Main idea

You expect an assistant that drafts an article, summarizes it, rewrites it and criticizes it to lighten the work of deciding. The opposite is what happens.

It increases the number of proposals, objections, variants and possible paths. Each one has to be evaluated: which objection deserves to be handled, which suggestion is too generic, which simplification betrays the idea, which nuance weighs the text down, which contradiction has to stay visible, which source is enough, which risk is acceptable, which trade-off genuinely commits the work.

The circulation of possibilities speeds up; the sorting stays whole — and it now bears on more objects.

Why it matters

This is counter-intuitive, and the reverse error is costly: believing the load falls leads you to drop your guard at exactly the moment it should rise.

It also gives a non-moral reason to keep human judgment. It is not a matter of principle about what a machine should or shouldn't do: it is that sorting becomes the bottleneck, and it hasn't been moved.

Nuances and limits

The increase isn't indefinitely bearable. Past a certain volume, you no longer judge: you accept by default, which brings you back to compliance.

And some help really does reduce the load, when it filters instead of proposing. The observation holds for the kind that widens.

Open questions

• From what volume of proposals does judgment stop being genuinely exercised?