Abstract
<p>Knowledge work is learning to hand itself over. Drafting, analysis, computation, even the role of a practice opponent can now be delegated to artificial intelligence, and the delegation works: the results come back, often better and faster than one's own. This paper asks what, in that handover, does not move. Drawing on Bernard Suits's analysis of games, I separate the achieved state of an activity from the value of participating in it and defend a non-transferability theorem: parts of an activity can be delegated, but no agent, human or artificial, can take over one's participation itself. When another realises the goal state on my behalf, the fact that I undertook the constitutive constraints and pursued the goal through them does not thereby come into existence. AI can deliver the state of affairs; it cannot deliver the fact of participation. At bottom, then, the central claim is not a claim about AI at all: the theorem is agent-relative, and holds whatever machines become — achievers, players, moral patients, or none of these. What it leaves open is the human question. Whether one can keep playing, lucidly, in full knowledge of the machine's strength, depends not only on individual temperament but on whether a culture grants dignity to those who do unnecessary things in earnest. Japanese professional shogi, which met machine superiority a decade ahead of other knowledge domains and kept its play, is examined as a completed historical case — public evidence that the question in this paper's title has already been answered, once, in the affirmative.</p>