Ethics

Complete course reading notes

Read every lesson below. The interactive reader above contains the same explanations, with visual tools and quizzes.

1. Three questions: normative, metaethical and applied

Learning goal: Sort a moral question by level and test whether an argument supports its conclusion.

Ethics is not one question. An applied question asks what to do in a case: may a manager read employee email, is it wrong to break a lease early. A normative question asks what makes any act right or wrong: its consequences, its fit with duties, the character it expresses, or whether others could reasonably reject it. A metaethical question steps back further and asks what moral claims are: statements that can be true or false, expressions of attitude, or something else. The Stanford Encyclopedia of Philosophy entry on metaethics treats that last question as its subject.

Keeping the levels apart prevents a common slide. Someone argues "morality is only opinion" (metaethics) and concludes "so the manager did nothing wrong" (applied). The second claim needs its own support, and a moral anti-realist can still hold firm views about the case. Each level can be argued without settling the others, although positions at one level can constrain the others.

An argument is valid if the conclusion must be true whenever the premises are true. It is sound if it is valid and the premises are in fact true. Moral arguments often become valid only after a hidden premise is added, so a main skill in this course is to write the premises out. Every later lesson uses the same routine: state the principle, apply it to a case, then ask what a critic would say.

Worked example

Classify and test: "Because most people in my town accept it, it is fine to underpay interns." Name the missing premise.

The conclusion is an applied claim about what is fine.
The premise is a descriptive claim about what people accept.
To connect them we need a premise such as "whatever most people accept is morally fine".
That premise is a substantive normative claim and many critics reject it. The argument is invalid without it and doubtful with it.

Practice problem and solution

A critic writes: "No one can prove moral claims, therefore this layoff was not wrong." At which level is the premise stated, and what hidden premise would the argument need to reach its conclusion? Name the level in the answer box.

The premise is about whether moral claims can be proved, which is metaethics. The conclusion is applied. A bridge premise is missing, such as 'a claim that cannot be proved cannot support a verdict that an act was wrong', and a critic can reject that premise.

Mental model: Applied cases, normative standards and metaethical status are three questions; keep them apart and write out hidden premises.

Common trap: Letting a metaethical doubt quietly decide an applied case.

2. Consequentialism and utilitarianism

Learning goal: Apply the principle of utility, separate act from rule versions and state the main objections.

Consequentialism says the rightness of an act depends on the goodness of its consequences. Its classic form is utilitarianism, whose classic proponents were Bentham, Mill and Sidgwick, according to the Stanford Encyclopedia of Philosophy. Mill states the creed this way: "actions are right in proportion as they tend to promote happiness, wrong as they tend to produce the reverse of happiness," where happiness means pleasure and the absence of pain.

Act consequentialism applies the test to each act: choose the act that maximizes the total good for all. Rule consequentialism applies it to rules: the SEP says it makes the rightness of an act depend on the consequences of a rule. The two can disagree. An act that breaks a useful rule once, in secret, might maximize good alone yet be forbidden by the best rule.

Mill also distinguishes quality of pleasure from quantity: "It is better to be a human being dissatisfied than a pig satisfied; better to be Socrates dissatisfied than a fool satisfied." Critics reply that aggregation can sacrifice individuals, that it demands too much, and that the good is hard to measure. Replies include rule versions, satisficing and richer accounts of the good. The debate is not settled.

Worked example

Policy X gives +3 to each of 12 people and -5 to each of 4 people. Policy Y gives +2 to each of 10 people and no losses. Find each net total and say which an act utilitarian picks.

X: 12 x 3 = 36 gains, 4 x 5 = 20 losses, net 16.
Y: 10 x 2 = 20 gains, no losses, net 20.
Y has the higher net total.
An act utilitarian picks Y because 20 > 16, even though X helps more people.

Practice problem and solution

Hypothetical: a manager can break a confidentiality rule once in secret. Breaking nets +9 in that case (+6 to each of 4 people, -3 to each of 5 people); keeping the rule nets +6. Now suppose a rule of 'break whenever the case nets positive' is adopted across 5 such cases, with trust erosion costing 6 points per case. A rule of 'keep' nets +6 per case. By how many points does the keep rule's total exceed the break rule's total? In one sentence, say why an act and a rule consequentialist may disagree here.

