultimate moral standard while giving liberty, individuality, dignity, higher pleasures, and emotional cultivation a larger place in a good human life. The SEP account of Mill’s moral and political philosophy places those revisions inside his continuing commitment to utilitarianism and reform.

Classical utilitarianism therefore developed inside a British reform environment centered in London, with Scottish and continental sources feeding it. Its practitioners regularly moved between philosophy, journalism, political economy, law, and public administration. That history helps explain why utilitarian reasoning fits the college’s AI tutor decision so readily. The framework was built to compare institutional alternatives, trace their effects across a population, and test whether an established practice actually improves the lives of the people living under it. The framework still has to explain what counts as well-being and how competing effects should be compared. Those questions lead into the philosophical structure of utilitarianism itself.

The Question Utilitarians Ask First

Utilitarianism belongs to a larger family called consequentialism. Consequentialists hold that whether an action, rule, or policy is morally right depends on what follows from it. A virtue ethicist asks what a person of practical wisdom would do and what character the action expresses. A Kantian asks whether the reasoning behind the action could be willed as a universal law and whether it treats persons as ends in themselves. A consequentialist asks a different opening question. What actually happens, to whom, and how does the resulting state of the world compare with what would have happened under the other choices available? The Stanford Encyclopedia of Philosophy’s entry on consequentialism is a useful place to see the fuller taxonomy.

Utilitarianism is the most famous consequentialism because it gives a definite answer to the next question inside the family. Which consequences count, and how do we compare them? The utilitarian answer is that consequences count as welfare, well-being, or interests. The moral question is how good or bad the outcome is for the lives of the beings affected. Consequentialists who use a different standard belong to the wider family outside utilitarianism.

That answer forces harder questions in any institutional AI decision. The affected group may include direct users, bystanders, workers whose jobs change, and communities whose data enters a private system. Decision-makers also need an account of well-being. A passing grade is a benefit only if it tracks learning the student can carry into later courses. A rise in completion is a proxy for welfare, and that proxy can drift from the good it was meant to track. The time horizon changes the comparison as well: this term may look different from a student’s four-year arc. Finally, the institution must decide how to compare a small benefit spread across many people with a serious harm concentrated on a smaller group.

Consequentialism leaves those questions open. Utilitarianism gives one systematic answer, and the rest of the chapter unpacks what that answer commits us to and where it strains.

What the Happiness Slogan Leaves Out

Students often meet utilitarianism through the slogan “the greatest happiness for the greatest number.” The slogan helps a little. It signals that consequences count, that many people count, and that we are supposed to aim for good outcomes and avoid bad ones. It also hides more than it reveals. This chapter unpacks classical total-maximizing utilitarianism into five commitments. Other philosophical taxonomies divide the tradition differently.

The first commitment is consequentialism itself. Actions, rules, and policies are judged by outcomes. The second is welfarism: the outcomes that carry moral weight are the ones affecting welfare, well-being, or interests. Bentham and Mill both defend hedonism, which holds that welfare consists in pleasure and the absence of pain. Later utilitarians develop preference- and interest-based accounts; some objective-list accounts belong to the broader consequentialist family. The tradition keeps welfare at the center while continuing to argue about what welfare is.

The third commitment is impartiality. Each person’s welfare counts on the same terms as anyone else’s. Rank, wealth, geographic distance, and personal connection do not change the basic weight of a person’s welfare. In Chapter V of Utilitarianism, Mill summarizes Bentham’s position with the dictum “everybody to count for one, nobody for more than one.”

The fourth commitment is aggregation. Utilitarianism adds up the welfare of the people affected so that choices spreading benefits and harms across many lives can be compared. The fifth is maximization. The morally best choice is the one whose overall welfare total is highest among the options available.

Each of these commitments has been challenged inside the utilitarian tradition, and contemporary utilitarians disagree with each other about what welfare is, whether pure maximization is defensible, and how impartiality should be balanced against special obligations to family, colleagues, or students. You cannot debate utilitarianism carefully if you treat it as a single lump. When you see a criticism of utilitarianism, ask which of the five commitments the criticism actually targets.

