Title: Mapping Ethical Discourse on the Anthropomorphization of Large Language Model-Based Conversational Agents

Abstract:

The growing prevalence of large language model (LLM)-based conversational agents (CAs) has brought the phenomenon of anthropomorphisation—the assignment of human-like traits to non-human entities—to the forefront. In contrast to previous generations of chatbots, modern LLM-driven CAs frequently produce linguistic and interactional signals, such as first-person self-reference alongside affective and epistemic expressions. Empirical studies indicate that these features can significantly boost user engagement. However, this tendency toward anthropomorphisation also sparks ethical debates, raising issues such as deception, excessive reliance, and the potential for exploitative relational dynamics. Conversely, certain scholars posit that such anthropomorphic interactions can foster inclusion, enhance well-being, and support autonomy.

Although interest in this topic is surging, existing literature remains fragmented across various fields. There is considerable inconsistency in how researchers define, operationalize, and normatively assess anthropomorphisation. To address this, our scoping review maps ethically focused research on the anthropomorphization of LLM-based CAs, drawing from five academic databases and three preprint repositories. We synthesize findings related to (1) conceptual underpinnings, (2) ethical opportunities and challenges, and (3) methodological strategies.

Our analysis reveals a consensus on attribution-based definitions of anthropomorphisation, yet significant divergence exists in how these concepts are operationalized. The prevailing normative framework is predominantly risk-oriented, and there is a scarcity of empirical studies that connect observed interaction effects to practical governance guidance. We conclude by proposing a future research agenda and offering design and governance recommendations to facilitate the ethical integration of anthropomorphic cues in LLM-based conversational agents.

Source: arXiv Generated at: 2026-06-03 00:00:00 UTC