Department of Psychology, University of British Columbia, Canada.
School of Psychology, University of Plymouth, United Kingdom.
Conscious Cogn. 2024 Sep;124:103733. doi: 10.1016/j.concog.2024.103733. Epub 2024 Aug 7.
The rise of powerful Large Language Models (LLMs) provides a compelling opportunity to investigate the consequences of anthropomorphism, particularly regarding how their exposure may influence the way individuals view themselves (self-perception) and other people (other-perception). Using a mind perception framework, we examined attributions of agency (the ability to do) and experience (the ability to feel). Participants evaluated their agentic and experiential capabilities and the extent to which these features are uniquely human before and after exposure to LLM responses. Post-exposure, participants increased evaluations of their agentic and experiential qualities while decreasing their perception that agency and experience are considered to be uniquely human. These results indicate that anthropomorphizing LLMs impacts attributions of mind for humans in fundamentally divergent ways: enhancing the perception of one's own mind while reducing its uniqueness for others. These results open up a range of future questions regarding how anthropomorphism can affect mind perception toward humans.
强大的大型语言模型 (LLM) 的兴起为研究拟人化的后果提供了一个极具吸引力的机会，特别是关于他们的暴露可能如何影响个人看待自己（自我认知）和他人（他人认知）的方式。我们使用心智感知框架，研究了能动性（做事的能力）和体验（感受的能力）的归因。参与者在接触 LLM 响应之前和之后评估了他们的能动性和体验能力，以及这些特征在多大程度上是人类独有的。接触后，参与者增加了对自己能动性和体验品质的评价，同时降低了他们对能动性和体验被认为是人类独有的看法。这些结果表明，将 LLM 拟人化会以根本不同的方式影响对人类心智的归因：增强对自己心智的感知，同时降低对他人心智独特性的感知。这些结果提出了一系列关于拟人化如何影响对人类心智感知的未来问题。
