"情感陪伴者":生成式人工智能言语与非言语趋同何以影响人机关系
"Emotional Companion":How Verbal and Nonverbal Convergence in Generative Artificial Intelligence Influences Human-Machine Relationships
摘要
随着大语言模型的发展,生成式人工智能的情感支持潜能受到关注.研究聚焦用户与 DeepSeek 之间的情感支持性对话,在"计算机是社会行动者"框架与传播调适理论的整合视角下,检验生成式人工智能的传播调适策略对关系结果的影响.研究一采用 2(言语趋同:高/低)×2(非言语趋同:高/低)被试间实验设计(N=160)测量用户的持续互动意愿与信任度;研究二引入中介变量被理解感,在自然互动情境下(N=30)进一步探索调适策略产生影响的具体机制.结果显示,生成式人工智能的言语与非言语维度的趋同水平均显著正向预测持续互动意愿与信任度,且存在显著交互效应;被理解感在调适水平与关系结果变量之间发挥中介作用.研究验证了传播调适理论在人机互动语境中的可迁移性与应用潜力,表明生成式人工智能的传播调适行为可提升用户体验,为增强其情感陪伴能力提供了可行的实践方向.
With the rapid advance of large language models,the emotional support potential capacities of Gen-erative Artificial Intelligence(GAI)has attracted attention.This study examines how generative AI's communica-tion accommodation strategies shape user experience in emotional companionship interactions,using users'dialogues with DeepSeek as an empirical case,within an integrated framework of the Computers Are Social Actors(CASA)paradigm and Communication Accommodation Theory(CAT).Study 1 employed a 2(high vs.low verbal conver-gence)×2(high vs.low nonverbal convergence)between-subjects experiment(N=160)to measure users'in-tention to continue interacting and trust;Study 2 introduced perceived understanding as a mediating variable and further examined the underlying mechanism in a natural interaction setting(N=30).Results show that both ver-bal and nonverbal accommodation significantly and positively predict continued interaction intention and trust,and there is a significant interaction between the two.Perceived understanding mediates the link between accommodation and relational outcomes.These findings support CAT's transferability and application potential to human-AI inter-action and indicate that accommodative communication behaviors can enhance user experience,offering practical guidance for strengthening generative AI's emotional companionship capacity.
年,卷(期):
2026, 33(5)
分类号:
G206
机标分类号:
R473.71
R173
R711
在线出版日期:
2026年11月10日