What the model gives back

Most of the argument for where the model belongs is stated in the negative: it does not diagnose, does not prescribe, does not decide. That restraint is right, but it can make the model sound like a tool defined by what it refuses to do. The positive case is simpler and more important. The model's job is to give the doctor back the one thing the system has been stealing from him: attention.

A doctor with forty-eight seconds and ninety patients is not short on knowledge. He is short on time and attention, spent before he reaches each patient by the history he has no room to take, the bag he has no time to read, the note he has to write while the next patient waits. Every one of those is work the model can do around the visit, so the doctor walks into the forty-eight seconds already knowing what matters and walks out without losing the next minute to paperwork.

Seen this way, the restraint and the usefulness are the same design. The model does everything around the decision precisely so the human has the attention to make the decision well. It is not competing with the doctor's judgment. It is clearing the ground so his judgment has room to operate, in a system that had crowded it out.

This is the quiet reason the boundary holds in practice and not just in policy. A doctor does not resent a tool that hands him a prepared patient and takes the paperwork off his hands; he relies on it, and relying on it, he stays the decider. The model that tried to decide for him would meet resistance and deserve it. The model that gives him back his attention is welcomed, and earns the right to sit underneath his judgment by making that judgment possible at the pace the country demands.