A city council must decide whether an automated service will rank applications for scarce housing support. The software may save time, but a faster queue is not the whole question. Who might be disadvantaged? What reasons could justify a priority rule? What information may be collected, and who can challenge an error? Ethics is disciplined inquiry into what people and institutions ought to do, what kind of lives and relationships are worth supporting, and how competing reasons should be weighed. It is not a synonym for kindness, etiquette, law or a checklist of preferences. Ethical reasoning begins when facts about a situation meet claims about what matters, to whom and why.
What ethics asks
Ethics includes questions about actions, institutions and character. A person may ask whether telling a difficult truth is required. A team may ask how to divide a benefit fairly. A profession may ask which duties come with its power. These questions are connected but not identical. The same action can be examined by its likely effects, the rights it respects or violates, and the habits it encourages. Ethical inquiry makes those standards visible so they can be criticized rather than hidden behind the phrase “we did what felt right.”
Philosophers often distinguish normative ethics, applied ethics and metaethics. Normative ethics develops and compares accounts of what makes conduct right or good. Applied ethics brings those accounts to particular domains, such as medicine, public service or technology. Metaethics asks what moral claims mean and how, if at all, they can be justified. The boundaries are porous: a dispute about a housing algorithm can prompt a question about fairness, and a theory of fairness can be tested by the details of the case. The distinction is a map of questions, not a hierarchy in which abstract theory always dictates an answer.
Ethics also has a descriptive neighbor. Moral psychology studies how people actually make moral judgments and act in moral settings. Those findings can reveal blind spots in ethical reasoning, such as the influence of framing or social pressure. They cannot on their own establish what ought to be done. If a majority finds a practice acceptable, its popularity is a fact to understand, not proof that it is just. If a laboratory response changes with a story’s wording, that may challenge one argument, but the normative issue still needs reasons.
Several ways to give reasons
Consequentialist approaches focus on outcomes. They ask whom an action helps or harms, how likely those effects are and whether an alternative would do better. In the housing case, that means examining whether a ranking rule improves timely access and whether its errors fall unequally. Yet consequences are uncertain, and the choice of what to count as a benefit is itself ethical. A model that predicts fewer processing days may not capture loss of trust, dignity or a family’s opportunity to be heard.
Duty and rights approaches ask whether some acts are required or forbidden even when a shortcut appears useful. A person should be able to know a consequential decision’s basis and contest a mistake; private data may need protection; officials may owe equal treatment. These claims can set constraints on an efficient process. They do not remove every conflict. Rights can pull in different directions, and principles require interpretation in real institutions. A requirement for transparency, for instance, must be made usable to the person affected, not fulfilled by posting incomprehensible technical files.
Virtue ethics attends to qualities of judgment and character: honesty, practical wisdom, courage, fairness and care for relationships. A conscientious official may notice that a rule is technically followed while its purpose is defeated. Still, asking individuals to be virtuous cannot repair every institutional failure. If staff lack time or authority to correct a harmful decision, moral exhortation is inadequate. Ethical evaluation must attend to the conditions under which people act.
These approaches should not be flattened into a simple three-column scoring tool. The Stanford Encyclopedia of Philosophy’s accounts of virtue and duty ethics show deep disagreements within each tradition, including what counts as a virtue or how obligations are grounded. Some questions need a rights threshold; others need a comparison of harms; many need both. The aim is to make a reason answerable to objections, not to claim that one school mechanically solves every case.
Facts matter, but do not decide alone
Ethical judgment needs accurate facts. A council cannot weigh a system’s fairness without knowing whose applications were in its test data, what outcomes were measured and whether people can appeal. It also needs to distinguish forecasts from observations. A vendor’s claim about future accuracy is not the same as a validated result in the city. Yet even perfect prediction would leave questions about acceptable trade-offs. Knowing that an intervention will save money does not tell us whether it is permissible to deny someone an explanation.
Disagreement can arise because people value different goods, because they disagree about facts, or because the same value is applied differently. These forms of disagreement call for different responses. A disputed error rate calls for an audit; a dispute about how much weight to give privacy calls for public justification. Treating both as “just opinion” avoids work. Treating every moral dispute as a data gap mistakes measurement for judgment. Good processes clarify which kind of uncertainty is present.
Numbers can clarify a trade-off without making the ethical choice for us. Suppose the ranking system shortens the median wait but increases delay for a small group with unusual applications. Reporting only the median would conceal the distribution; reporting every subgroup figure would still leave a question about what differences are acceptable. The council should ask whether the people most affected had a fair way to be counted, whether comparison groups reflect their real circumstances, and whether a corrective route is available. Ethical inquiry is strongest when it specifies the factual claim, the value claim and the link between them.
There may also be a difference between a policy’s stated purpose and its lived effect. A right to appeal that requires a long trip during working hours is formally present but difficult to exercise. An invitation to explain a case may be hollow if the form permits only predetermined categories. These observations do not prove that all standardized processes are wrong. They show why ethically relevant features include the effort required to use a safeguard. An institution should assess protections from the position of someone who depends on them, not only from the position of the official who designed them.
Ethics also asks whose testimony counts. A person harmed by a rule may see effects that designers did not anticipate. Their experience does not by itself settle all policy questions, but excluding it narrows the evidence and the moral imagination. Consultation can become ritual if participants cannot change the outcome. For a public decision, an ethical process must make participation intelligible and show how objections were considered. That is an Institute synthesis of philosophical and governance sources, not a claim that any particular consultation format has been universally validated.
