Developing and Validating the Chinese Version of the Attitudes Toward Large Language Models Scale (AT-LLM Chinese)

Large Language Models (LLMs) have become integral to education, business, and public life, yet cross-culturally validated tools for assessing public attitudes toward them remain scarce. This study adapted and validated two established five-item instruments, the Attitudes Toward General LLMs (AT-GLLM) and Attitudes Toward Primary LLMs (AT-PLLM) scales, for use in the Chinese context. Each scale includes two items measuring acceptance and three items measuring fear. A sample of 576 Chinese LLM users completed the Chinese versions of both scales alongside the Attitudes Toward Artificial Intelligence (ATAI) measure and a self-efficacy scale. Confirmatory factor analyses supported the expected two-factor structure, acceptance and fear, for both scales, with acceptable model fit indices. Reliability coefficients ranged from α = .54 to .74, with the lower value corresponding to the two-item acceptance subscale, as expected given its brevity. Measurement invariance testing across low- and high-frequency LLM users confirmed configural, metric, scalar, and strict invariance, indicating that the constructs operate equivalently across experience levels. External validation showed that ATAI-acceptance strongly predicted LLM acceptance, whereas ATAI-fear predicted LLM-related fear, supporting convergent validity. Self-efficacy did not significantly predict LLM attitudes once general AI attitudes were accounted for. These findings confirm the psychometric soundness and cross-cultural applicability of the AT-GLLM and AT-PLLM scales in China. By providing validated instruments for one of the world’s largest and most active AI ecosystems, this study advances global understanding of LLM attitudes and offers tools for guiding responsible, trust-oriented LLM design and policy development.