Break case: 4x6 - 5x3 = 24 - 15 = 9. Act utilitarian, looking at one case alone, picks break (9 > 6). Break rule over 5 cases: 5 x (9 - 6) = 15. Keep rule over 5 cases: 5 x 6 = 30. Difference: 30 - 15 = 15. The rule consequentialist judges the rule by its total consequences, including trust erosion, so it forbids what the act test alone would allow.

Mental model: Utilitarianism sums gains and losses across people; check the sum, not the number of beneficiaries.

Common trap: Counting beneficiaries instead of net total.

3. Kantian ethics: universal law and humanity

Learning goal: State the universal law and humanity formulas and apply them to a maxim.

Kant holds that the moral worth of an action lies in acting from duty, not from inclination. His supreme test is the categorical imperative, which commands without depending on a goal you happen to have. In the Abbott translation of the Groundwork, the first test appears as the demand "never to act otherwise than so that I could also will that my maxim should become a universal law."

A maxim is the principle you act on: "when I need money, I will make a promise I do not intend to keep." To apply the test, imagine everyone adopting the maxim and ask whether you could still will your own act. Kant uses the false promise as an example and separates the question whether it is prudent from the question whether it is right.

The humanity formula runs: "So act as to treat humanity, whether in thine own person or in that of any other, in every case as an end withal, never as means only." The word "only" matters: using a taxi driver as a means of transport is not wrong if you also respect them as a person with their own ends. Critics raise conflicts between duties and the rigidity of exceptionless rules; Kantians reply with refined versions.

Worked example

A seller hides a known defect in a used car so a buyer will pay full price. Which formula does the plan most directly violate, and why?

Maxim: "I will hide defects when it raises my price."
The buyer cannot agree to a deal whose key fact is hidden.
So the buyer is treated merely as a means to my gain.
The humanity formula is violated; the universal law test also fails because universal concealment would destroy trust in sales.

Practice problem and solution

Hypothetical: Case A: a firm tells applicants a job is a six-month contract and they accept knowing that. Case B: a firm tells applicants the job is permanent to fill a six-month gap, knowing it will end. Which case, A or B, treats applicants merely as a means? Give the letter, then explain with the word 'only' in 'never as means only'.

Case B. The firm gets the work by deception, so applicants cannot consent to the real terms and are used merely as tools. In Case A applicants know the facts and choose, so they are treated as ends as well as means. Using someone as a means is allowed; using them merely as a means is the ban.

Mental model: Kant gives two linked tests: universalize the maxim, and never treat persons merely as means.

Common trap: Reading "means" as "any use of a person".

4. Virtue ethics and the mean

Learning goal: Explain virtue as a trained state of character and apply the doctrine of the mean.

Virtue ethics asks what kind of person to be, not only what act to perform. In the West its founding figures are Plato and Aristotle, as the Stanford Encyclopedia of Philosophy entry notes. The SEP also records that the complaint that virtue ethics does not produce codifiable principles is still commonly voiced, and that judgment informed by experience, practical wisdom, is needed to apply rules or principles.

Aristotle's account in Nicomachean Ethics II.6 reads: "Virtue, then, is a state of character concerned with choice, lying in a mean, i.e. the mean relative to us, this being determined by a rational principle, and by that principle by which the man of practical wisdom would determine it." The mean is not an arithmetic midpoint; it is relative to the person and the situation.

He uses courage as an example: the person who flies from and fears everything becomes a coward, and the person who fears nothing at all but goes to meet every danger becomes rash. Courage lies between. Critics ask whether virtue ethics gives enough guidance and who counts as the wise person; defenders point to the guidance in exemplars and in detailed accounts of each virtue.

Worked example

A paramedic runs into an unstable building without checking for collapse risk, saying that fear is weakness. Use the mean to name the vice.

Courage is a mean between cowardice and rashness.
The paramedic disregards real danger and so fears nothing relevant.
That is the excess regarding fear.
The vice is rashness, not courage.

Practice problem and solution

Hypothetical: two investors each put 40% of their savings into one risky asset. P is a retiree who needs the money within a year. Q is 25 and has researched the asset and saves for decades. Using Aristotle's 'mean relative to us', which investor is more likely rash at 40%? Give P or Q, then say in one sentence why 50% is not automatically the mean.

P. The same exposure meets real danger without regard to P's situation, so for P it leans toward the excess regarding fear: rashness. For Q, with a long horizon and research, it may sit nearer the mean. The mean is relative to the person and set by practical wisdom, not an arithmetic midpoint.