Counting Consequences Without Pretending They Are Easy

Jeremy Bentham (1748–1832) gave one of the first systematic modern versions of utilitarianism. He was a legal reformer more than an armchair philosopher, and he wanted a moral theory that could evaluate laws, prisons, punishments, and public policies on the same terms as private choices. His starting move was to refuse appeals to custom, rank, tradition, or authority as self-justifying. If a legal practice was defensible, Bentham thought, it should be defensible by the difference it made to the welfare of the people affected.

Bentham grounds welfare in pleasure and pain. He opens Chapter I of An Introduction to the Principles of Morals and Legislation with a striking sentence: “Nature has placed mankind under the governance of two sovereign masters, pain and pleasure. It is for them alone to point out what we ought to do, as well as to determine what we shall do.” One useful reading sees the sentence moving from a claim about what motivates human beings to a claim about what should guide moral judgment. Bentham then defines utility as the tendency of an action “to produce benefit, advantage, pleasure, good, or happiness” for the party whose interests are at stake, or to prevent the opposite. The Econlib edition of Chapter I provides the full passage.

Two features of Bentham’s picture are useful for AI ethics. The first is his account of what a community is. Bentham treats the community as the sum of the interests of its members. He rejects a separate community-level interest floating above individual welfares. When an institution claims that an AI policy promotes “student success,” a Bentham-shaped question asks whose lives become better and by how much. Aggregate claims about a community have to cash out as effects on identifiable people to count as welfare claims.

The second useful feature is the hedonic calculus. In Chapter IV, Bentham lists seven properties relevant to estimating how much a pleasure or pain contributes to an evaluation: intensity, duration, certainty, propinquity (nearness in time), fecundity (whether it tends to produce further pleasures or pains), purity (whether it comes mixed with the opposite), and extent (how many people share it). Bentham accepts estimation and summation in principle, while also saying that strict calculation need not precede every moral judgment. PHIL 123 uses the calculus as a consequence map. The map directs attention toward probability, timing, downstream effects, mixed costs, and the number of people affected while leaving the result qualitative when the evidence cannot support exact arithmetic. Bentham’s qualifications appear in the Econlib edition of Chapter IV.

Bentham’s calculus makes the comparison more disciplined. A quick service benefit may be intense and highly certain for direct users, while a privacy harm may be diffuse in the short term but high in fecundity because a leak or unplanned downstream use produces further harms. Effects on bystanders also count even when they never intended to participate. The calculus reveals how incomplete an institutional comparison remains when it counts only its fastest and most visible benefits.

Bentham’s reforming instincts widened the moral circle in ways later utilitarians developed further. He argued that the morally relevant question about non-human animals is whether they can suffer, and his reform projects addressed prisons as well as some legal inequalities affecting women. Later philosophers usually classify Bentham as an act-focused utilitarian, meaning that his official criterion of rightness applies to individual actions and their consequences. That classification is a reasonable summary of where his principle points. It is anachronistic if it is taken to mean Bentham chose between “act” and “rule” versions of the theory the way twentieth-century philosophers did. The act-versus-rule terminology developed later and reads back onto Bentham a distinction he never had to make. The SEP entries on the history of utilitarianism and on Jeremy Bentham give the fuller picture.

How Mill Ranks Different Kinds Of Pleasure

John Stuart Mill (1806–1873) grew up inside the Benthamite project. His father James was one of Bentham’s closest collaborators, and Mill was tutored in utilitarian reasoning from childhood. He kept the utilitarian project intact through his mature work while revising several features of Bentham’s account that struck him, and generations of readers after him, as flat.

