Cultural Variations in Human-AI Partnership: Initial Cross-Cultural Validation of the Transactive Memory System with GenAI (TMS-GenAI) Measurement Tool

Paper Title

Cultural Variations in Human-AI Partnership: Initial Cross-Cultural Validation of the Transactive Memory System with GenAI (TMS-GenAI) Measurement Tool

Publication Info

• Topic area: Cross-cultural validation of a measurement tool for human-AI cognitive partnerships.
• Keywords: Transactive Memory System, Generative AI, human-AI collaboration, cognitive offloading, cross-cultural validation, Extended Mind theory, Cognitive Self-Esteem.

Background and Problem

• Problem / challenge: No validated measurement tool exists to assess the quality, structure, and cultural variation of human–Generative AI transactive memory systems.
• Significance: Understanding human-AI cognitive partnerships is critical for equitable technology design and effective collaboration across diverse contexts.
• Motivation and related work: Prior research has extended Transactive Memory System (TMS) theory to human-AI partnerships, emphasizing AI as epistemic collaborators. However, existing TMS scales focus on human teams and lack constructs for AI collaboration, leaving a gap in measurement tools for assessing cognitive offloading and partnership with AI.

Solution

• Proposed approach: Development and initial validation of the TMS-GenAI measurement tool, which integrates classic TMS constructs with dimensions specific to human–Generative AI collaboration.
• Novelty:
  1. Introduction of a theoretically grounded measurement tool for human–Generative AI partnerships.
  2. Empirical validation across culturally distinct populations (Turkiye and the United States).
  3. Extension of TMS theory to include constructs like Generative AI Familiarity and cognitive offloading behaviors.
• Procedure and key techniques:
  • Development of a 28-item scale based on TMS, Extended Mind, and Cognitive Self-Esteem theories.
  • Administration to university students in Turkiye (N=437) and the United States (N=476).
  • Exploratory Factor Analysis (EFA) with Principal Axis Factoring and Promax rotation to identify factor structures.
  • Cross-cultural comparison using Tucker’s Congruence Coefficient and Adjusted Rand Index (ARI).

Results

• Concrete findings:
  • Six-factor structure identified in Turkish participants, explaining 67.31% of variance.
  • Five-factor structure identified in U.S. participants, explaining 61.59% of variance.
  • Core constructs (Ability to Think, Ability to Remember, Generative AI Offloading) showed high cross-cultural stability (Tucker’s Congruence ≥ 0.98).
  • Collaboration-related constructs (Specialization and Coordination) exhibited cultural divergence.
• Advantage over baselines: First validated tool for assessing human–Generative AI transactive memory systems, addressing gaps in existing TMS and cognitive offloading measures.
• Experiments / evaluation:
  • Factor analysis confirmed structural validity with high KMO values (Turkiye: 0.922; U.S.: 0.871) and significant Bartlett’s Test of Sphericity (p < 0.001).
  • Reliability analysis showed Cronbach’s alpha ≥ 0.85 for overall scale and ≥ 0.70 for most subscales.
  • Cross-cultural comparisons revealed partial structural non-invariance, with differences in how specialization and coordination are organized.
• Limitations and future work:
  • Reliance on self-report data and university student samples limits generalizability.
  • Context-specific items (focused on educational tasks) require adaptation for professional or organizational settings.
  • Future research should employ Confirmatory Factor Analysis (CFA), expand item pools, and test measurement invariance across diverse populations.

Summary

This study introduces the TMS-GenAI measurement tool to assess human–Generative AI cognitive partnerships, validated across Turkish and U.S. university students. The tool integrates TMS, Extended Mind, and Cognitive Self-Esteem theories, capturing dimensions like Ability to Think, Ability to Remember, Generative AI Offloading, Credibility, Specialization, and Coordination. Results show high cross-cultural stability for cognitive self-evaluations and offloading constructs, alongside cultural divergence in collaboration-related dimensions. The findings provide a foundation for future research and practical applications in diverse human–AI interaction contexts.