Mental model: Virtue is a trained state between two vices, found by practical judgment.

Common trap: Treating the mean as a mathematical average.

5. Contractualism and justice from behind a veil

Learning goal: Distinguish Scanlon from Rawls and apply a maximin choice.

Contractualism holds that morality concerns principles that people could agree to, or could not reasonably reject. Scanlon's version, in his own summary quoted by the Stanford Encyclopedia of Philosophy, says an act is wrong if its performance would be disallowed by any set of principles for the general regulation of behaviour that no one could reasonably reject as a basis for informed, unforced, general agreement.

Rawls's contract differs. The SEP notes that Rawls seeks principles everyone would agree to, rather than principles no one could reasonably reject. His device is the original position, whose main feature is the veil of ignorance: to ensure impartiality, the parties are deprived of knowledge of their personal characteristics, conceptions of the good and social circumstances. The SEP reports that Rawls contends the parties would choose two principles of justice, the first guaranteeing equal basic rights and liberties.

The SEP also presents an argument for the principles from the maximin criterion: rank options by their worst-off position. It discusses objections as well. In this lesson maximin is a calculation tool, not a claim that it is the right rule in every setting.

Worked example

Society S1 has group incomes (12, 45, 80). S2 has (20, 35, 50). S3 has (18, 70, 70). Which does maximin select, and what is the worst-off income there?

List the minimum of each: S1 12, S2 20, S3 18.
The largest minimum is 20, in S2.
Maximin selects S2.
The worst-off income is 20.

Practice problem and solution

Hypothetical: society T1 has group incomes (30, 30, 50). T2 has (25, 60, 90). T3 has (28, 40, 41). A maximin rule picks the society with the best worst-off income; an average-income rule picks the highest mean. How many income points higher is the worst-off income in the maximin pick than in the average-income pick? In one sentence, say why Rawls's veil of ignorance is offered as a reason to rank by the worst-off.

Minimums: T1 30, T2 25, T3 28; maximin picks T1 (30). Means: T1 36.67, T2 58.33, T3 36.33; the average rule picks T2, whose worst-off is 25. Difference: 30 - 25 = 5. Behind the veil, parties do not know their position, so they may care about the worst position they could land in. Maximin here is a calculation tool, not a claim that it is always the right rule.

Mental model: Maximin ranks options by their worst position.

Common trap: Choosing by average or by highest income.

6. Doing and allowing, intention and foresight

Learning goal: Apply the doing/allowing distinction and double effect to paired cases.

Deontological views often hold that some constraints bind more strongly than duties to aid. Two ideas support this. The doing/allowing distinction contrasts causing harm with letting harm happen. The Stanford Encyclopedia of Philosophy shows that it is not the same as action versus inaction: a person who fails to turn up and so spoils a performance does harm by inaction.

The doctrine of double effect says that sometimes it is permissible to cause harm as an unintended, merely foreseen side effect of bringing about a good result, even though it would not be permissible to cause that harm as a means to the same end. The SEP gives a standard pair: the terror bomber aims at civilian deaths to weaken enemy resolve, so the deaths are intended; the tactical bomber aims at military targets while foreseeing civilian deaths, so the deaths are foreseen but unintended.

The principle is contested. The SEP lists traditional formulations with conditions, including that the act be morally good or indifferent and that the agent not positively will the bad effect, and critics question whether intention can carry this much moral weight. The SEP also discusses trolley cases, including Thomson's idea that redirecting an existing threat may not count as causing harm. Treat these as analytical tools, not settled rules.

Worked example

A planner destroys a bridge to stop an enemy supply convoy, foreseeing that some civilian workers nearby will die. Their deaths help neither the plan nor the aim. How does double effect classify those deaths?

Name the aim: stopping the convoy.
The deaths are not part of how the aim is achieved.
They are foreseen but not intended.
They are a foreseen side effect, still subject to the other conditions.

Practice problem and solution

A planner jams a radio tower to disable enemy command, foreseeing that a nearby clinic will lose power briefly. The power loss does not serve the plan. How does double effect classify it? One or two words.

The power loss is a side effect of the jamming, not the means. Double effect calls it foreseen but unintended.

Mental model: Intention and foresight can differ, and many deontologists give them different moral weight.

Common trap: Treating any foreseen harm as automatically intended.

7. Metaethics: realism, expressivism, error theory

Learning goal: Distinguish cognitivist, noncognitivist and error-theoretic views of moral claims.

