Which ethical dilemmas are real, and which are just temptations

Most things called ethical dilemmas are not dilemmas. They are cases where the right answer is obvious and doing it is expensive.

The two problems have nothing in common. A temptation is solved by resolve, or by removing the incentive that created it.

A genuine dilemma, two obligations that both bind, where honoring one means failing the other, cannot be solved by being a better person. It has to be decided, and something real gets broken either way.

Calling the first kind a dilemma is how organizations make a plain failure sound philosophically interesting.

What to take away

Telling them apart

Ask one question: if there were no cost to me or my organization, would the answer still be hard?

If the difficulty disappears once you remove the self-interest, it was a temptation. If it remains, because someone is still worse off whichever way you go, it is a dilemma.

A second check, useful when the first is unclear: can you describe the option you are drawn to in the plainest available language without it sounding bad? Temptations fail this test. Real dilemmas do not: both options sound defensible, which is exactly the problem.

The shapes a genuine conflict takes

Most real dilemmas fall into a small number of recurring structures. Recognizing the shape does not resolve the case, but it tells you which considerations are load-bearing.

The most consequential of these is the third, because it is the one that most easily disguises itself. Deferred, diffuse harm has no advocate in the room. The decision that produces it is always made by people who will not experience it.

A method that survives contact with a real decision

Frameworks fail in practice because they demand information nobody has under time pressure. This is a shorter sequence, meant to be usable in a meeting.

1. State both obligations, out loud, as obligations. Not "the commercial consideration" against "the ethical consideration", that framing has already decided the case by conceding that only one side is moral. Both sides are obligations: to shareholders, to employees, to a customer, to someone who has no relationship with you at all. Name them as such.

2. Name who bears each cost, specifically. Not "the business", not "the community". Which people, how many, how badly, and can they do anything about it. Costs that land on unnamed people get discounted automatically.

3. Ask what you are assuming that is convenient. Every difficult decision rests on a factual assumption that happens to favor the cheaper option. Find it. Then ask what it would take to check it, and whether you have time.

4. Look for the option that preserves part of both. It is usually not on the table because the framing was binary. Partial disclosure, delayed action with a firm date, exit with a long transition, proceeding while funding the mitigation. These are typically worse than either pure option on paper and better in the world.

5. Check the reversibility. If one option can be undone and the other cannot, that asymmetry matters more than most of the arguments. Prefer the recoverable path when the evidence is thin.

6. Decide who decides. Some of these should not be settled by the person in front of them, because their incentives are wrong or the consequences run wider than their authority. Escalating is a decision, not an evasion, but it has to be a decision with a deadline.

7. Write the reasoning down before you know how it turned out. Two paragraphs: what we knew, what we chose, what we accepted as the cost, what would have changed our mind. Written afterwards, this becomes a justification. Written in advance, it is the only thing that makes the next case easier.

Worked cases

The first three are hypothetical, constructed to sit on the line rather than to have an answer. The fourth is drawn from the public record.

The colleague who is struggling. A team member's performance has fallen sharply. In confidence, they tell you they are dealing with a serious family illness and ask you not to tell anyone. Their work is now being covered by two others who do not know why and are becoming resentful, and one of them is being marked down for missed deadlines that are not theirs.

The conflict is between a promise of confidence and a duty to people being harmed by keeping it. What is not in conflict: nobody has done anything wrong. The workable move is usually to go back to the person and expand the permission, not break it.

You can almost always disclose the fact of a temporary constraint without its content, and asking permission to do so is not a betrayal.

What is not workable is doing nothing and letting a third party absorb an unexplained cost. Where an illness or caring situation intersects with employment obligations, the handling may be legally constrained; involve someone who knows the rules where you are before deciding what can be said.

The forecast everyone wants. You are asked to produce a projection for a decision that has, in practice, already been made. Your honest number does not support it. The request is not to lie: it is to use a more optimistic set of assumptions, each individually arguable.

This looks like a dilemma but is mostly a temptation with good manners. The real conflict is your duty to give your honest professional view and your duty not to substitute your judgment for the decision-maker's.

Separate the two: present your number and the alternative assumption set labeled as theirs, then let the decision be made on visible grounds.

What you must not do is present their assumptions in your name. If the analysis will be relied on externally (by investors, lenders, or a regulator), the position is different again and there may be duties attached to it. For a US public company that includes the SEC's disclosure rules; the rest is a question for counsel.

The customer you should probably refuse. A prospective customer wants to buy your product for a use you find objectionable but which is lawful. Refusing costs a large contract and several jobs on your side. Your competitor will sell to them.

The competitor argument is the weakest and always the loudest. Whether someone else will do a thing has no bearing on whether you should. But the jobs on your side are a genuine obligation to real people.

