Developing and Validating the Spanish Version of the Large Language Models Dependency Scale (LLM-D12-SP) 开发和验证大型语言模型依赖量表的西班牙语版本（LLM-D12-SP）

TL;DR

• The study validates the Spanish version of the Large Language Model Dependency Scale (LLM-D12-SP), a psychometric tool assessing psychological dependency on LLMs.
• The scale measures two dimensions: Instrumental Dependency (task and decision reliance) and Relationship Dependency (companionship and social interaction reliance).
• Confirmatory factor analysis confirmed the two-factor structure, with good internal consistency (Cronbach's alpha = 0.89 total).
• The scale showed discriminant validity and external associations with internet addiction and perceived LLM trustworthiness, but weak links to need for cognition.
• This validation extends prior English and Arabic versions, supporting cross-linguistic use of the scale in organizational and research contexts.

Why It Matters

As LLMs become integral to work and communication, understanding psychological dependency is critical for ethical AI deployment and user well-being. This validated Spanish-language tool enables researchers and practitioners to assess dependency in a rapidly growing linguistic demographic, supporting culturally sensitive interventions and policy development. The findings also reinforce the need for cross-linguistic validation of AI-related psychometric instruments to ensure equitable global research and application.

Technical Details

• The LLM-D12-SP is a 12-item self-report scale with two subscales: Instrumental Dependency (6 items) and Relationship Dependency (6 items).
• Data were collected from 386 Spanish-speaking participants (mean age 28.0 years, SD = 6.1; 55% male), recruited via online platforms.
• Confirmatory factor analysis (CFA) confirmed the original two-factor model fit (CFI = 0.95, RMSEA = 0.06), supporting the scale’s structural validity.
• Internal consistency was high: Cronbach’s alpha = 0.89 (total), 0.86 (Instrumental), 0.85 (Relationship).
• Discriminant validity was established via correlation analyses showing moderate inter-subscale correlation (r = 0.52), indicating related but distinct constructs.
• External validity was assessed through correlations with internet addiction (positive association), perceived LLM trustworthiness (positive association), and need for cognition (weak or no association).

Industry Insight

Organizations integrating LLMs into workflows should consider monitoring employee dependency using validated tools like the LLM-D12-SP to prevent over-reliance and support healthy human-AI collaboration. The cross-linguistic validation of this scale enables multinational companies and researchers to standardize dependency assessments across diverse linguistic groups, promoting inclusive AI governance. Future efforts should focus on longitudinal studies to track dependency changes over time and evaluate the impact of training or policy interventions on LLM use behaviors.

TL;DR

• 本研究完成了西班牙语版大语言模型依赖量表（LLM-D12-SP）的首次验证，填补了西语人群心理依赖评估工具的空白。
• 量表包含工具性依赖与关系依赖两个维度，经386名西语参与者验证，具有良好的内部一致性（Cronbach's α=0.89）和结构效度。
• 验证结果支持该量表在跨文化语境下的适用性，为组织环境中LLM使用的心理影响研究提供了可靠测量工具。

为什么值得看

该研究解决了西语区LLM快速普及背景下缺乏本土化心理评估工具的痛点，为后续跨文化人机依赖比较研究奠定基础。其验证方法严谨，结果可直接应用于企业员工数字健康评估及AI伦理干预设计。

技术解析

• 采用Confirmatory Factor Analysis（CFA）验证量表双因子结构，确认工具性依赖（任务/决策支持）与关系依赖（陪伴/社交）的区分效度。
• 样本覆盖386名西语使用者（平均年龄28岁，55%男性），通过Cronbach's α评估内部一致性，总分α=0.89，子量表α均>0.85。
• 外部效度检验显示依赖程度与网络成瘾、LLM信任度呈正相关，与认知需求无显著关联，验证了量表的行为预测能力。
• 量表开发严格遵循跨文化心理测量学规范，确保语言等效性与文化适配性，为后续多语言推广提供方法论参考。

行业启示

• 企业应引入此类标准化量表评估员工对LLM的依赖风险，尤其关注工具性依赖引发的决策惰性关系依赖导致的人际疏离。
• 开发者需在AI产品中嵌入依赖监测机制，针对高依赖用户设计认知干预模块，避免过度工具化或情感化使用。
• 跨文化AI伦理研究需优先建立多语言心理评估基准，避免单一文化视角导致的依赖风险误判，推动全球人机协作规范制定。