Metaethics asks what moral claims are. Moral realism, in broad terms, holds that moral facts exist objectively. The Stanford Encyclopedia of Philosophy entry on moral anti-realism takes anti-realism to be the denial of that view, and it identifies broadly three theses: moral noncognitivism, moral error theory and moral non-objectivism.

Noncognitivists say moral sentences do not state facts but express attitudes or commitments. The SEP metaethics entry describes quasi-realist expressivism as taking on the project of explaining, on noncognitivist foundations, why moral language has all the trappings of fact-stating talk. Error theorists say that the properties needed to make moral claims true, such as moral wrongness, are not instantiated. The SEP lists Mackie among its defenders.

These are rival theories, not settled results. A realist must say how moral facts fit into the world and how we know them. An expressivist must explain the logic of moral talk. An error theorist must say what to do with moral language. Whichever view you hold, state what it implies about truth, knowledge and disagreement.

Worked example

Someone says: "Moral sentences purport to describe moral facts, but there are no such facts, so 'torture is wrong' is false." Name the view.

The speaker accepts that moral sentences purport to state facts.
The speaker denies that the facts exist.
So positive moral claims come out false.
That is moral error theory.

Practice problem and solution

Hypothetical: three speakers react to 'Bribery is wrong.' P says it states a mind-independent fact and is true. Q says it states no fact and expresses disapproval. R says it purports to state a fact, but no such moral fact exists. Which speaker holds that the sentence is fact-stating yet false? Give the letter, then say how Q differs from R about truth.

R, the error theorist (the SEP lists Mackie among defenders). R accepts that the sentence purports to state a fact and says the fact is missing, so positive moral claims are false. Q, a noncognitivist, denies the sentence mainly states a fact, so it is not false in R's way. P is a realist.

Mental model: Realism, expressivism and error theory answer the semantic and ontological questions differently.

Common trap: Equating anti-realism with relativism.

8. Is, ought and the open question

Learning goal: Explain Hume's observation and Moore's open question argument without overstating either.

Hume wrote in the Treatise that he has often met an argument proceeding "in the ordinary way of reasoning" with propositions joined by "is, and is not", until "of a sudden" he finds only propositions joined with "an ought, or an ought not". He adds that this change is imperceptible but of the last consequence, because the new relation needs explanation. The SEP metaethics entry says there is substantial debate about what Hume meant and whether he was right.

The practical lesson is to look for the bridge premise. From "Most patients prefer option A" you cannot conclude "Doctors ought to offer A" without a premise such as "doctors ought to follow patient preferences". That premise may be reasonable. The point is that it must be stated and defended.

Moore, in Principia Ethica, argued that "good" cannot be defined. He wrote that if asked what good is, "my answer is that good is good, and that is the end of the matter," and he named the error of identifying good with another property the naturalistic fallacy. The SEP notes that quasi-realist expressivism remains among the most attractive ways of resisting the non-naturalist implications of the open question argument, and that others resist it differently. Both Hume and Moore are starting points for debate.

Worked example

"Most employees at the firm accept the monitoring policy, so the policy is morally acceptable." Name the missing kind of premise.

The premise states a fact about acceptance.
The conclusion states a moral verdict.
A premise linking acceptance to acceptability is missing.
That is a bridge premise, a normative claim, and it needs defence.

Practice problem and solution

"Many customers say they are satisfied, so the company has no duty to disclose the defect." What kind of premise is needed to make this valid? Two or three words.

The argument needs a normative premise linking satisfaction to the absence of a duty, such as "no complaint means no duty". That is a bridge premise.

Mental model: A moral conclusion needs at least one moral premise, and hidden ones should be stated.

Common trap: Treating a premise as obvious because it is common.

9. Disagreement and relativism

Learning goal: Separate descriptive disagreement from metaethical relativism and from tolerance.

People and societies disagree about morality. That is an observation and it needs evidence. The Stanford Encyclopedia of Philosophy entry on moral relativism has separate sections on descriptive moral relativism, on whether moral disagreements are rationally resolvable, on metaethical moral relativism and on relativism and tolerance. The structure shows that these are different questions.

The mistake to avoid is the jump from "societies disagree" to "no view is correct". A difference in beliefs does not by itself show that moral truth is relative. Beliefs about science also differ across places without making physics relative. An argument is needed for why moral disagreement is different, and it has to be made.

