Okanohara: AI looks like an agent, so blame lands on the last button-presser

Okanohara's starting point is a philosophical one: how should we rethink the human role in the age of AI? He approaches responsibility, freedom, and judgment from the engineering side — how AI is designed and controlled — rather than from ethics alone. The excerpt published here comes from his Iwanami Shinsho book 'Human and AI.'

His central observation is that observers cannot directly verify another being's inner life, yet human societies have always functioned by assuming intent and responsibility anyway. That same mechanism, he argues, now applies automatically to AI: because it has a self-model, explains its reasons, and behaves consistently over time, society reads intent into it, forms trust or distrust, and hands it responsibility and rights. Okanohara stresses this is neither illusion nor misunderstanding — it is how social attribution has always worked.

The stakes he identifies are structural. When the entity that makes a judgment and the entity that bears its consequences are not the same, responsibility becomes ambiguous, and the burden can slide onto individuals and frontline staff. How the responsibility for AI-involved decisions should be allocated is a question the excerpt raises rather than resolves, and its practical weight is likely to depend on whether organizations accept his framing when they write their own internal rules.