This article develops a fictionality-based approach to artificial intelligence, proposing that human-AI interaction unfolds through the dynamics of character engagement. Drawing on narrative theory, cognitive science, and AI studies, it examines large language models such as ChatGPT, Claude, and LAURA (modeled on Laura Palmer) as fictional objects of relation-emergent agents whose apparent minds arise through interpretive stances, narrative framing, and phenomenological opacity. Integrating Dennett's notions of real patterns and intentional stance with theories of characterization and fictionality, it identifies four illusions-of independent agency, independent identity, codependency, and independent poetics-through which users perform mind-building and world-probing exchanges with AI. Distinguishing between designed and emergent fictionality, and between AI immersivity and emersivity, the article shows how linguistic patterning and cognitive projection together sustain real yet imagined forms of intersubjectivity. While such patterned engagements enable genuine cognitive and creative extensions, they also expose the ethical perils of counterfeit people. Fictionality, however, offers a training ground for double exposure-the capacity to navigate what is both real and fictive at once. The article concludes by proposing a view of human-AI interactions as introspection by computation: AI as a technology that extends self-knowledge through the externalization of mind.
原文链接
分享
引用