Mill’s best-known revision concerns the account of happiness. Bentham treats pleasures as differing only in the quantitative properties of the calculus. Mill argues that pleasures also differ in quality. Some pleasures involve the exercise of higher human capacities such as reasoning, imagination, moral feeling, and self-development. Others are simpler bodily satisfactions. The line Mill is remembered for makes the point sharp: “It is better to be a human being dissatisfied than a pig satisfied; better to be Socrates dissatisfied than a fool satisfied.” Mill’s test for the distinction relies on the judgment of people who have experienced both kinds of pleasure. Those competent judges, he argues, reliably prefer the higher pleasures even when the lower ones are more intense on Bentham’s scale. Mill’s Utilitarianism is at Project Gutenberg, and the SEP entry on Mill’s moral and political philosophy gives a careful contemporary reading.

Mill’s distinction changes how a college evaluates AI-supported instruction. A student’s short-term satisfaction with a graded assignment is a different kind of good from the long-term development of judgment, careful reading, and moral seriousness. A system can produce quick satisfaction at scale while degrading the capacities a college education is supposed to develop. Completion and pass rates still count, but they remain proxies. If students cannot reconstruct their reasoning in the next course, higher completion has failed to establish the learning gain the metric was supposed to represent.

Mill also takes moral rules more seriously than a caricature of utilitarianism suggests. In everyday moral life, he thinks we should be guided by what he calls secondary principles: the ordinary moral rules about honesty, promise-keeping, fair dealing, and respect for others. Those rules are stable social achievements, learned over long stretches of human experience. Mill argues they are justified by their tendency to protect welfare over the long run. He also gives special weight to justice, to individual rights, and to liberty. In On Liberty, he defends what has come to be called the harm principle, holding that the only legitimate ground for coercing an adult against their will is the prevention of harm to others. The Project Gutenberg edition of On Liberty is easy to consult.

Because Mill puts so much weight on rules, rights, justice, and liberty, later philosophers sometimes call him a rule utilitarian. The reading captures something real. Mill has strong rule-utilitarian and indirect-utilitarian tendencies, and his treatment of justice in Chapter V of Utilitarianism pushes hard in that direction. It is safer to say Mill has those tendencies than to flatly call him a rule utilitarian, because his theory still uses welfare as the ultimate ground for the rules. Their moral force comes from the goods they protect. The IEP entry on Mill’s ethics sorts through the interpretive dispute in more detail.

For our purposes, Mill leaves us with a richer utilitarianism than Bentham’s. Happiness remains the ultimate standard, while dignity, higher human capacities, justice, individual liberty, and stable secondary principles play a larger role in moral reasoning.

Why Act-By-Act Calculation Breaks Down

When contemporary philosophers speak of act utilitarianism, they usually mean the theoretical claim that an action is morally right when its actual consequences produce at least as much overall welfare as any available alternative. Ordinary agents must decide before those consequences are known, so classroom arguments compare reasonably expected consequences instead. The criterion of rightness and the practical method of decision are related but distinct. The IEP article on act and rule utilitarianism gives a useful survey.

Act utilitarianism states the framework in its most transparent form: evaluate the specific choices in front of you by their consequences, including who benefits and who is harmed. That directness gives an institution a clear starting point.

Under real conditions the theory strains in ways students should notice. A college choosing an AI system rarely knows how the tool will affect study habits across four semesters or how stored data will be used three years later. Missing evidence about long-term learning, vendor behavior, and downstream data use sharply limits confidence in an act-by-act verdict.

A second strain is demandingness. If every action is judged against every available alternative, ordinary moral life becomes an endless optimization problem. Williams’s integrity objection adds a distinct concern. In “A Critique of Utilitarianism”, he argues that impartial maximization can estrange agents from the projects and commitments through which they act as particular people. A moral theory distorts agency when it treats a person’s projects, promises, and relationships as inputs that must always yield to the aggregate.

A third strain is that maximizing aggregate welfare can license doing serious harm to a small group in exchange for small benefits spread across many. Taken alone, act utilitarianism has trouble ruling out stipulated pressure cases in which sacrificing one person produces the highest total. We will develop that objection more fully later in the chapter.

A helpful move at this point is to separate two questions. The criterion of rightness identifies which action would in fact be morally right. A decision procedure guides ordinary agents who must choose before they know the actual consequences. Bentham concedes that strict felicific calculation need not precede every judgment. Practical guidance under uncertainty can therefore rely on simpler procedures even while the underlying criterion remains exposed to the demandingness and concentrated-harm objections.

