Whose Thought Was That? Authorship, Perceived Agency, and the Design of Wiser Human-AI Relations

We explore the notion that AI systems now have the capacity to produce written work almost autonomously, and the resulting rift this has caused concerning how authorship is attributed. People do not feel ownership of the work that is produced. But when asked whether it belongs to LLM, they say no. We argue and compare this to other systems and try to pinpoint the reason why authorship simply disappears in the gap found between user and machine, and what that gap says about AI, the self and the boundary between them. We point out that what matters less is the magnitude of the generation, but rather the appearance of agency, and that once agency is experienced, we rush to reify a boundary. We further argue that this boundary is imputed rather than discovered. We further highlight that no taxonomy can measure the precise quantities submission guidelines will demand, because it does not exist and will only get more challenging with time. We propose instead a criterion for designing human-AI relations, one judged by what a person can do afterwards, paying close attention as to whether that qualifies with or without the model.