Algorithmic Bias and AI-Generated Literature: Rethinking Authorship, Intention, and Critical Hermeneutics in the Age of AI

The proliferation of Large Language Models (LLMs) and deep generative neural networks (DGN) in creative writing, across a wide range of domains, has disrupted the current paradigms of literary production and critical theory. This paper challenges the intersection of algorithmic bias and synthetic literature by showing how autoregressive architectures can perpetuate, amplify and naturalize biases found in web-scale training corpora, thereby propagating historical biases. This study looks at the survivability of the concept of artistic intentionality and critical hermeneutics in a method of statistical token prediction of literary text, using the theoretical frameworks of Roland Barthes's post-structuralist critique of the author, Michel Foucault's author-function and N. Katherine Hayles' distributed cognition. Reflecting on recent interventions in Critical (AI) studies (2023-2026), the paper explores computational creativity, positing that contemporary generation systems have reduced the semantic insight in favor of syntactic fluidity, and have compelled cross-entropy-led linguistic flattening rendering non-Western epistemologies, vernaculars and marginalised idioms vulnerable to disenfranchisement. Finally, the paper proposes a post-humanist critical discourse, which is neither technologically deterministic nor reactionary humanist; it sees literature as an act of emotional, historical and socio-political resistance.