What helps is narrowing the question from "do we approve of this use" to "would we be a proximate cause of the harm, and can we constrain it": contract terms, limits on functionality, refusal of support for particular deployments.

If the honest answer is that you cannot constrain it and the harm is serious, that is a decision for the level of the organization that can absorb it, not for a sales director with a quota.

The case that was not hypothetical: Volkswagen. In 2015 Volkswagen admitted that its diesel cars used software to detect an emissions test and cut pollution controls while the test ran. The company pleaded guilty in the US in 2017 and paid $4.3 billion in criminal and civil penalties, part of US penalties, settlements and owner remedies totalling more than $20 billion.

The shape is truth against harm, with the harm arriving later and landing on people with no voice in the room: the buyers, and everyone downwind of the tests.

The public record shows the device went in when meeting US nitrogen oxide limits looked impossible on the budget and schedule available. Nobody framed it as an ethics question. That is how a temptation becomes policy.

The cost of not deciding

While a decision sits open, harm continues. Positions harden, the option set narrows, and the eventual call gets made under worse conditions by people with less room.

US law prices some of that cost directly. A False Claims Act case brought by an employee can carry damages of three times the government's loss plus a per-claim penalty. The employee's share of the recovery is typically 15 to 30 percent.

Awards under the SEC whistleblower program run from 10 to 30 percent of sanctions above $1 million, and the largest exceed $100 million.

Delay has a clock in the other direction too. An employee who wants to bring a federal discrimination claim generally has 180 days to file with the EEOC, or 300 days in states with a work-sharing agreement. Silence past that date ends the claim rather than the problem.

Deferral also corrupts. The longer an organization lives with an unresolved problem, the more its members adjust. Original discomfort fades, and attrition closes the question rather than judgment.

If you are going to defer, make it explicit. What are we waiting for, by when, and who decides if it does not arrive? An open decision with a date is a decision. One without a date is a choice for the status quo that nobody has to own.

Newer shapes of the same problems

The structures above are old. What changes is where they show up.

Decisions made by systems. When a model declines someone, prices them, ranks them, or flags them, accuracy against fairness becomes a threshold that someone sets, and that choice distributes errors between groups. It is usually made by whoever is closest to the code, which is a governance problem before it is a technical one. New York City requires an annual bias audit of automated employment decision tools used in hiring, and Colorado's 2024 AI Act adds duties for high-risk systems when it takes effect in 2026. Where such a decision draws on data that moved between organizations, what must be settled first is set out in the UK regulator's data sharing code of practice.

Monitoring of employees. Remote and hybrid work has made extensive monitoring technically easy and cheap. The conflict is between a legitimate interest in knowing work is being done and an obligation not to surveil people in their homes. In the US, the NLRA protects concerted activity, including some discussion of pay and conditions on company systems. Several states require notice before monitoring starts, and intrusive surveillance can become evidence in an EEOC harassment claim. What an employer may monitor, and what it must tell people, is legally governed and varies; a worker's route to raise a concern about it may carry its own protection, as the UK's whistleblowing framework illustrates.

Generated content, and data that turns sensitive later. When machine-generated text, images, or voices reach customers, applicants, or readers, the guide is the plain-words test: would the recipient's reaction change if they knew? If yes, the non-disclosure is doing work, and work done by non-disclosure is the definition of the problem. Data collected innocuously can become dangerous with a legal change, a conflict, or a breach; for US public companies, the SEC's cybersecurity rule, in force since December 2023, requires a material incident on Form 8-K within four business days. Deletion schedules are where the retention conflict gets decided, usually by nobody in particular.

Signs you are rationalising

Worth knowing by heart, because they are recognizable from the inside if you are looking.

One of these is a prompt to check the facts again. Three at once usually means the answer is already known.

Related reading on this site: codes of conduct covers which of these a rule can settle in advance. The leadership ethics pillar covers who these decisions land on and why they get deferred.

The employee ethics pillar covers the version you face without institutional cover. Ethical frameworks covers the reasoning traditions people reach for when the shapes above do not resolve on their own.

Common questions

How do I tell a dilemma from a temptation?

A dilemma has two obligations that both have a claim on you. A temptation has one obligation and one thing you want. Most situations described as dilemmas are the second, and the giveaway is that only one option needs a justification.

What do I do when there is genuinely no good option?

Choose the one you can explain to the person who bears the cost, record why, and say plainly what you accepted as the price. Deciding badly and recording it honestly is recoverable; not deciding usually is not.

Does a framework resolve these?

No. It makes the disagreement between obligations explicit, which is what lets a group argue about the right thing. The choice remains a judgment, and it stays yours.