Into the Hyperreal: AI Hallucinations and Trolling as Simulacra

Engineering notes

Engineering notes will be added by the aipentium editorial team.

Chinese explanation / 中文解读

中文解读待补充：本站会优先为大语言模型、生成式AI、ChatGPT相关技术、计算机视觉、深度学习等高价值论文补充中文说明。

Original abstract

This paper offers a conceptual reframing of AI hallucination and online trolling phenomena through Jean Baudrillard’s theory (1994) of simulation and simulacra, arguing that both operate according to structurally similar epistemic logics. Research has characterized large language model (LLM) hallucinations as a structural feature of probabilistic text generation systems optimized for plausibility, and trolling as a behavior that leverages ambiguity, irony, and repetition to sustain circulation. Across both domains, authority is shown to emerge from platform visibility and genre conformity rather than evidentiary grounding. Using illustrative cases and prior research, this paper demonstrates how corrective measures frequently fail to restore reference and instead, get absorbed into the same circulatory economy of signs within a hyperreal information order. By reframing AI hallucinations and online trolling as features of a shared sociotechnical logic of simulation, this paper contributes a conceptual account of how contemporary information systems reshape epistemic authority, credibility, and responsibility.

Links and sources

Comments