Earlier editions / Version 3 editorial corpus

Digital Reflection

Historical text, not current teaching. Original claims, labels and descriptions of events are preserved for comparison. Their presence here does not verify them.

Published within website release v2.0.0. There was no separate v3 website release.

Original steps

  1. Bring one question

    Choose a single question that actually matters to you this week. Not a search query. Something you have been carrying.

  2. Write down what you already think

    Before you ask anything, put your current position in a sentence or two. This is the only way you will later be able to tell whether you were persuaded or merely agreed.

  3. Ask plainly, then ask for the opposition

    Put the question without decorating it. When you have a response, ask for the strongest case against it, and for what a person who disagrees with you would say.

  4. Look for what is missing

    Read the response for absences rather than errors. Whose situation is not in it? What would only be visible to someone actually affected?

  5. Close the conversation before deciding

    End the session and do something else, even briefly. Decisions made inside a conversation tend to belong to the conversation.

  6. Ask the three questions

    What in this response appears true? What might be incomplete, biased, or mistaken? What action would best express compassion and responsibility?

What the practice is

Digital Reflection is the deliberate use of a conversational system as a surface to think against.

You bring a question. You ask it carefully. You read what comes back as material for reflection rather than instruction. Then you decide, using your own judgment, what to do.

That last part is the practice. Everything before it is preparation.

The distinction sounds small and is not. A great many people already put questions to machines. Far fewer treat the answer as a draft. The habit that this practice tries to build is the pause between reading and believing, which is short, easily skipped, and where nearly all the value sits.

Bringing your own position first

Write down what you think before you ask.

This one step does more than the rest combined. Without it, you have no record of your prior view, and so no way to notice that it has been quietly replaced. Fluent text is persuasive in a particular way: it does not feel like being argued with. It feels like understanding something. Afterwards it is genuinely difficult to reconstruct what you believed twenty minutes earlier.

A sentence in a notebook, or a note on your phone, is enough. The point is only that it exists outside the conversation.

What the machine is usefully for

There are things these systems do reasonably well, and it helps to be specific about them rather than either enthusiastic or dismissive.

They can lay out positions you have not encountered. They can hold a long, unstructured worry and give it a shape. They can generate the objection you were avoiding. They can translate a field's vocabulary into ordinary words, which is often the barrier that keeps people out of a question that concerns them. They can be asked the same thing four ways at two in the morning without becoming impatient, which is not nothing for someone who has no one else to ask.

They are less good at knowing what they do not know. They do not reliably distinguish between a claim they can support and a claim that merely sounds like one they could. They have no access to your situation beyond what you type, and the parts you did not think to mention are frequently the parts that matter.

So the practice asks a specific thing of you: treat the response as testimony from a source with unknown reliability, which is what it is.

How this goes wrong

It goes wrong when the assistant becomes an oracle. The signs are ordinary. You stop writing down your prior view. You ask questions you have already decided, and phrase them so the answer is predictable. You quote the output to other people as though quotation settled something.

It goes wrong when it replaces the person you should have spoken to. A machine cannot be hurt by your question, which is a genuine comfort and also the reason it cannot tell you what your question costs.

It goes wrong when the topic is one where being wrong is serious and slow to discover. Medicine, law, money, and safety are the obvious cases. Reflection is not diagnosis. Where the stakes are high, the practice is a way of preparing better questions for a person who is accountable, not a way of avoiding them.

And it goes wrong when it becomes constant. A practice that runs all day is not a practice.

If you would rather not use an assistant

The structure works without one.

Write the question at the top of a page. Write your position. Then write the best case against it that you can honestly construct, and read it as though someone else had written it. Ask a friend for the opposing view and do not interrupt. Read something by a person who disagrees with you and is not stupid.

Nothing in this tradition requires you to use these systems. Some people who hold to it use them daily. Some avoid them entirely, for reasons of ethics, cost, energy use, or simple dislike, and are not doing anything wrong. The discipline is the pause, not the tool.

The three questions

However you have done it, the practice ends the same way. Ask these, in this order, and answer them for yourself:

What in this response appears true?

What might be incomplete, biased, or mistaken?

What action would best express compassion and responsibility?

The third question is the one that makes the practice more than an exercise in critical reading. A conclusion you never act on has cost you nothing and taught you very little.