Rules As Consequence-Protecting Practices

Rule utilitarians take the strains seriously and reframe the question. The formulation used in this chapter asks which code would have the best expected consequences if generally accepted. Individual actions are then evaluated by whether they comply with that justified code. Other versions emphasize compliance, internalization, or public adoption. The SEP entry on rule consequentialism sets out the contemporary landscape.

The shift plays out most clearly at institutional scale. Colleges, hospitals, and government agencies set policies, procurement rules, disclosure requirements, training expectations, and platform defaults that shape many downstream actions. AI systems amplify this pattern because a recommender rule, triage standard, or chatbot default is repeatedly applied after deployment. Rule-level thinking asks what happens when a choice becomes standard practice and which code produces the best overall outcomes for everyone affected.

The Michigan Integrated Data Automated System (MiDAS) case makes the rule-level problem concrete. A Michigan performance audit explains that the system was intended to increase automation, improve customer service and data accuracy, and lower operating costs. In Cahoo v. SAS Analytics, the Sixth Circuit recounts allegations that, during a fully automated period from 2013 to 2015, fraud determinations were issued without meaningful human review and triggered tax-refund intercepts, wage garnishment, and a fraud penalty equal to four times the benefits received or sought. The complaint alleged that about 93 percent of the determinations later reviewed were false. Because the case was at the motion-to-dismiss stage, the court accepted the complaint’s allegations as true; the percentage should not be treated as a final audit finding, and the dates attached to the underlying review remain uncertain. A separate 2022 Michigan Attorney General release announced a proposed settlement in the Bauserman litigation.

The utilitarian post-mortem focuses on the rule governing automated adverse determinations. Efficient enforcement of a public benefits program has consequences that count, but a policy without meaningful human review or an accessible appeal path exposes wrongly flagged claimants to severe losses. A rule utilitarian should compare that policy with one requiring review before collection and a clear opportunity to contest the determination. Public trust, the labor absorbed by wrongful appeals, and the costs of litigation and administrative correction belong in that comparison alongside enforcement speed.

A similar pattern shows up in health-care algorithms. A widely used commercial tool predicted future health-care costs as a proxy for health need when deciding which patients should receive extra care management. In a 2019 Science study, Ziad Obermeyer and his coauthors found that the proxy hid unmet need among Black patients: because Black patients had historically received less care per unit of illness, they generated lower costs, so the risk score underestimated their need relative to equally ill white patients. Kara Manke’s Berkeley summary explains the mechanism in accessible terms. The study establishes the proxy failure. The further claim that institutions should adopt a general rule requiring proxy audits before scaled deployment is this chapter’s rule-utilitarian inference from the evidence.

Rule utilitarianism grounds rules in their consequences, while Kantian deontology grounds duties in rational agency and the intrinsic dignity of persons. The two theories can endorse the same policy even though a change in expected welfare can change the rule-utilitarian judgment.

Rule utilitarianism also faces a paired pressure test. A code followed rigidly through a disastrous exception risks what J. J. C. Smart called rule worship. Yet a code rewritten with an exception for every case in which breaking it would maximize welfare starts to collapse toward act utilitarianism. Contemporary rule consequentialists answer that a code must be judged together with the costs of teaching, internalizing, publicizing, and enforcing it. That reply explains why a simple and stable rule may outperform a maze of exceptions. It leaves two hard questions open: where a disaster threshold should sit, and whether we can predict the real costs of rival codes well enough to choose among them. Hare’s account addresses the same practical tension through a different theory of deliberation.

When the Rule Is Not Enough

R. M. Hare (1919–2002) offered a version of utilitarianism that gives useful guidance for PHIL 123 and for AI ethics generally. In Moral Thinking, Hare distinguishes two kinds of moral thinking a competent agent uses in different situations. The SEP entry on Hare sets out the broader philosophical context.

