摘要
With the rapid advancement of large-scale language models,artificial intelligence(AI)systems such as ChatGPT have increasingly exhibited human-like characteristics.However,understanding the extent to which these systems resemble actual human traits remains a critical question.This study evaluated three versions of ChatGPT(3.5,4o,and 4o mini)across 84 dimensions of psychological characteristics,revealing notable differences from human norms and identifying key areas where alignment with human traits has progressively improved with each version update.By clustering the dimensions into broader psychological subdomains,the study highlighted consistent patterns of divergence and improvement across model iterations.These findings underscore the evolving psychological profile of ChatGPT and its increasing convergence with human characteristics.
基金
supported by the National Natural Science Foundation of China(Grant Nos.32371125,32071081)。
