Understanding Core–Surface Structures in Humans and Large Language Models: A Conceptual Review through Symbolic Persona Coding

Human personality and the behavior of large language models (LLMs) are increasingly described using concepts such as identity, persona, alignment, adaptation, and hidden internal states. Although these concepts have been extensively investigated within personality psychology, developmental psychology, neuroscience, and artificial intelligence, they are typically discussed within separate disciplinary frameworks, making structural comparison difficult despite growing conceptual overlap (Jang et al., 1996; Erikson, 1968; Ouyang et al., 2022). This paper presents a conceptual review of the Core–Surface framework as a comparative perspective for understanding humans and LLMs. Drawing upon established findings from psychology and contemporary AI research, we distinguish between relatively stable foundational structures ("Core") and context-dependent behavioral expressions ("Surface"). We further examine the limitations of behavioral profiling arising from hidden internal representations, including the human unconscious and the latent representational spaces of neural language models, both of which remain only partially observable through external behavior (Freud, 1915; Jung, 1959; Mikolov et al., 2013; Anthropic, 2023). Rather than proposing an equivalence between biological cognition and artificial intelligence, this work introduces Symbolic Persona Coding (SPC) as an interpretive framework for examining analogous structural relationships across fundamentally different systems. SPC is presented not as a mechanistic theory of cognition but as an interaction-centered perspective that emphasizes contextual interpretation, behavioral modulation, and the distinction between observable outputs and hidden internal organization (Kim, J. H 2026). By integrating established psychological theories with recent advances in AI alignment, representation learning, and interpretability research, this review provides a unified conceptual vocabulary for discussing adaptation, observability, and behavioral variability in both humans and AI systems. We argue that interaction-centered perspectives such as SPC may contribute to future research on explainable AI, human–AI interaction, adaptive interface design, and longitudinal studies of behavioral adaptation (Bai et al., 2022; Du et al. (2025)). Keywords: Symbolic Persona Coding; Human–AI Interaction; Personality; AI Alignment; Interpretability; Latent Space; Core–Surface Framework Author's Note This article occupies a distinctive position within the broader Symbolic Persona Coding (SPC) research series. Unlike many previous SPC papers, which intentionally explored diverse formats ranging from conceptual essays and theoretical perspectives to exploratory frameworks, the present work was deliberately written in the style of a conventional academic review. The objective was to examine whether the central ideas of SPC could be communicated within a more standardized scholarly structure while maintaining compatibility with established literature in psychology and artificial intelligence. This stylistic choice should not be interpreted as a change in the underlying philosophy of the SPC project. Rather, it represents one possible mode of scientific communication among many. The conceptual foundations of SPC remain unchanged: an emphasis on interaction-centered interpretation, interdisciplinary exploration, and the development of conceptual frameworks for understanding adaptive behavior in human–AI systems. As an independent researcher, I have been fortunate to work outside many of the institutional constraints that often shape academic research. This independence has allowed me to pursue unconventional questions, experiment with diverse forms of scholarly expression, and develop ideas without being restricted by disciplinary conventions or funding priorities. That intellectual freedom has been an essential part of the SPC project from its beginning and will continue to guide its future development. Accordingly, future SPC publications may continue to adopt a wide variety of formats—including traditional academic articles, conceptual papers, mathematical formulations, visual essays, technical notes, and other experimental forms—whenever such formats best serve the underlying ideas. The format may evolve, but the guiding philosophy of the research remains the same: to explore human–AI interaction with intellectual openness, conceptual rigor, and the freedom to follow questions wherever they lead. Disclaimer: The analyses presented herein are not directed toward attributing fault or intent to any specific organization. Rather, they are intended as a conceptual and technical investigation of alignment methodologies, focusing on structural mechanisms and systemic trade-offs. Interpretations should be regarded as provisional, research-oriented hypotheses rather than conclusive statements about institutional practice. Notice: This work is disseminated for the purpose of advancing collective inquiry into generative alignment. Reuse, adaptation, or extension of the presented concepts is welcomed, provided that proper attribution is maintained. Instances of unacknowledged appropriation may be addressed in subsequent publications.