The intuitive level is where most of your moral life happens. At this level, you rely on stable moral principles you have already internalized, including honesty, promise-keeping, respect for trust, and protection of people who cannot easily protect themselves. These principles do good work over the long run because they are stable, learnable, and reliable enough for other people to depend on. They also spare you the cognitive load and temptations of calculating every choice from scratch. Bentham’s practical concession and Mill’s secondary principles are historical precursors to this account, though Hare gives intuitive principles a more substantial role than disposable rules of thumb.

The critical level is where you shift when the intuitive level cannot settle the case. Two ordinary principles may conflict, forcing you to decide which claim is more serious. An inherited principle may have been built for a situation that no longer matches the case in front of you. Available evidence may also give you serious reason to think the principle will misfire. High stakes should make you slow down and review the evidence, but they do not by themselves show that a principle should be overridden. Hare’s critical thinking weighs preferences impartially, selects and revises principles for intuitive use, and governs their application. PHIL 123 translates that distinction into four practical triggers; Hare develops the levels in a different form.

Two-level utilitarianism is easy to confuse with rule utilitarianism because both give ordinary principles a large role. Rule utilitarianism makes conformity to the best justified code the criterion of rightness. Hare keeps preference-based act utilitarianism at the critical level and uses it to select, revise, and resolve conflicts among intuitive principles. PHIL 123 borrows Hare’s distinction between levels as a deliberative tool across act and rule arguments, a broader classroom use than Hare’s own theory.

Now return to the transcript-storage default. The pilot committee is deciding whether AI tutor transcripts and audio recordings should be stored by default, off by default with opt-in, or stored only under specific conditions.

Start with the intuitive level. Stable principles oppose gathering personal data by default without informed consent, repurposing data collected for one aim, and placing the burden of protection on the least-resourced users. Honesty also requires the college to explain how student work and voices will be used. These principles initially support storage off by default with a clear opt-in.

The pilot chair might press back. She might say the case is genuinely unusual and calls for critical-level reasoning. The transcripts are how faculty catch common misunderstandings that the tutor is producing at scale. The developmental writing department gained ten points on pass rates last term partly because two English instructors read transcripts and noticed the tutor was praising confident errors in student drafts. Storage off by default may protect privacy at the cost of learning support the college could not otherwise provide. She may also say the intuitive-level rules do not squarely settle the case because two of them pull against each other. Protecting privacy pulls against improving learning support. Reasonable people can disagree about which claim is more serious. That is the sort of internal conflict two-level thinking treats as a critical-level trigger.

At the critical level, the committee has to reason directly about consequences. Default storage may help faculty identify common misunderstandings, but it creates concentrated risks for students whose speech is identifiable or whose classmates are audible in the background. A rule prohibiting institutional storage reduces those risks while removing a source of instructional evidence. Limited-retention opt-in preserves some faculty access after informed consent, restricted access, and scheduled deletion. Its costs include administrative work and an incomplete sample if many students decline.

The decisive comparison concerns trust, instructional value, and enforceability under general adoption. Limited-retention opt-in appears to have the strongest expected welfare balance on the stipulated evidence because it reduces concentrated privacy risk while preserving a bounded route to instructional review. That verdict depends on understandable consent, enforced access rules, and full support for students who decline. New evidence about participation, enforcement, learning gains, or privacy risk could change the comparison.

The conclusion remains provisional because the case supplies limited evidence. A competent agent should be able to defend the judgment from the information available now and revise it when the expected consequences change.

Two-level thinking has real limits. Hare worries about the temptation to declare a case unusual whenever you want to break a principle that would otherwise apply to you. As the SEP discussion of Hare’s levels explains, Bernard Williams presses a different worry: intuitive commitments may become psychologically unstable when agents simultaneously understand them as utility-producing instruments that critical reasoning can override. If a principle has authority only while it maximizes utility, an agent’s awareness of its provisional status may weaken the settled commitment the intuitive level requires. Hare’s functional answer assigns each level a distinct job, but the tension remains. A responsible move to the critical level therefore requires a public explanation of the features that call for direct reasoning, one that another person can evaluate.

