Human Judgment and Responsibility in Multi-Agent Systems

Human Judgment and Responsibility in Multi-Agent Systems

Xufen Tu · Independent Interdisciplinary Researcher Public Citation Version: v1.0

When several automated systems contribute to an outcome, their combined involvement can make responsibility harder to locate. People affected by the outcome may know that many participants were involved without knowing who remains answerable to them.

My research observation is that more participants do not necessarily bring clearer accountability. Even when each contribution appears useful, the consequences of the overall result may be difficult to explain or address. That difficulty matters most to people who must live with those consequences.

My current position is that distributed participation does not, by itself, remove human or institutional responsibility. An outcome that affects people still raises a public question about who can answer for it. The number or capability of the systems involved cannot settle that question.

This position does not assign responsibility in any particular case. It does not prescribe how automated systems should be organized or how disagreements among them should be resolved. Different circumstances call for careful human judgment, and human judgment can itself be mistaken.

The public concern is whether affected people can identify someone answerable for a consequential outcome, even when many automated contributions are involved.