Control and accountability in hybrid AI systems

The text analyzes the problem of responsibility in human-AI hybrid systems, contrasting nominal control (the formal presence of a human) with actual control (meaningful human control). The author indicates that the ethics of such systems depend on their architecture, rather than solely on the morality of individual components. Key importance is placed on the concepts of tracking (the system's responsiveness to human moral reasons and context) and tracing (the ability to attribute an action to a specific agent). This approach aims to resolve the problem of the responsibility gap, where technical complexity disperses knowledge and agency, rendering the human merely an "administrative" obstacle in the decision-making process rather than a real supervisor.