From Voice to Emulation: Tracing the Scholarly Persona across Human and Algorithmic Academic Writing

Abstract

The expansion of AI text generation in academic environments raises essential questions about how scholarly persona is constructed when writing is not produced by a human knower. This study applies a three-dimensional model—dialogic positioning, evaluative practice, and identity projection—to compare persona construction in first-year and fourth-year L2 academic writing and in AI-generated texts. A parallel corpus of 150 paragraphs responding to the same prompt was analyzed through normalized frequency profiling supported by contextual interpretation. First-year writing enacts dialogicity through direct reader engagement, uses affect-driven evaluation to assert value, and projects identity through dense autobiographical presence. Fourth-year writing regulates dialogicity via stance modulation and metadiscourse, employs purpose-linked appreciation and cautious judgment, and asserts a confident academic self responsible for claims. AI-generated texts display structural dialogicity through patterned hedging and connectives, apply neutralized and interchangeable evaluation without criteria, and erase identity by removing accountable selfhood. These contrasts motivate the Voice–Emulation Continuum, a theoretical contribution distinguishing socially situated scholarly voice from algorithmic emulation lacking epistemic agency. Pedagogically, AI output is positioned as a diagnostic contrast that helps writers recognize and reinforce the stance ownership, evaluative justification, and identity accountability constitutive of human academic writing.