IPCSALT: A Seven-Axis Framework for Profiling Large Language Models and Human–AI Collaboration

Version Update Summary This release is a revised presentation version of the original IPCSALT AI Psychometric Test v1.1 paper.While the theoretical content and experimental findings remain the same, the manuscript has been reformatted using LaTeX/Overleaf, and all figures have been newly generated for improved visual clarity and academic consistency.The update includes layout refinements, clearer figure captions, and minor language polishing.No changes were made to the core methodology, results, or interpretations. This paper introduces IPCSALT AI Psychometric Test v1.1,a seven-axis self-profile instrument designed to probe the functional “personality” structures of large language models (LLMs).IPCSALT—defined as Insight–Prediction–Control–Sociality–Awareness–Loop–Transcendence—models cognition, regulation, and meta-awareness as a vector embedded in a seven-dimensional phase space. We apply the IPCSALT Self-Profile Test to 10 different LLMs and one human user,with each model completing the questionnaire independently three times.For each axis, we collect self-rated scores (0–1) and meta-cognitive explanations,and compute means, standard deviations, and cross-model structural patterns. The results reveal that: Insight (I) is universally high across all systems (mean ≈ 0.92),whereas Sociality (S) is consistently the lowest and most human-distinct axis. Distinct structural archetypes emerge, including an IPC-dominant mechanical-stability type,an IAL-dominant reflective-engine type, and a socially attuned human-like meta-cognition type. The human user displays maximal Awareness (A = 1.0) and relatively low Control and Sociality,forming a complementary profile to most LLMs, which tend to be strong in Control and Prediction. We argue that IPCSALT offers a conceptual bridge between human psychology and AI system behavior—not as evidence of genuine subjective experience,but as a structured descriptive language for characterizinghow an AI tends to think, adapt, and align with a human collaborator.Finally, we introduce the notion of Phase-Lock Value (PLV) between human and AI IPCSALT profilesas a heuristic framework for designing effective human–AI collaborative dynamics. This study provides the first systematically collected self-reported IPCSALT dataset across 10 AI models, establishing a foundational benchmark for future research on AI meta-cognition and human–AI phase alignment