The Objection Utilitarianism Has To Face

This chapter’s principal objection is that maximizing aggregate welfare can license serious harm to a smaller group in exchange for benefits spread across many. Rule utilitarianism and two-level reasoning provide resources for preventing many practical versions of that harm. The remaining question is whether those resources can also justify protecting the person whose loss would improve an accurate aggregate.

State the objection in its strongest form. Aggregation adds welfare across persons. It treats a small harm to a large group and a serious harm to a small group as commensurable, and it can prescribe the second whenever the arithmetic favors it. A healthy patient whose organs would save five others is the familiar stipulated pressure case. Institutional examples show the same structure without relying on a life-or-death thought experiment. MiDAS traded due-process protections for some unemployment claimants against intended administrative efficiency. The health algorithm traded medical attention for Black patients against the convenience of using predicted cost as a proxy for health need.

In Section 5 of A Theory of Justice, John Rawls argues that utilitarian aggregation fails to take seriously the separateness of persons. A gain to one person does not literally compensate another person for an uncompensated loss. Better measurement can reveal who bears the burden, but a correct aggregate may still authorize it. Rawls therefore challenges aggregation as a standard of justification even when the data is accurate. Williams’s earlier integrity objection targets a different problem: what impartial maximization does to the agency and projects of the person who must act.

Utilitarianism has a serious practical reply. Institutional legitimacy, the stability of moral rules, and the willingness of vulnerable groups to participate in shared systems are consequences too. MiDAS illustrates the point: the intended efficiency gain came with wrongful collections, damaged livelihoods, appeals, and litigation costs. Effects on institutional trust also belong in the comparison once evidence establishes them. The supported consequences are already enough to make the proposed efficiency policy lose on utilitarian grounds.

Subgroup analysis can expose harms hidden inside an aggregate. The Obermeyer study shows the mechanism: a cost proxy hid unequal unmet need, so the risk score treated equally ill patients differently. This chapter’s contemporary application of rule and two-level reasoning supports an ordinary institutional principle requiring subgroup and proxy audits before scaled deployment. The principle protects long-run welfare by making concentrated harms visible before they are normalized.

The criterion of rightness and the decision procedure can also be separated. Even if the utilitarian criterion says that an act of sacrifice would be right in some possible case, real institutions may be unable to identify that case reliably. Giving officials broad permission to override protective rules can produce predictable harm. Finite agents should normally use stable principles that protect vulnerable people and subject proposed exceptions to public scrutiny.

The reply handles many practical cases in which fuller consequence counting reverses an apparently efficient policy. It does not dissolve Rawls’s objection. A Kantian can add that protecting a person only because protection improves the aggregate gives the wrong form of justification. On that view, a person should be protected because other people do not get to trade away that person’s welfare for their own benefit. Utilitarian and Kantian reasoning often converge on a policy while continuing to disagree about why the policy is justified and what should happen if the aggregate calculation changes.

Carry one unresolved question into the Kant chapter and the framework comparison: would the vulnerable person remain protected if sacrificing them produced the better aggregate? Utilitarianism protects through welfare-promoting rules and decision procedures. Kantian ethics treats the person’s rational agency and dignity as grounds the aggregate cannot replace.

How Far the Circle Reaches

Peter Singer is one of the best-known contemporary utilitarians. His essay “Famine, Affluence, and Morality” argues that distance does not cancel our obligation to prevent serious suffering when we can do so without sacrificing something comparably important. His work on animal ethics similarly extends moral consideration beyond human beings capable of speaking for themselves. Liam Hampton’s profile of Singer gives a compact overview.

Singer has often been associated with preference utilitarianism, which evaluates welfare partly through the satisfaction of preferences or interests. Preference views can make sense of choices someone makes about their own life even when the choice does not maximize pleasure. Both sentience-based hedonistic views and preference-based views can include non-human animals; they explain the relevant welfare differently. Singer’s position has also changed. In “Doing Our Best for Hedonistic Utilitarianism”, Singer and Katarzyna de Lazari-Radek defend a hedonistic form of act utilitarianism rather than the preference view commonly associated with Singer’s earlier work.

