Scholarlink

<p>Recent work in computational psychiatry has explored artificial agents as designed models of mental and neurological disorders, typically by introducing pathological parameters. This paper advances an inverse claim: large language models (LLMs) already instantiate a structural configuration resembling dementia with Lewy bodies (DLB). We argue that the core disturbance in DLB lies in instability of mechanisms that stabilize experience as real and contextually grounded, while generative and cognitive capacities remain relatively preserved. Hallucinations and fluctuations are thus interpreted as breakdowns in reality endorsement rather than failures of perception or reasoning. Analogously, LLMs generate coherent and contextually appropriate content but lack intrinsic mechanisms for stabilizing reference, truth commitment, and continuity across contexts. So-called hallucinations are therefore better understood as expressions of missing operations required for reality stabilization, rather than as technical errors of generation. Drawing on Metaqualia Theory (MTQ) as a structural vocabulary, we demonstrate that both DLB and LLMs exhibit a dissociation between generative capacity and reality stabilization. This structural homology reframes hallucination in both biological and artificial domains, suggesting that contemporary AI systems may already function as inadvertent structural models of disorders of subjectivity, with implications for psychiatry, AI ethics, and theories of responsibility.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Large Language Models as Inadvertent Models of Dementia with Lewy Bodies: How a Disorder of Reality Construction Illuminates AI Hallucination

  • 期刊论文

  • 发表日期：

    2026年4月11日

查看全文
  • Minoru Matsui

摘要

Recent work in computational psychiatry has explored artificial agents as designed models of mental and neurological disorders, typically by introducing pathological parameters. This paper advances an inverse claim: large language models (LLMs) already instantiate a structural configuration resembling dementia with Lewy bodies (DLB). We argue that the core disturbance in DLB lies in instability of mechanisms that stabilize experience as real and contextually grounded, while generative and cognitive capacities remain relatively preserved. Hallucinations and fluctuations are thus interpreted as breakdowns in reality endorsement rather than failures of perception or reasoning. Analogously, LLMs generate coherent and contextually appropriate content but lack intrinsic mechanisms for stabilizing reference, truth commitment, and continuity across contexts. So-called hallucinations are therefore better understood as expressions of missing operations required for reality stabilization, rather than as technical errors of generation. Drawing on Metaqualia Theory (MTQ) as a structural vocabulary, we demonstrate that both DLB and LLMs exhibit a dissociation between generative capacity and reality stabilization. This structural homology reframes hallucination in both biological and artificial domains, suggesting that contemporary AI systems may already function as inadvertent structural models of disorders of subjectivity, with implications for psychiatry, AI ethics, and theories of responsibility.