Version Responsibility in AI-Era Knowledge Production and the Reconstruction of Authorship

Abstract

The widespread application of generative artificial intelligence is profoundly reconstructing the paradigm of knowledge production, shifting it from the traditional model of “a single human author and a single original text” toward a model of human–machine collaboration. In this process, traditional authorship is substantively deconstructed because its internal unity of intentionality and responsibility is interrupted, which in turn produces black-boxed knowledge production and accountability dilemmas. To address this problem, this paper introduces the concept of version responsibility. The concept suspends the metaphysical debate over AI originality and redirects governance toward the question of who bears ultimate social and ethical responsibility for a specific public version of knowledge. Drawing on the RID cognitive dynamics model in Knowing and Saying: An Ontological Investigation of Human Cognition , the paper explains version responsibility through three dimensions: factual verification, transparent disclosure, and consequence accountability. It argues that the philosophical essence of version responsibility is to defend the priority of human beings in the dimension of problem pressure, or the D-dimension. On this basis, the paper proposes an institutional principle of functional outsourcing and responsibility centralization, and develops governance frameworks for high-risk fields such as academic publishing, medical decision support, and judicial sentencing. The goal is to rebuild a truthful and reliable public knowledge ecology while still releasing the cognitive dividends of AI.