When Not to Leave a Decision to Chance

Random decision tools are for choices where the outcome is acceptable either way. That condition is doing a lot of work, and most bad uses of a coin or a wheel come from ignoring it. Here is where the line sits, and what to do on the other side of it.

Try it now โ€” Spin a wheel โ†’

A three-question test

Before randomising a decision, ask three questions. Is the outcome hard or impossible to reverse? Does it affect someone who has not agreed to be bound by chance? Is it outside the authority of the person doing the drawing? If the answer to any of them is yes, the decision should not be delegated to a randomiser. The test is deliberately blunt. It is meant to catch the case where a coin looks like a light way to settle something that is not light at all.

Safety decisions

Anything where a bad outcome means injury or death is not a candidate. This includes medical treatment choices, whether to drive when tired, electrical work, medication dosing, and any decision about whether a piece of equipment is safe. Randomness can help you choose between two options that a competent person has already judged acceptable. It cannot substitute for that judgement, and using it to choose whether to take a risk confuses selection with assessment. The coin does not know the load rating.

Commitments to other people

You can randomise which of two acceptable ways to keep a commitment. You cannot randomise whether you keep it. Showing up, delivering work you were paid for, and caring for someone who depends on you are not lottery outcomes. The same applies to promises: a coin that decides whether you attend a wedding is not a fair way to resolve your ambivalence, it is a way to make the other person absorb your indecision. Decide the commitment, then use chance only for the details around it.

Money and legal consequences

Prices, contract terms, employment decisions, insurance questions and anything subject to an audit or a fiduciary duty need a reasoned justification. A randomiser supplies no justification, and in many settings it would be evidence of a process failure rather than a defence. Two exceptions are worth noting. Random allocation of a limited, neutral benefit โ€” a place in an oversubscribed course, a slot in a schedule โ€” is often accepted precisely because it is defensible. And a prize draw where entrants knowingly accept the mechanism is its own category, with its own rules.

When the bad outcome is unbounded

A coin between two low-cost outcomes is a reasonable tool. A coin between two very different lives is not. If one side of the flip could cost you your housing, your health or a relationship, the expected cost is dominated by that side, and a fifty per cent chance of it is not a balanced decision. Symmetric probability is not symmetric consequence. Before flipping, ask what the worst realistic outcome of each side is. If the answers are not in the same order of magnitude, stop.

Which, not whether

The most useful distinction is between choosing which option and deciding whether to accept a risk. Chance is good at the first: pick a restaurant, pick a topic, pick who goes first among people equally able to go first. Chance is bad at the second: whether to take the job, whether to make the investment, whether to accept a medical procedure. The first is a selection among acceptable alternatives. The second is a judgement about whether the alternatives are acceptable at all, and that judgement has to come first.

Randomness does not transfer responsibility

Using a coin does not transfer ownership of the outcome. You chose to flip, you chose the two sides, and you chose to be bound by the result. If the outcome is bad, the accountability sits with the person who set up the process. This matters practically, not just morally: a team that randomises a decision with real consequences and then blames the wheel has learned nothing and will make the same choice again. Owning the process is what makes future draws worth participating in.

What to do instead

If the three-question test fails, the decision needs a different tool. Write the criteria and score the options. Set a deadline and decide by it. Take the smallest reversible step and see what you learn. Ask the person affected. None of these is as satisfying as a spin, and all of them produce a decision you can defend. The discomfort of choosing deliberately is usually a sign that the decision matters, which is precisely when a randomiser is least appropriate.

Where chance belongs

Chance belongs in low-stakes choices, genuine ties, exploration, and the allocation of scarce but neutral goods. It belongs where the cost of deliberating exceeds the cost of a suboptimal outcome, and where the people affected accept the mechanism in advance. Those conditions cover more of daily life than they sound like: what to eat, who presents first, which idea to prototype. Keeping the boundary clear is what allows you to use chance freely inside it without pretending it works outside.

Related guides

Fairness in Random Picks ยท Coin Flips for Real Decisions

Liked this? Embed a decision wheel on your own site for free.