Four lenses, one case: comparing ethical frameworks examples

Four lenses, one case: comparing ethical frameworks examples works only where the named frameworks disagree. If every way of reasoning about a case produces the same answer, you did not need four lenses. You needed five minutes.

The four frameworks used here are utilitarianism, Kantian deontology, virtue ethics and Rawlsian justice. The three examples below are chosen for divergence. Each is hypothetical, invented to sit on a line, and none describes anything that happened. Run them yourself before reading the commentary; the value is in noticing which framework you reach for first.

What to take away

A service has a defect that produces a wrong answer in a small fraction of uses.
A former employee, dismissed for conduct you consider serious, applies for a role elsewhere.
You hold detailed usage data collected to operate a service.
The common misuse of frameworks is retrospective: choose the answer, then select the framework that endorses it, and present the result as principled reasoning.

The four frameworks, by name

Utilitarianism.Jeremy Bentham and John Stuart Mill. Which option produces the better overall result, measured in welfare summed across everyone affected? Strength: it takes effects on real people seriously. Weakness: it will justify harming an individual for an aggregate gain, and it needs predictions nobody has. Rule utilitarianism asks which rule produces the best results.
Kantian deontology.Immanuel Kant set it out in the Groundwork of the Metaphysics of Morals (1785). Act only on a principle you could will as a universal law, and treat people as ends, never merely as means. Strength: it protects people who would otherwise be traded away. Weakness: it is stubborn when the rule was written for a case unlike this one.
Virtue ethics.Aristotle's Nicomachean Ethics asks what a person of practical wisdom would do, and what the choice makes of you. Elizabeth Anscombe's 1958 essay "Modern Moral Philosophy" revived the approach; Philippa Foot and Alasdair MacIntyre's After Virtue (1981) extended it. Strength: it captures the corrosion repeated small compromises cause. Weakness: it gives thin guidance on specifics.
Rawlsian justice.John Rawls, A Theory of Justice (1971). Test the rule behind a veil of ignorance, not knowing which person you will be. Rawls argued that position yields equal basic liberties, and allows inequalities only where they help the worst off. Strength: it forces you to name who bears the cost. Weakness: it stalls when interests cannot be reconciled.

Two related frameworks sit near these four. W. D. Ross's prima facie duties, in The Right and the Good (1930), bind unless a stronger duty overrides. T. M. Scanlon's contractualism, in What We Owe to Each Other (1998), asks whether your principle could be justified to each person affected. It is the sharpest version of the fairness test.

Case one: the flaw nobody has hit

A service has a defect that produces a wrong answer in a small fraction of uses. It has not caused a known harm. Fixing it requires taking the service down for a period that will disrupt every customer, including the large majority unaffected by the defect.

Utilitarianism weighs expected harm against certain disruption, and probably fixes it during a quiet window. The calculation is close, and it depends on numbers you can estimate.
Kantian deontology asks whether you promised correctness. If you did, the small frequency is not a defense. You are shipping something that does not do what you said.
Virtue ethics asks what an organization that knowingly leaves a wrong answer in place becomes, and how the next such decision goes once this one has been made.
Rawlsian fairness asks whether you could explain the delay to the specific person who receives the wrong answer next week. That is usually the question that ends the discussion.

The divergence is instructive: utilitarianism permits waiting, and the other three do not. When one framework is alone in permitting the cheaper option, that is worth noticing.

Case two: the reference

A former employee, dismissed for conduct you consider serious, applies for a role elsewhere. The prospective employer calls you. Your policy is to confirm dates and title only.

Utilitarianism pushes toward warning themthe next employer and their staff avoid a repeat.
Kantian deontology pulls the other way. You have obligations of fairness to a person who cannot answer what is said about them in a call they do not know is happening. You may also have obligations under your own policy and your agreements with them.
Virtue ethics is split, which is unusual and worth sitting with. Both discretion and candor are things an honest person values.
Rawlsian fairness asks whether the former employee could accept the rule you are applying. Most people would accept "we tell nobody anything" more readily than "we tell people when we think it is warranted," because the second gives you a discretion they cannot check.

