基于记忆的技术竞争采纳三状态模型：替代机制、多归属与流失
A memory-based three-state model of competing technology adoption: substitution regimes, multi-homing, and churn
浏览论文内容
中文总结 AI 辅助
该研究提出基于记忆的三状态智能体竞争模型，可复现四类技术替代机制，揭示相似市场替代曲线或有不同微观成因，为关联技术替代轨迹与个体行为提供紧凑基线。
中文摘要 AI 辅助
技术、产品、平台及行为惯例常通过渐进式采纳、依赖强化的使用及临时多归属展开竞争。我们构建了一个均匀混合的基于智能体的三状态竞争模型，涉及在位选项（X）与挑战者（Y）。智能体分为X专属用户、Y专属用户及双采纳者（Z）三类。采纳基于记忆：专属用户仅在有限学习窗口内积累足够与采纳相关的接触后，才会转向另一选项；留存同样基于记忆：双采纳者仅在有限留存窗口内各选项均获得足够强化时，才会继续使用两者。该微观机制可复现与Adner-Kapoor技术替代四类机制（创造性破坏、稳健共存、弹性幻觉及稳健弹性）类似的总体使用特征，无需明确表征生态系统、互补者、价格或战略投资。从相同的小型挑战者种子出发，基准模拟仅在采纳负担、留存负担、采纳后使用偏好及双采纳者的“教导”作用上存在差异。滚动使用份额可复现四类总体替代模式，而状态解析轨迹与相图则揭示了不同的微观路径。因此，相似的市场层面替代曲线未必有唯一因果解释：尽管生态系统机制在诸多实证案例中可能至关重要，但仅有限记忆学习与留存即可产生定性相似的机制。该模型提供了一个紧凑基线，将技术替代轨迹与可观测的个体层面采纳、多归属及终止行为关联起来。
英文摘要
Technologies, products, platforms, and behavioral routines often compete through gradual adoption, reinforcement-dependent use, and temporary multi-homing. We formulate a homogeneous, well-mixed, three-state agent-based model of competition between an incumbent option (X) and a challenger (Y). Agents are exclusive users of (X), exclusive users of (Y), or dual adopters (Z). Adoption is memory-based: an exclusive user adds the alternative only after enough adoption-relevant encounters within a finite learning window. Retention is also memory-based: a dual adopter continues to use both options only if each is sufficiently reinforced within a finite retention window. This microscopic mechanism reproduces aggregate usage signatures analogous to the four Adner--Kapoor technology-substitution regimes---creative destruction, robust coexistence, the illusion of resilience, and robust resilience---without explicitly representing ecosystems, complementors, prices, or strategic investment. Starting from the same small challenger seed, the benchmark simulations differ only in adoption burden, retention burden, post-adoption usage preference, and the teaching role of dual adopters. Rolling usage shares reproduce the four aggregate substitution patterns, while state-resolved trajectories and phase portraits reveal distinct microscopic pathways. Thus, similar market-level substitution curves need not have unique causal interpretations: although ecosystem mechanisms may be essential in many empirical cases, finite-memory learning and retention alone can generate qualitatively similar regimes. The model provides a compact baseline linking technology-substitution trajectories to observable individual-level adoption, multi-homing, and discontinuance.
