人机协同决策中的偏好互适应演化：基于决策场论动态仿真

Preference Co-adaptation Evolution in Human-Machine Collaborative Decision-Making: A Dynamic Simulation Based on Decision Field TheoryRen Zongqiang1,2, Shen Xin1,2, Wei Da3

摘要:

人工智能的兴起让人机协同决策正成为未来组织决策新常态。如何在复杂人机交互场景中推动人机互适应智能决策成为当前研究热点。本文基于决策场论,构建了一个面向人机协同决策的偏好动态演化模型创新性地引入偏好耦合、注意力趋同以及信任不确定性的社会化双向反馈机制,克服了传统人机交互技术与规则驱动型方法(如规则、流程、博弈等固定框架)“间断式”互动的局限性,推动人机协同向“更连续”、“更高效”和“更创新”的方向演化。稳健性检验表明,模型在关键参数扰动和偏好序列检验中表现出良好的稳定性。并且模型仿真实验验证了人机协同比单一主体主导式具有更显著的可信度和高效能优势。此外,参数实验进一步验证了高响应双向反馈机制是驱动人机协同绩效提升的核心条件。

Abstract:

With the rise of artificial intelligence, human-machine collaborative decision-making is emerging as a new paradigm in organizational decision processes. A key research challenge is how to enable adaptive and intelligent co-decision-making in complex interactive scenarios. Drawing on Decision Field Theory, this study constructs a dynamic preference evolution model tailored to human-machine collaboration. The model innovatively incorporates mechanisms of preference coupling, attention convergence, and uncertainty in trust through a bidirectional social feedback loop. These additions address the limitations of traditional rule-based human-machine interaction frameworks—such as static rules, predefined processes, and discrete game structures—by promoting more continuous, efficient, and innovative collaboration. Robustness tests demonstrate the model’s stability under key parameter perturbations and preference sequence variations. Furthermore, simulation results confirm that human-machine collaboration yields significantly higher credibility and efficiency than single-agent decision modes. Parametric experiments further verify that high-responsiveness bidirectional feedback is a critical driver for enhancing collaborative decision performance.