Adaptive Self-Selection and Collaborative Selection in Self-and-Other Models: A General Framework for Dynamic Selection-Mode Switching

Abstract

Selection is a fundamental operation in intelligent systems, computational systems, optimization processes, and model-based reasoning. Existing approaches commonly rely on independent selection, centralized decision-making, voting, ranking, or collaborative selection mechanisms. However, a system that consistently uses only one selection mode may fail to exploit the complementary advantages of independent evaluation and inter-model cooperation. This paper proposes an Adaptive Self-Selection and Collaborative Selection (ASCS) framework within a conceptual Self-and-Other Models architecture. The central idea is to allow a model to select an option independently when autonomous selection is sufficiently reliable and to engage other models when collaboration is expected to improve decision quality. The system adaptively switches between these modes according to uncertainty, disagreement, estimated selection quality, computational cost, historical performance, task characteristics, and resource constraints. The proposed framework is model-level and architecture-oriented rather than a new large language model. Its principles can be applied to heterogeneous computational models, optimization algorithms, predictive models, planning systems, knowledge representations, and other entities capable of generating or evaluating candidate options. The paper defines the conceptual architecture, mathematical formulation, selection policies, switching mechanism, learning process, evaluation methodology, limitations, and research hypotheses. It also distinguishes the proposed framework from related research on adaptive agent selection, ensemble decision-making, and collaborative multi-agent systems. The framework is presented as a research proposal rather than an empirically validated method. Its principal research question is whether a unified, adaptive mechanism for switching between self-selection and collaborative selection can provide a more favorable balance among selection quality, computational cost, robustness, and autonomy than fixed-mode alternatives.