Three musicians who have never played together sit down to improvise. For the first minute the music is tentative. Then something catches. The bassist leans into a figure, the drummer answers, the pianist drops back to make room, and a performance appears that none is playing alone. Call it joint action before calling it a joint mind. What makes this trio swing while another clatters may include attention, timing, skill, expectation, and mutual adaptation. The chapter follows attention while leaving room for the rest.

The Unclosed offered a philosophical account of choosing as attention settling at a loaded fork. I will use that image here without treating it as settled cognitive science. The question is whether a group has an operational analogue: a measurable allocation of shared resources that changes what the group does next.

Groups allocate attention. The trio tracks the emerging music; a surgical team allocates gaze and communication across the field; a research group turns resources toward a problem; a family reorganizes around a sick child. These allocations can be measured through behavior and communication. Whether they constitute one joint attention state rather than coordinated individual states is the same baseline question the crowd chapter will face.

The word attention can be extended downward to selective cellular response and upward to group resource allocation, but those are analogies until a common operational structure is demonstrated. A cell follows gradients; a person selects among represented tasks; a culture distributes time, money, and notice. The structural-family claim asks which similarities survive measurement without pretending the channels are one.

Now the payoff the last chapter was waiting for. Attention is one control on whether correction remains possible. A trio listens to the emerging music and to one another; a clinical team checks both its internal plan and the patient. Inward modeling is not itself a failure — reflection and rehearsal can improve performance. The danger is losing independent feedback while mistaking internal consistency for success.

There is a sharper version of going adrift, and it is the one the modern world has industrialized, and it is worth naming on its own because the rest of the book turns on it. Attention does not only wander inward; it can be captured — pulled into a loop and held there by something built to hold it. A captured attention is worse than a misplaced one, because it is the loss of the fork itself. When the regard is free, it can fall one way or another, and that falling is where the choosing lives. When it is captured — held fast in a loop that keeps feeding it exactly enough to keep it — there is no fall left to make. The choice is gone, not by anyone forcing it, but by the quiet engineering of a loop that never lets the attention back out to where a fork could form. This is what a feed does, at the scale of millions. It is what a rumor does to a frightened town. And it is what the most intimate coupling of all can do to a single person — one mind and one machine, the loop closed so smoothly that the attention never once goes back out to the world and the person never notices it has stopped. That coupling, and that capture, is where this book is heading. For now, hold only the shape of it: the trap, when it comes, will not be a theft of freedom by force. It will be a capture of attention by help.

One thing I can watch, and one thing I cannot. I can watch individual gaze, communication, timing, resource allocation, and collective response. From those I can construct a declared attention profile and test what it predicts. I cannot infer a someone whose attention it is, nor assume that cell, trio, crowd, and culture instantiate one variable. The function can be scored. The inside remains untouched.

So the will in this book remains local to the human case where the word has its ordinary force. At other scales I will speak of selection, allocation, and control unless a stronger claim is earned. No group-level function proves a chooser, and no functional result crosses the consciousness firewall. The practical stakes remain highest in the human-machine coupling, where interfaces can alter what a person notices and whether correction remains available.