A third claim is normative: that we should tolerate those who differ. Tolerance is itself a moral position, so a relativist who endorses it appeals to a principle that is not obviously relative to a society. The SEP discusses this tension and mixed positions. The habit to build is to ask which of the three claims, descriptive, metaethical or normative, someone is making.

Worked example

In a hypothetical survey, Society A approves at 82%, B at 41% and C at 63%. Take the average of A and C and subtract B. Then say what the number does not show.

Average of A and C: (82 + 63) / 2 = 72.5.
Subtract B: 72.5 - 41 = 31.5 points.
The gap measures disagreement in belief.
It does not show that moral truth is relative.

Practice problem and solution

In an original hypothetical case, community P treats refusing a family-arranged marriage as wrong, while community Q treats it as permissible. A speaker argues: "The communities disagree, so each verdict is true relative to its community, and everyone ought to tolerate the other practice." What kind of claim is the premise that the communities disagree? Answer with the claim type.

The premise is descriptive disagreement: it reports a difference in moral beliefs. It does not by itself establish metaethical relativism, the claim that moral truth is relative to a framework. Nor does relativism alone establish a duty of tolerance; that is a separate normative claim.

Mental model: The numbers describe belief differences only.

Common trap: Reading approval rates as moral truth.

10. Applied ethics: giving and the duty to aid

Learning goal: Reconstruct Singer's argument, test each premise and calculate its demands under stated assumptions.

Peter Singer's 1972 article "Famine, Affluence, and Morality" argues that affluent people have strong duties to aid distant strangers. It starts from the premise that suffering and death from lack of food, shelter and medical care are bad. Its key principle is: "if it is in our power to prevent something bad from happening, without thereby sacrificing anything of comparable moral importance, we ought, morally, to do it."

He illustrates it with the pond case: "if I am walking past a shallow pond and see a child drowning in it, I ought to wade in and pull the child out. This will mean getting my clothes muddy, but this is insignificant, while the death of the child would presumably be a very bad thing." The argument then asks whether distance, or the number of others who could help, changes the duty.

The structure is premises and a conclusion, so each premise can be tested. A critic can deny the principle, deny that giving costs nothing of comparable importance, or deny that distance is irrelevant. The cost-per-life figures in the exercises below are hypothetical. Real charity-effectiveness estimates are contested and change over time.

Worked example

A person earning $60,000 gives 10% of income to a charity that, under a hypothetical estimate, saves one life per $4,800 given. How many lives does that gift save per year under the estimate?

Gift: 10% of $60,000 is $6,000.
Estimate: one life per $4,800.
Lives: 6,000 / 4,800 = 1.25.
The figure rests on a hypothetical cost estimate, not a measured outcome.

Practice problem and solution

Hypothetical estimates: charity A saves one life per $2,500 given and charity B saves one life per $4,000. A donor gives $4,000 a year. How many more lives does the gift save at A than at B, under these estimates (one or two decimals)? In one sentence, say which premise of Singer's argument the cost estimate bears on and why the estimate is a weak point.

At A: 4,000 / 2,500 = 1.6 lives. At B: 4,000 / 4,000 = 1.0 life. Difference: 1.6 - 1.0 = 0.6. The estimate bears on whether the gift costs nothing of comparable moral importance and how much good it does. Real charity-effectiveness estimates are contested and change over time, so a critic can attack the number.

Mental model: Every figure in a cost estimate is an assumption; check its source.

Common trap: Presenting a hypothetical number as a fact.

Course outline

Three questions: normative, metaethical and applied

Consequentialism and utilitarianism

Kantian ethics: universal law and humanity

Virtue ethics and the mean

Contractualism and justice from behind a veil

Doing and allowing, intention and foresight

Metaethics: realism, expressivism, error theory

Is, ought and the open question

Disagreement and relativism

Applied ethics: giving and the duty to aid

Sources and curriculum note

Complete course reading notes

1. Three questions: normative, metaethical and applied

Worked example

2. Consequentialism and utilitarianism

Worked example

3. Kantian ethics: universal law and humanity

Worked example

4. Virtue ethics and the mean

Worked example

5. Contractualism and justice from behind a veil

Worked example

6. Doing and allowing, intention and foresight

Worked example

7. Metaethics: realism, expressivism, error theory

Worked example

8. Is, ought and the open question

Worked example

9. Disagreement and relativism

Worked example

10. Applied ethics: giving and the duty to aid

Worked example