HAIIA-L3: A Structural Model of the Self-Concept Enabling Engineering Realization

Recent large language models (LLMs) have reached a level at which they can generate long narratives and consistent characters. However, there is still no well-organized account of how such models implicitly maintain self-consistency, long-term motivation, and the persistence of values. In previous work, the author proposed PPFE (Primitive Psychological Feature Extraction) as a method for quantifying sentence- and clause-level affective fluctuations, making it possible to visualize local variations of emotion and motivation (L2)[1]. Through actual text analysis, however, it became evident that the emergence of emotions and the selection of actions cannot be explained solely by instantaneous affective states; rather, they presuppose a higher-level structure of self-concept. Based on this observation, the present paper proposes a self-concept model termed HAIIA-L3. L3 is defined as a superordinate layer that integrates individual events and affective changes (L2) and recursively maintains and updates self-identity across past, present, and future. The self-concept is not treated as a static representation, but is described as a dynamical system with a double-loop structure consisting of internal self-recurrence (λ_self) and intersubjective approval recurrence (λ_inter). When the balance between preservation and update in this double-recurrence system collapses, the fixed point of the self-concept undergoes an irreversible phase transition. This paper refers to this phenomenon as an Egophase Shift, and interprets it as a structural discontinuity analogous to stress accumulation and fault rupture in earthquakes. The goal of L3 analysis is not to predict such phase transitions at a single point in time, but to identify, in engineering terms, their probability and critical regions based on indicators of instability in the recursive structure. Such phase transitions of the self-concept are not mere state changes, but involve a reconfiguration of the very space in which the self is situated. As a coordinate system for describing this space, the present study reinterprets Maslow’s hierarchy of needs. While Maslow’s model has often been understood as a theory of the “contents” of needs, this work positions it instead as a coordinate system for placing the self-concept. On this basis, states that arise when transitions to higher levels fail are organized as a “dark-side structure,” represented by an inverted triangle. Phenomena such as loss, emptiness, and distorted compensatory behavior are rarely expressed as strong emotion words, yet can be understood as structural voids produced by the loss of relational anchoring. Although this study does not attempt to define consciousness itself, it provides an intermediate representational framework for human understanding by separating and linking language processing (L1), affective field computation (L2), and self-concept (L3). The proposed model offers an implementable and extensible foundation for integration with lightweight language engines, literary analysis, personality estimation, and dialogue systems.