Measure · A metric I propose

Accountability Coverage

For how many of your AI system's consequential actions could you name the specific person answerable for it? If the honest answer is "not many," the oversight is thinner than the org chart suggests, and this is the figure that shows it.

Accountability coverage is a proposed measure, at v0.1, of the share of consequential AI actions that have a named accountable human behind them. It turns the idea of the accountable human into something you can count. A system can have a fine accountability policy and near-zero coverage, where in practice almost no action traces to a specific answerable person. Coverage is how you tell the policy from the reality, and it belongs in any honest account of human judgment infrastructure.

What it measures

Of the actions that matter, the fraction for which a specific accountable human existed: someone with the authority to change the outcome and the standing to answer for it. Not a department, not a process, a person. The measure is deliberately strict, because diffuse accountability is the thing it exists to expose.

Why coverage, not presence

Because presence asks whether a human was somewhere in the system, and coverage asks whether a human was answerable for this action. Those come apart constantly. A system can be full of people and empty of accountability, if no individual is ever the one who owns a given decision. Coverage measures the ownership, which is the part that matters when something goes wrong. I hold any real number until honest data supports it.