Anthropomorphism, Sentiment, and AI Hype on Reddit: The Moderating Role of LLM Chatbot Use Case

Abstract

The anthropomorphism of chatbots has been studied extensively, dating back to the first chatbot, ELIZA, created by Joseph Weizenbaum in the 1960s. Within fields like human-computer interaction (HCI), communication, and computer science, anthropomorphism has been investigated in relation to user behavior and narratives that influence public perceptions of chatbots. While human-like design and language were once treated as uniformly beneficial for attitudes and sentiment, recent work has complicated this assumption, pointing to ambivalent or even negative attitudes, as well as risks like increased AI hype. In addition, both anthropomorphism and its social responses vary based on individual and contextual factors (e.g., subjective vs. objective contexts). Modern large language model (LLM) chatbots like Claude and ChatGPT perform a variety of social and functional tasks for users with considerably different needs—and also face unprecedented levels of anthropomorphism and hype. Focusing on online communities of AI-interested publics, my study examines differences in anthropomorphism and its relationship to sentiment across three use cases (coding, writing, and media), adding nuance to HCI theories about task context and unpacking their relevancy to LLM chatbots. Moreover, building on existing hype theory, I construct a novel measure of AI hype driven by co-occurring future-oriented language, certainty, and risk/reward framing. Using a computational text analysis of conversations in LLM-focused Reddit communities, I find that anthropomorphism varies significantly by chatbot use case, as do the relationships between anthropomorphism and negative affect or hype. These findings provide new visibility into how chatbot anthropomorphism and attitudes vary across contexts—and how these patterns manifest in discussion of LLM risks and broader utopian and dystopian narratives.