The Centre for Effective Altruism defines effective altruism as a project that uses evidence and reason to find effective ways of helping others and acts on what it finds. Applied utilitarianism is one important influence on the movement, especially its impartial concern for distant strangers, but effective altruism is not identical with utilitarianism and includes people with different moral theories. If you meet the movement in a later class or public argument, identify the specific utilitarian commitments the argument draws on rather than treating the labels as interchangeable.

These contemporary arguments apply utilitarian impartiality to distant suffering and non-human animals while reopening the question of what welfare consists in. Utilitarianism remains a family of positions whose members disagree about pleasure, preferences, interests, and the demands of impartiality.

Building A Utilitarian Argument

You now have enough of the framework to build a utilitarian argument in the PHIL 123 four-move form. Before writing the premises, answer two questions. First, are you directly evaluating an act or a general rule? Second, can justified intuitive principles settle the case, or does the case require critical reflection? These questions do different jobs. Act and rule utilitarianism identify what receives direct evaluation. Two-level reasoning guides how you deliberate.

P1 must identify the version of utilitarianism you are using. In an act argument, the standard applies directly to the actions genuinely available in this case. In a rule argument, the standard first applies to competing rules or codes. You compare the expected consequences if each code were generally accepted, then show whether the action follows the justified code. A genuine rule-utilitarian argument compares rival codes; a comparison of one individual act remains act-level reasoning.

Most utilitarian arguments succeed or fail in P2 and P3. A weak P2 says only that an option “helps more people” or “improves outcomes.” A strong P2 identifies the live alternatives, the affected people or sentient beings, the kind of welfare at stake, the likely magnitude and duration of the effects, their distribution, and the evidence behind the prediction. It also marks important uncertainty. When the evidence does not support a precise utility score, use a careful qualitative comparison.

P3 prevents the argument from jumping from one benefit to an overall verdict. It states why the evidence favors one act or rule over its rivals. The comparison should consider probability, duration, distribution, and concentrated harm when those features carry the dispute. If severe harm to a subgroup disappears inside a general claim about “most people,” the bridge needs repair.

Two-level reasoning helps you decide when to rely on an ordinary principle and when to reopen the problem. Begin with the ordinary welfare-protecting principles relevant to the case. Move into critical reflection only if those principles conflict, fail to cover a novel case, seem likely to misfire seriously, or are themselves under evaluation. A high-stakes decision deserves careful review, but high stakes alone do not authorize an override. Once critical reflection is justified, you must still complete the act or rule comparison. The phrase critical level cannot substitute for P2 or P3.

The transcript-storage argument illustrates the full structure. The committee evaluated the general adoption of three policy rules. It identified a conflict between privacy and learning-support principles, which justified critical reflection. It then compared the expected effects of default storage, no institutional storage, and limited-retention opt-in for the affected groups. On the stipulated evidence, the opt-in rule won provisionally because it preserved some educational benefit while reducing concentrated privacy risk. New evidence could change that comparison, so the conclusion remains bounded.

When a four-line argument hides too much, expand the step that needs more support. Separate the observable facts from the prediction, state the source of the prediction, or add a line showing that the action conforms to the selected rule. Your goal is an argument another person can inspect. More labels are useful only when they reveal reasoning that would otherwise remain hidden.

In PHIL 123, utilitarianism is often the strongest starting framework when a choice affects many people at scale. It gives us a way to think about aggregate outcomes, institutional practices, and the welfare of people who are not in the room. AI policy, public health, resource allocation, and platform design all spread consequences across populations.

PHIL 123 uses Kantian deontology as a stronger starting framework when an aggregate benefit would come at the cost of using a specific person or group as an instrument for someone else’s ends. When an AI system produces gains for most students by extracting data from a small group without meaningful consent, or when a public benefits system runs on false accusations against people who cannot afford to fight back, Kantian reasoning explains the wrong more directly. Your framework comparison should locate the point where the welfare total and the treatment of a particular person pull apart, then explain which standard should govern the judgment.

References

Scholarly layer

Person record