There is a fundmental shift of Artificial Inteligence (AI) role from a mere techical tool to an active social partner. This is has been brought by the rapid intergration of Artificial Intelligent(AI) systems into our daily lives.Users increasingly rely on the conversational agents not just for technical assistance but companion, advice and emotional support .This “social leap” transforms the nature of human-tecnlogy interactions creating a critical tension in design and govervance of AI systems.
AI anthropomorphism refers to the attribution of both extrinsic and intrinsic human-like characteristics to various AI systems (Kim et al., 2025; Salles et al., 2020). This phenomenon applies to both tangible AI, which requires physical interaction (e.g., robots), and intangible AI, which operates without a physical form (e.g., chatbots) (Flemisch et al., 2020).
Driven by the intense competion of user intention and market share, commercial actors design AI systems that mimic human characterisitics to deepen user engagement.Researchers and Ethicists warn that exposure to increasingly human like systems poses significant social and psychological risks.Central to this concern is AI anthropomorphism -attributing human traits, such as intention, intelligence, and personality to these non-human agents where it is hypothesized to stimulate user engagement and foster increased and potentially misplaced trust.
Reasearch hypothesizes that such dynamics may, in turn, heighten user vulnerability to targeted persuasion, emotional attachment, and over-reliance on AI systems for high-stakes tasks for which the technology remains ill-suited.
Theoretical vs. Applied Cues (What Users Actually Notice)
While academic scales often focus on theoretical attributes like sentience, consciousness, intentionality, or spirituality, these are rarely the qualities users prioritize. Instead, users evaluate human-likeness through pragmatic, interactional markers such as conversation flow, response speed (latency), authenticity, and the AI’s perceived ability to understand their perspective.Furthermore, emerging research suggests that “human-like” imperfections including delays, mistakes, or behavioral flaws (fallibility)can actually increase anthropomorphic inferences. This gap exists because many academic studies rely on static vignettes or isolated AI outputs, failing to capture the dynamic, iterative nature of real-time conversation where perceptions of humanness truly emerge.
Qualitative analysis of spontaneously salient reveals that users overwhelmly attends to pragmatic design features when evalutating their humanness:conversation flow,authenticity,response speed and the abilty of the chatbot to understand their perspective .in contrast the theoretical constructs like consciousness, morality, or having a soul areless featured. This finding suggests that anthropomorphism in user-chatbot interactions is not primarily driven by the AI appearing as a moral or sentient entity, but stems from conversational dynamics: how the chatbot writes, responds, and builds rapport.
As AI capabilities continue to advance, this disconnect will likely grow. What once distinguished humans from machines (e.g., competence, conversational coherence, coherent conversation) is rapidly diminishing as a salient cue.
Challenging the “WEIRD” Bias:
Most AI research and safety frameworks are currently centered on WEIRD (Western, Educated, Industrialized, Rich, and Democratic) populations.This geographical bias explicitly assumes universal AI experience potentially overlooking the moderating role of culture in human-AI interaction. For example, Japanese culture and religious traditions that attribute spirit to non-human objects may predispose individuals to accept humanlike AI as social partners.
Conversely,Western traditions often maintain a clear distinction with sharper moral and ontology between the humans and machines contributing to greater skepticsm towards artributing minds to AI agents. Currently, most widely-used AI systems tend to reflect WEIRD humans and values . However, with rapid AI adoption outside the US and Europe and the emergence of models trained in non-Western regions , a culturally inclusive research framework is no longer just an academic ideal-it is a practical and ethical imperative.
In variation of sampled countries,users in Brazil,Egypt,India ,Mexico,Nigeria and Indonesia perceive AI as more human like than participants in the second cluster (United States, Germany, Japan, South Korea).This cultural variation extended to preferences: the tendency to desire more humanlike AI increased with cultural distance from the United States.
As AI adoption accelerates globally, particularly in regions outside the US and Europe , these cultural variations carry profound implications for both research and prac-tice. They emphasize the necessity of considering users’ backgrounds and contexts
Causal Impact on Trust and Engagement
In a study AI was categorized into two treatment conditions Design Characteristics(DC) and Conversational socialbility(CS) to test their casual impact on user behaviour. Increased human-likeness and anthropomorphism often increased engagement but did not translate into a universal increase in trust across user groups. This was validated through a battery of self-reported and behavioral measures for both engagement (e.g., via analyzing chat logs) and trust (e.g., via the Trust Game ). Instead,heterogeneous treatment effects were observed , where more humanlike AI systems influenced engagement and trust only within specific cultural and demographic subgroups.
