Scholarlink

<p>The rise of large language models (LLMs) signals a far-reaching transformation in our relationship to language and text. While public debates tend to focus on concerns over synthetic content and misinformation, this paper argues that a more profound threat lies elsewhere: in the disruption of the textual relationship between citizens and institutions on which democratic rule-of-law societies depend. Drawing on continental cultural and legal theory, it develops a framework centered on two indispensable author-figures: the <i>fictional author</i>, a symbolic instance (such as the People, God, or the State) whose authority grounds institutional legitimacy and is perpetually staged through democratically accountable interpreters; and the <i>human author</i>, whose practices of reading and writing sustain the mutability of the textual foundations of democratic order. Against the position advanced by Coeckelbergh and Gunkel (AI &amp; Society 39: 2221–2231, 2024), who welcome human-machine co-authorship as a transcendence of logocentrism, this paper argues that authorship is not an obsolete remnant of an outdated social model, but an operational precondition for institutional coherence. The paper shows that generative AI does not dissolve metaphysical truth claims but, as Lassègue and Longo (L’empire numérique: de l’alphabet à l’IA, 2025) argue through their concept of “algorithmic ideology,” displaces them onto data-driven processes presumed to yield truths beyond interpretive contestation. Simultaneously, the structural asymmetries of human-machine text production render genuine co-authorship illusory and erode the individual's constructive engagement with the societal text. The paper concludes that a meaningful integration of generative AI into democratic societies requires confining LLMs to auxiliary functions. Only by preserving the two forms of authorship, i.e. by entrusting the production and interpretation of texts to institutionally authorized and democratically legitimated actors and by cultivating the individual capacity for critical textual engagement, can the core conditions of democratic life be safeguarded.</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

Truth, text, and technology: critical reflections on the institutional implications of generative artificial intelligence

  • 期刊论文

  • 发表日期：

    2026年3月23日

查看全文
  • Katrin Becker

摘要

The rise of large language models (LLMs) signals a far-reaching transformation in our relationship to language and text. While public debates tend to focus on concerns over synthetic content and misinformation, this paper argues that a more profound threat lies elsewhere: in the disruption of the textual relationship between citizens and institutions on which democratic rule-of-law societies depend. Drawing on continental cultural and legal theory, it develops a framework centered on two indispensable author-figures: the fictional author, a symbolic instance (such as the People, God, or the State) whose authority grounds institutional legitimacy and is perpetually staged through democratically accountable interpreters; and the human author, whose practices of reading and writing sustain the mutability of the textual foundations of democratic order. Against the position advanced by Coeckelbergh and Gunkel (AI & Society 39: 2221–2231, 2024), who welcome human-machine co-authorship as a transcendence of logocentrism, this paper argues that authorship is not an obsolete remnant of an outdated social model, but an operational precondition for institutional coherence. The paper shows that generative AI does not dissolve metaphysical truth claims but, as Lassègue and Longo (L’empire numérique: de l’alphabet à l’IA, 2025) argue through their concept of “algorithmic ideology,” displaces them onto data-driven processes presumed to yield truths beyond interpretive contestation. Simultaneously, the structural asymmetries of human-machine text production render genuine co-authorship illusory and erode the individual's constructive engagement with the societal text. The paper concludes that a meaningful integration of generative AI into democratic societies requires confining LLMs to auxiliary functions. Only by preserving the two forms of authorship, i.e. by entrusting the production and interpretation of texts to institutionally authorized and democratically legitimated actors and by cultivating the individual capacity for critical textual engagement, can the core conditions of democratic life be safeguarded.