KJCKorea
Journal Central

This paper critiques the instrumentalist limitations of contemporary AI ethics—which currently relies heavily on declarative and ex-post regulatory models—and reconstructs a dynamic model of AI ethics by integrating the technological philosophies of Gilbert Simondon and Bruno Latour. Simondon’s theory of ontogenesis conceptualizes AI as a processual individual undergoing concretization within an associated milieu, calling for the cultural establishment of a technical culture that mediates between technical practices and social values. Furthermore, Latour’s theory of mediation views AI as a network of detours and delegations that translates goals and actions, proposing moral repetition—which introduces friction into black-boxed chains of action—as a concrete ethical practice. The convergence of these two perspectives shifts AI ethics away from the static application of norms. Instead, it transforms AI ethics into a dynamic practice of governance that visualizes, continuously monitors, and revises the genetic lineages and delegation pathways throughout the entire lifecycle of the technology.

Science and technology operate together with the expectations that experts invest in them. Expert expectations guide people’s perception and actions concerning technology in particular directions and lend them legitimacy. The public, however, does not always accept expert expectations as they are presented. This study seeks to reveal the process through which expert expectations about AI are reconstructed as they become articulated with the folk theories of viewers. Specifically, it takes as its case the six most-viewed videos returned by a YouTube search using the Korean term “인공지능” (artificial intelligence), together with 816 comments drawn from the 1,739 comments that received at least ten likes, and analyzes both the ways in which the expert justifies these expectations and the ways in which viewers receive them. The expert foresees sweeping structural changes driven by AI while leaving preparation for them to individuals, and pairs firm certainty with reservations that limit the expert’s accountability for those expectations. Viewers, for their part, draw on personal experience, secondhand accounts, and familiar narratives to endorse, refine, or contest these expectations. By shedding light on how viewers reconstruct these expectations, the study raises the question of responsibility for experts who voice expectations about a future that cannot yet be verified.

This article examines the risk of interpreting failed online data collection as public nonresponse. It audits records concerning twenty-two films by Hong Sang-soo and fourteen novels by Choi In-hun across fifty-two collection routes. The resulting 1,872 work-source records are treated as units of procedural inspection, not as counts of independent responses. The analysis distinguishes three diagnostic branches applied to 347,992 stored rows: exact-text repetition, contextual and classification checks, and source-score combinations. These checks address different questions and do not establish a representative audience sample. The article distinguishes access failure, unresolved work context, classification failure, and non-observation within a specified observation window. It proposes the infrastructural selectivity of reception visibility to describe how platform access and researchers’ processing rules jointly determine what becomes available as reception evidence. Its contribution is to locate the point at which technical non-observation can become an unwarranted judgment about cultural silence. Preserving document context and recording the conditions of exclusion make this judgment open to scrutiny. The argument connects critical infrastructure studies with data auditing while keeping file preservation distinct from semantic and interpretive validation. It thereby identifies evidential conditions that should precede claims about audience attitudes or cognitive experience.