Who Lives With the Choice? AI-Mediated Deliberation, Asymmetric Historical Incorporation, and Recursive Human Agency

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

Generative artificial intelligence is moving from information retrieval and task execution into the structure of personal deliberation. Unlike books, search engines, or conventional recommendation systems, large language models can respond to a user's specific context by reframing the problem, generating previously unconsidered actions, proposing evaluative criteria, organizing evidence, constructing reasons, and simulating counterfactual futures. Artificial intelligence therefore increasingly affects not only what information a person receives, but how a choice is formed. Existing research has examined algorithmic advice, intelligent choice architecture, distributed cognition, human--AI agency, transformative choice, long-run reliance, and responsibility gaps. This paper asks a different question: what happens when a choice is partly formed through human--AI interaction, but only one participant must convert one of the candidate futures into a lived and path-dependent personal history? The paper develops three linked concepts. First, an \emph{AI-mediated deliberative field} describes the problem space, candidate actions, evaluative dimensions, weights, reasons, and counterfactual futures through which a person deliberates. Second, \emph{historical incorporation} describes the process by which a realized event changes the state from which the same person must continue: knowledge, resources, relationships, preferences, identity, opportunity sets, and subsequent decision conditions. Third, \emph{recursive chooser formation} captures the fact that a consequence-bearing decision changes the person who will make the next decision. The central claim is that \emph{distributed deliberation does not imply distributed history}. In many present advisory uses of generative AI, the model can contribute substantially to the construction and weighting of possible futures while the human user remains the primary locus of historical actualization and consequence-bearing change. This creates \emph{asymmetric historical incorporation}: the choice may be jointly formed, while its realized consequences are incorporated primarily into one participant's biography. Repeated use can then generate a recursive dynamic in which AI-mediated choices alter the future chooser who returns to AI in the next round. The paper does not claim that AI necessarily diminishes human agency, that human--AI coupling constitutes a new conscious subject, or that historical consequence-bearing determines moral or legal responsibility. Instead, it proposes a framework for studying human agency longitudinally when deliberation becomes increasingly AI-mediated but lived consequences remain unevenly distributed.