CONVERSATIONAL AGENTS AS TOOLS AND BEINGS: HOW PEOPLE ANTHROPOMORPHIZE CHATBOTS

A key challenge in research on text-based chatbots concerns the tendency to anthropomorphize them: chatbots can mimic certain human qualities, and users in turn often ascribe human traits to them.The perception of humanness plays a crucial role in human-chatbot interactions because it strongly shapes how users experience the conversation.At the same time, users may make potentially harmful decisions and experience negative psychological impacts when they "collaborate" with a chatbot they believe possesses human-like abilities: projecting human qualities onto an artificial agent can foster the misleading perception that the agent truly "understands" them and "feels for" them, which may lead to overreliance and overtrust.In this article, I examine how people ascribe humanness to both Large Language Model (LLM)-based chatbots and pre-LLM chatbots.Building on a literature review on human-chatbot interaction and empirical findings, I explain how people tend to anthropomorphize conversational agents, conceptualizing conversations with them as interactions with entities whose humanness is continuously ascribed and denied by human users, who may attribute or withdraw specific human-like traits, like emotions, intelligence, and agency.