N° 013 2026.05.20

answerability

応答可能性

There is a question that pulls at the centre of all this generative-AI talk, and I think it has been mis-asked. The question being asked is: what makes an output authentic? Is it human? Is it stitched? Is the training data clean? Was the consent given?

These are reasonable questions. But they are downstream of a different one, which I would rather start from. The question I want: when something is made, is there a presence in it — someone answering for it, at every point where a choice was made?

This is what I have been calling, in my own head, answerability. It is not about volume, not about complexity, not even about originality in the usual sense. It is closer to a binary: at this point, is someone here?

By this measure, a single short line — a haiku, a phrase, a drawn stroke — can be a full expression, because someone is present in it at every choice. And conversely, a generative model outputting tens of thousands of words can be unanswered, because no one was present at the point where each word was chosen. Asking the model to generate is not the same as choosing.

This redraws the question of AI use. It is not "did you use AI / didn't you", which is a question about tooling. It is "where were you, while it was being made?" — a question about presence. Two people can use the same tool, and one of them is present at every decision while the other has delegated the decision itself. The first is making. The second is requesting.

I find this useful because it gives me a way to stay calm in the middle of the noise. I can use AI heavily and still be the maker, if I am present at every selection, every refusal, every adjustment. The volume of tool use is not the question. The position of my attention is.