Scholarlink

Under the empowerment of Large Language Models (LLMs), library virtual assistants have transitioned from “machine agents” to “AI digital humans,” significantly enhancing the intelligence, integration, and efficiency of library services. However, previous research has primarily treated library virtual assistants as AI devices, focusing on technical trust while neglecting interaction trust within library-specific contexts and the role of digital humans as AI agents. This study explores the factors influencing users’ perceived trust and acceptance of digital human services. Grounded in the contextual characteristics of library services, we propose an interaction trust framework for library AI digital humans, integrating technical, emotional, and experiential dimensions. The framework encompasses six constructs: Perceived Performance (PP), Scenario Context (SC), Anthropomorphism (AN), System Transparency (ST), Multimodal Interaction (MI), and Interaction Comfort (IC). Data were collected via an online survey of 350 library users and analyzed using Structural Equation Modeling (SEM) to test the proposed hypotheses. The results indicate that PP, SC, AN, ST, MI, and IC each have a significant positive effect on interaction trust, with IC partially mediating the influence of AN and MI on trust. These findings validate the rationality of user perception logic, provide empirical support for optimizing the interaction experience of library AI digital humans, and enhance the understanding of the interplay between interaction factors and trust design. This study promotes a shift in AI agent design from a “technology-centric” to a “context-centric” paradigm.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Exploring Human-AI Interaction Perception Factors and Collaborative Trust in Library AI Digital Humans

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Jun Liu,
  • Hongtao Wu,
  • Binxin Hu,
  • Xiaoling Yin

摘要

Under the empowerment of Large Language Models (LLMs), library virtual assistants have transitioned from “machine agents” to “AI digital humans,” significantly enhancing the intelligence, integration, and efficiency of library services. However, previous research has primarily treated library virtual assistants as AI devices, focusing on technical trust while neglecting interaction trust within library-specific contexts and the role of digital humans as AI agents. This study explores the factors influencing users’ perceived trust and acceptance of digital human services. Grounded in the contextual characteristics of library services, we propose an interaction trust framework for library AI digital humans, integrating technical, emotional, and experiential dimensions. The framework encompasses six constructs: Perceived Performance (PP), Scenario Context (SC), Anthropomorphism (AN), System Transparency (ST), Multimodal Interaction (MI), and Interaction Comfort (IC). Data were collected via an online survey of 350 library users and analyzed using Structural Equation Modeling (SEM) to test the proposed hypotheses. The results indicate that PP, SC, AN, ST, MI, and IC each have a significant positive effect on interaction trust, with IC partially mediating the influence of AN and MI on trust. These findings validate the rationality of user perception logic, provide empirical support for optimizing the interaction experience of library AI digital humans, and enhance the understanding of the interplay between interaction factors and trust design. This study promotes a shift in AI agent design from a “technology-centric” to a “context-centric” paradigm.