Here the general rule beats the case-by-case judgment, even though case-by-case looks better in this instance. That is the recurring lesson of reference cases: a policy you apply uniformly protects people from your own bias in ways your judgment in the moment cannot.

What you can say about a former employee is also legally constrained in most places, and the constraints differ. This is a conversation to have with counsel before it is one to have with a caller. Where the person has raised a concern about wrongdoing, separate protections may apply, of the kind described in the UK's whistleblowing framework.

Case three: the useful data

You hold detailed usage data collected to operate a service. Analyzing it would let you spot customers heading toward a serious problem and intervene, probably helpfully. The customers agreed to collection for service operation. They did not agree to this.

Utilitarianism says the intervention helps people, so do it.
Kantian deontology says the permission is the permission, and a benefit does not retroactively widen it.
Virtue ethics asks whether you are the kind of organization that decides what is good for people using information they gave you for something else.
Rawlsian fairness offers the way throughask them. If the use is good for them, most will agree, and the ones who would not are exactly the people whose refusal you should be honoring.

This is the pattern where a fourth option beats all four analyses. When consent is the missing element, obtaining it usually costs less than the argument about whether you need it.

Secondary uses of personal data are also regulated, differently in different places, so the legal question runs alongside the ethical one rather than after it. Guidance on what makes consent valid is a useful discipline here even outside its jurisdiction, because it forces the question of what the original permission actually covered.

Where the lenses actually split

Where the four lenses split

Situation type

Small harm to few: Duties, fairness
Rule fits badly: Consequences
One-off compromise: Character
Cannot explain to affected: Fairness
Benefit without asking: Fairness, duties

Dissenting lens

Small harm to few: People being traded
Rule fits badly: The actual outcome
One-off compromise: What repetition does
Cannot explain to affected: Their standing as parties
Benefit without asking: Their right to decide

What it protects

Small harm to few
Rule fits badly
One-off compromise
Cannot explain to affected
Benefit without asking

Which framework dissents

Small harm to few, large benefit to many: Kantian deontology, Rawlsian fairness
A rule that fits badly here: Utilitarianism
A one-off compromise "just this once": Virtue ethics
An option you cannot explain to those affected: Rawlsian fairness
Acting for someone's benefit without asking: Rawlsian fairness, Kantian deontology

What the dissent is protecting

Small harm to few, large benefit to many: The people being traded
A rule that fits badly here: The actual outcome
A one-off compromise "just this once": What repetition does to you
An option you cannot explain to those affected: Their standing as parties, not obstacles
Acting for someone's benefit without asking: Their right to decide

The failure mode

The common misuse of frameworks is retrospective: pick the answer, then choose the framework that endorses it, and call it principled reasoning.

It is recognizable. Someone who reaches for utilitarianism when the rule is inconvenient, and for Kant when the consequences are inconvenient, is not reasoning. They are decorating.

Two defenses. First, apply all four, briefly, in writing, before deciding. It takes ten minutes, and the disagreement between them is the whole product. Second, ask what this reasoning would commit you to in the next case. If you would not accept the same argument when it points the other way, you have not made an argument.

Related reading on this site: ethical frameworks sets out the four lenses, ethical dilemmas covers the underlying shapes these cases take, codes of conduct covers what a written rule can settle in advance, and leadership ethics covers who decides.

Common questions

Which frameworks are the four lenses?

Utilitarianism comes from Bentham and Mill and weighs outcomes. Kantian deontology comes from Kant and weighs duties. Virtue ethics comes from Aristotle and weighs character. Rawlsian justice comes from Rawls and asks what people could accept behind a veil of ignorance. Ross's prima facie duties and Scanlon's contractualism sit near the fairness lens.

Why are all three cases invented?

Because a real case brings its own disputed facts, and the point here is the reasoning rather than the verdict. Each is built to sit on a line so that the four frameworks disagree.

What if the four frameworks all agree?

Then you did not need them, and the decision was not difficult. The value is entirely in the cases where they diverge, which is why these three were chosen.

Is one framework more reliable than the others?

No, and treating one as the default is how retrospective justification starts. The useful discipline is applying all four briefly, in writing, before deciding rather than after.