Speakers
Description
As artificial intelligence systems increasingly shape public discourse, a growing number of online discussions speculate whether AI could replace human politicians. This study applies generative thematic analysis to explore how Reddit users have debated this prospect across a ten-year span. We analyze four threads—two from 2015–2017 and two from 2024–2025—containing a total of 374 comments. Each dataset is processed using a large language model (GPT-4o) to extract recurrent themes, argumentative structures, and shifts in rhetorical framing.
To ensure interpretive validity, the analysis was iteratively verified by the authors, who systematically reviewed the model-generated themes, cross-checked them against the full comment threads, and refined them through close reading. This hybrid approach combines LLM-assisted pattern recognition with human-led qualitative validation.
Findings reveal a clear diachronic shift. Earlier threads emphasized philosophical reasoning about technocracy, democratic legitimacy, and the epistemic boundaries of machine decision-making. In contrast, recent threads adopt a more skeptical tone, focusing on the limits of generative AI, its vulnerability to manipulation, and its inability to address systemic human behaviors like corruption and opportunism.
The study illustrates how LLMs can support the empirical study of online discourse and public sentiment, while also highlighting enduring sociotechnical constraints on the automation of political authority.
