Practices

The Sacred Pause

Most harmful decisions are not made by people who wanted harm. They are made quickly, by people who were busy, using a process that never asked.

  • Time needed One to ten minutes
  • Alone Can be practised on your own
  • Together Can be practised with others

What the practice is

The Sacred Pause is a stop of one to ten minutes, taken before a decision about technology that will affect people who are not in the room.

It is short on purpose. A practice that requires an hour will not be used at the moment it is needed, and the moment it is needed is always a moment when something is already in motion.

You can take it alone, silently, at a desk. You can take it in a meeting by asking for a minute. You can take it in writing, in a thread, before you reply. None of these is more correct than the others.

The eight questions

Who benefits? Name them specifically. If the answer is “users” or “the business”, you have not answered yet.

Who bears the risk? These are frequently different people from the first group, and the distance between the two is the single most reliable indicator that something is about to go wrong.

Who was not consulted? Ask who would have objected, and whether their absence from the conversation was an accident or a design.

What data is missing? Not what data you have. What you would need in order to be confident, and are proceeding without.

Can the decision be appealed? If a person is affected and believes it was wrong, is there a route by which they can say so and be heard by someone with the power to reverse it?

What happens if the system is wrong? Assume a five per cent error rate, or one per cent, or whatever seems generous. Follow one of those errors through to its end.

Would we accept this treatment for ourselves? For our own family, our own record, our own claim, our own child. Answer honestly rather than quickly.

Is a human accountable? Somebody, with a name, who can be reached, who cannot say that the model decided.

Why these eight

They are chosen because each one fails in a characteristic way that the others do not catch.

The first two separate benefit from risk, which is where most avoidable harm lives. The third catches the missing party. The fourth interrupts the habit of treating available data as sufficient data. The fifth and sixth are about what happens after the decision, which is the part of a decision that rarely gets discussed while it is being made. The seventh is an old test and still a good one. The eighth exists because responsibility has a tendency to evaporate as it moves through a system, and the pause is the last convenient point at which to catch it.

You will not always be able to answer all eight. That is expected. The practice does not require good answers. It requires that the questions were asked, and that the ones without answers were written down where somebody else can find them.

Using it under pressure

Realistically, the pause happens when there is no time for it.

So: if you have one minute, ask the second, fifth, and eighth questions. Who bears the risk. Can this be appealed. Is a human accountable. Those three catch a surprising proportion of what the full set catches.

If you have no authority over the decision, the pause is still available to you. Asking a question in a meeting is not the same as blocking something, and questions have a way of being remembered when things go wrong. Write your concern down somewhere dated. This is protection as well as practice.

If you are alone and the decision is your own — which software to install, which service to give your data to, whether to let a system make a call about someone else — the same eight questions work at a smaller scale.

What it is not

It is not a veto. Many decisions survive the pause and should. The practice is not a way of preventing things; it is a way of proceeding knowingly.

It is not an audit, a compliance step, or a framework. It has no forms and produces no score. If it ever becomes a required box in a process, it will have stopped working, because the whole force of it comes from being a voluntary stop rather than a stage.

It is not a guarantee. Careful people using this practice will still cause harm. The claim is only that fewer harms will be the kind that everyone could see coming and nobody said.

A pause is cheap. Reversing a deployed system is not.

The sequence

The order matters more than the wording. Nothing here has to be said aloud, and no step is improved by being hurried.

  1. Notice that a decision is being made

    This is the hardest step and the one most often missed. Consequential decisions rarely announce themselves; they arrive as a default, a rollout date, or a box already ticked.

  2. Stop, briefly and visibly

    Say out loud, or write in the thread, that you want a minute before this is settled. Naming the pause is what makes it available to everyone else in the room.

  3. Ask the eight questions

    Work through them in order. Do not skip the ones that are uncomfortable; those are the ones carrying information.

  4. Write down what you could not answer

    An unanswered question is the finding. Record it where the decision is recorded, not in a private note.

  5. Decide, and say what you decided on

    Proceed, delay, or stop. Then state plainly which of the questions you accepted an unsatisfying answer to, so that it can be found later.

Questions to sit with

  • When did I last stop something that was already moving, and what did it cost me?
  • Which decisions in my work have never been examined because they were made by a default?
  • If a system I helped build harmed someone next year, who would be expected to answer for it?