研究内容

Research contents

本研究以北京首钢公园的虚拟现实模拟为场景，探讨智能非玩家角色（NPC）引导的互动对用户体验的影响，重点分析环境背景知识（历史文化信息）在用户参与度、认知理解和情感唤醒中的作用。研究基于具身认知理论，将 30 名参与者分为实验组（接受 NPC 提供的历史文化背景信息）和对照组（无 NPC 引导），对比两组在认知理解（对建筑设计与城市空间的理解）、情感投入（生理及主观情感反应）、注意力分配（眼动行为）上的差异，旨在填补现有研究中对 NPC 引导互动与具身认知结合影响的空白，为 VR 在城市规划、教育及文化遗产领域的应用提供理论支持。

This study uses virtual reality simulation in Beijing Shougang Park as a scenario to explore the impact of intelligent non-player roles (NPC)-guided interaction on user experience, focusing on analyzing the role of environmental background knowledge (historical and cultural information) in user participation, cognitive understanding and emotional arousal. Based on embodied cognitive theory, the 30 participants were divided into experimental groups (received historical and cultural background information provided by NPCs) and control groups (without NPC guidance). The differences in cognitive understanding (understanding of architectural design and urban space), emotional investment (physiological and subjective emotional responses), and attention distribution (eye movement behavior) were compared. The purpose is to fill the gap in the impact of the combination of NPC-guided interaction and embodied cognition in existing research, and provide theoretical support for the application of VR in the fields of urban planning, education and cultural heritage.

研究方法

Research methods

研究采用混合研究方法，招募 30 名 18-35 岁参与者（男女各 15 名），随机分为实验组和对照组，均经伦理审批并签署知情同意书。实验在 Unreal Engine 5 构建的首钢公园虚拟环境中进行，实验组通过 ConvAI 插件实现与 NPC 的实时自然语言交互以获取历史文化信息，对照组无此引导。数据采集包括：眼动追踪（HTC VIVE Pro Eye 记录注视次数和时长）、生理指标（ErgoLAB 传感器收集 HRV、EDA、RESP）、问卷（Likert 量表及开放式问题评估认知、情感与满意度）。数据分析采用 t 检验（比较眼动和生理数据）、描述性统计与 ENA（分析问卷数据）、结构方程模型（SEM）整合变量以探究对情感参与和认知理解的影响，通过 CFI 和 RMSEA 评估模型拟合度。

The study adopted a mixed research method and recruited 30 participants aged 18-35 (15 men and women) and were randomly divided into experimental and control groups, all of which were ethical and signed informed consent. The experiment was conducted in the Shougang Park virtual environment constructed by Unreal Engine 5. The experimental group realized real-time natural language interaction with NPC through the ConvAI plug-in to obtain historical and cultural information. The control group did not have this guidance. Data collection includes: eye tracking (HTC VIVE Pro Eye recording number and duration of gazes), physiological indicators (ErgoLAB sensor collection HRV, EDA, RESP), questionnaire (Likert scale and open-ended question assessment cognition, emotion and satisfaction). The data analysis used a t-test (comparing eye movement and physiological data), descriptive statistics and ENA (analytical questionnaire data), and structural equation model (SEM) to investigate the impact on emotional participation and cognitive understanding, and model fit was evaluated through CFI and RMSEA.

研究结论

Research conclusions

研究结果显示，NPC 引导显著提升了用户的体验效果：实验组在兴趣区的注视次数（12±3.5）和时长（7.8±2.5 秒）显著高于对照组（8±2.8、4.6±1.9 秒），表明 NPC 有效引导了注意力；生理指标上，实验组的 HRV（54±11 ms）、EDA（0.58±0.15 µS）、RESP（18.2±2.6 次 / 分）均显著高于对照组，反映更高的情感唤醒；问卷结果显示实验组认知理解（4.3±0.5）和情感参与（4.2±0.6）得分更高，ENA 分析表明其概念网络更复杂；SEM 验证了 NPC 互动通过眼动行为（具身认知）和生理反应间接影响情感参与和认知理解，直接效应也显著（系数分别为 0.47、0.52），模型拟合良好（CFI=0.96，RMSEA=0.05）。综上，NPC 提供的背景知识能增强 VR 环境中的认知与情感体验，为相关领域应用提供实践参考。

The research results showed that NPC guidance significantly improved the user's experience effect: the number of gazes (12±3.5) and duration (7.8±2.5 seconds) in the experimental group were significantly higher than those in the control group (8±2.8, 4.6±1.9 seconds), indicating that NPC effectively guided attention; in terms of physiological indicators, the HRV (54±11 ms), EDA (0.58±0.15 µS), and RESP (18.2±2.6 times/min) in the experimental group were significantly higher than those in the control group, reflecting higher emotional arousal; the questionnaire results showed that the experimental group had higher scores of cognitive understanding (4.3±0.5) and emotional participation (4.2±0.6), and ENA analysis showed that its conceptual network was more complex; SEM verified NPC Interaction indirectly affects emotional participation and cognitive understanding through eye movement behavior (embodied cognition) and physiological responses, and the direct effect is also significant (coefficients are 0.47 and 0.52, respectively), and the model fits well (CFI=0.96, RMSEA=0.05). In summary, the background knowledge provided by NPC can enhance cognitive and emotional experience in the VR environment and provide practical reference for applications in related fields.