Preservation Is a Choice: Artificial Intelligence and the Conditions of Collective Learning

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).

Human societies possess unprecedented capacities to preserve information, yet informational survival does not guarantee that future communities will remain able to understand, evaluate, or learn from what survives. This paper develops capability preservation as a framework for examining that distinction. It argues that preservation concerns not only informational objects but also the relationships and epistemic capacities that allow later actors to reconstruct, contextualize, compare, evaluate, contest, revise, and renew inherited knowledge. Artificial intelligence makes this problem increasingly important because different forms of AI mediation can strengthen, weaken, redistribute, or relocate these capacities. Retrieval, translation, summarization, synthesis, and computational interpretation can expand access and reconstructability while also changing the pathways connecting inherited representations to their sources, contexts, alternatives, and methods. The paper proposes five dimensions of capability preservation: reconstructability, contextual recoverability, comparability, contestability, and renewability. These dimensions form the Capability Preservation Test, a diagnostic framework for evaluating what an informational environment leaves future actors capable of doing. Preservation should therefore be understood not simply as a choice about what information survives, but also as a choice about what remains possible for those who inherit it.