Inattentive Robot: When Should One Body Give Timely Attention to Many Masters?
When should an AI assistant intervene? The TimelyAI workshop frames this as a question of opportune assistance: when generative AI should proactively act, interrupt, or remain silent. This paper shifts that question from knowledge work to the domestic robot: a shared embodied AI with one body, many users, and conflicting temporal claims. Unlike a chatbot, a household robot cannot instantiate a separate assistant for every family member. It can only stand in one place, move toward one emergency, and use one pair of hands at a time. I introduce the inattentive robot: not a robot that lacks perception, but one that cannot attend to everyone at once. In a family home, the robot may be asked to cook breakfast, tutor a child, protect an elder, comfort a teenager, monitor a baby, care for a pet, or rescue a stranger in the street. Each request arrives with its own sense of timeliness. This provocation argues that domestic AI attention is not merely a scheduling problem but a political problem of priority, care, ownership, vulnerability, and interruption. We propose scenario-based attention conflict tests as a design method for exposing the implicit protocols by which shared robots decide whose time matters first.
When should an AI assistant intervene? The TimelyAI workshop frames this as a question of opportune assistance: when generative AI should proactively act, interrupt, or remain silent. This paper shifts that question from knowledge work to the domestic robot: a shared embodied AI with one body, many users, and conflicting temporal claims. Unlike a chatbot, a household robot cannot instantiate a separate assistant for every family member. It can only stand in one place, move toward one emergency, and use one pair of hands at a time. I introduce the inattentive robot: not a robot that lacks perception, but one that cannot attend to everyone at once. In a family home, the robot may be asked to cook breakfast, tutor a child, protect an elder, comfort a teenager, monitor a baby, care for a pet, or rescue a stranger in the street. Each request arrives with its own sense of timeliness. This provocation argues that domestic AI attention is not merely a scheduling problem but a political problem of priority, care, ownership, vulnerability, and interruption. We propose scenario-based attention conflict tests as a design method for exposing the implicit protocols by which shared robots decide whose time matters first.