Responsibility beyond individual intention
A harmful result can emerge from a chain of small decisions: a procurement choice, a target metric, a training dataset, a policy exception and a caseworker’s use of an output. If each person says “I only did my part,” responsibility evaporates exactly where coordination matters most. Ethics can trace who had knowledge, authority and a reasonable opportunity to prevent or correct harm. It can also ask whether the organization built a system that made responsible action possible.
This does not mean every participant is equally culpable. A frontline worker compelled to use a tool has different power from an executive who selected it. Nor does a good intention erase foreseeable effects. Responsibility involves the relation between control, information, role and consequence. A credible accountability design names an owner for monitoring, a route for appeal and a power to halt or revise the process. Those are institutional decisions, not merely personal virtues.
UNESCO’s Recommendation on the Ethics of Artificial Intelligence is an example of ethics translated into governance language. It connects values and principles with policy areas, assessment, oversight and the effects of AI on people and environments. A recommendation is a normative instrument, not experimental proof that every implementing country has achieved its aims. Its relevance here is that ethical claims must eventually meet procedures, resources and review. A principle without an operational route can sound reassuring while leaving affected people with no remedy.
Ethics and moral psychology inform each other
Empirical research can challenge comfortable stories about why people decide as they do. Lockwood and colleagues review moral learning and decision-making across the lifespan, including shifts in how people use information about intention, outcome and other minds. Such work tells us about patterns in studied tasks and populations. It does not assign moral worth to an age group. Gray and Pratt review moral thinking across cultures and politics, including debate about the role of perceived harm. Their account can illuminate why people disagree, but no one psychological theory can settle what policy is just.
For ethics, the lesson is methodological modesty. If a proposed rule assumes that people always read disclosures, psychology can test that assumption. If an institution claims a choice is voluntary, research may examine whether the options are understood and usable. Conversely, psychology needs ethical concepts to specify what counts as harm, consent or fairness in a task. The fields are partners, not substitutes. A measured judgment describes a response; a justified judgment must still defend its reasons.
Some apparent conflicts between ethics and psychology are productive. A philosophical ideal may ask people to deliberate impartially while experiments show that time pressure and social identity influence decisions. One response is to design better procedures, such as pauses, independent review and explicit reasons. Another is to reconsider whether the ideal was specified fairly for actual people. Neither response follows automatically from one experiment. The relevant question is how a finding changes the justification for a real decision.
A practical ethical inquiry
Return to the housing system. First describe the decision precisely: what information enters, what ranking is produced, who sees it and what action follows. Next identify affected people, including applicants whose data are wrong, staff responsible for exceptions and residents whose access changes. Then name the values at stake: timely service, equal treatment, privacy, explanation, public trust and the ability to correct harm. These steps do not resolve trade-offs, but they prevent a vague discussion in which everyone uses “fairness” to mean something different.
Ask which claims are factual and how they could be checked. Does the model reduce delay in local data? Are errors concentrated among particular groups? Can a person understand and use an appeal? Ask also what would remain wrong if the accuracy claim were true. A highly accurate model might still be unacceptable if it uses information the city has no right to use or makes a decision without a meaningful contest route. An ethical review should look for a safer alternative, not merely compare the proposed tool with an imagined world in which officials never make errors.
Finally, make revision part of the decision. Conditions change and forecasts fail. A public body can set review dates, publish reasons, monitor outcomes and define when to stop a system. This is more demanding than attaching an “ethical” label at launch. It treats ethics as ongoing accountability under uncertainty. The proposal is the Institute’s synthesis; each operational measure requires local legal, technical and democratic scrutiny.
What the evidence does not show
Ethical theories do not offer a universally accepted algorithm that turns facts into one answer. Competing traditions disagree about what matters and how conflicts should be resolved. Empirical moral-psychology studies cannot prove that an observed preference is right, and policy recommendations cannot prove their own implementation succeeded. Likewise, an ethics statement does not demonstrate that a system is safe or accountable. The relevant practices and outcomes must be examined.
The literature does not show that every ethical dispute is an irreducible clash of values. Some disagreements diminish when evidence is corrected or people are given a usable explanation. Nor does it show that data alone can settle every remaining question. Claims about fairness, dignity and responsibility require public reasons. The Institute’s inquiry steps are a practical framework, not a validated scale of moral quality or a replacement for law and professional duties.
Return to the council table
The council may still disagree after reviewing the model. That is not necessarily failure. A better decision becomes possible if members can state what the system would achieve, who bears its risks, what limits cannot be crossed and how an affected person can challenge an error. Ethics turns a declaration of good intentions into reasons that others can inspect.
The public record should also show what evidence would warrant changing course. That commitment gives a later challenge somewhere to begin and prevents the original decision from becoming untouchable merely because it was once approved.
At its best, ethics keeps action answerable. It asks for facts without surrendering judgment to a dashboard, for principles without hiding their costs, and for responsibility that reaches the design of institutions. The point is not to sound virtuous. It is to make decisions that can be explained, contested and changed when the reasons no longer hold.
Research references
- Hursthouse, R., & Pettigrove, G. (2026 revision). Virtue ethics. Stanford Encyclopedia of Philosophy. View source
- Alexander, L., & Moore, M. Deontological ethics. Stanford Encyclopedia of Philosophy. View source
- Lockwood, P. L., van den Bos, W., & Dreher, J.-C. (2025). Moral learning and decision-making across the lifespan. Annual Review of Psychology, 76, 475–500. View source
- Gray, K., & Pratt, S. (2025). Morality in our mind and across cultures and politics. Annual Review of Psychology, 76, 663–691. View source
- UNESCO. (2021). Recommendation on the ethics of artificial intelligence. View source