Rethinking AI anthropomorphism: A holistic conceptualization and scale across AI systems and service contexts

摘要

Building on Gestalt theory, this research introduces a holistic framework for defining artificial intelligence (AI) anthropomorphism, which guides the development of a parsimonious Scale of AI Anthropomorphism (SAIA). Through a systematic literature review, seven focus group discussions, and rigorous six-step scale development procedures, the research conceptualizes and validates a multidimensional framework that captures both external and internal human-like traits, including virtues and vices of anthropomorphic AI. Based on 2944 valid responses from five studies, a 40-item, six-dimensional parsimonious SAIA was developed, namely, human-like appearance, cognitive competency, adaptive capacity, social intelligence, morality, and fallibility. The scale demonstrated strong psychometric properties through comprehensive qualitative and quantitative validation. The SAIA serves as a robust tool for assessing the anthropomorphism of both tangible (e.g., robots) and intangible (e.g., chatbots) AI across normal service and service failure contexts. • Introduces a holistic framework for AI anthropomorphism using Gestalt theory. • Develops and validates a six-dimensional Scale of AI Anthropomorphism (SAIA). • Captures both external and internal human-like traits, including virtues and vices. • Demonstrates strong psychometric properties through five studies with 2944 responses. • Provides a generalizable tool for assessing AI anthropomorphism across AI systems and service contexts.