| 双主体驱动的人机复杂偏好融合及应用 |
| The Method and Application of Dual-Agent Driven Human-Machine Complex Preference Integration |
| 摘要点击 101 全文点击 0 投稿时间：2024-08-22 修订日期：2026-04-03 |
| 查看/发表评论 下载PDF阅读器 |
| 中文关键词 人机偏好；人机协同；偏好融合；混合智能；人机决策 |
| 英文关键词 human-machine preference; human-machine collaboration; preference fusion; hybrid intelligence; human machine decision-making |
| 基金项目 国家自然科学基金项目（面上项目，重点项目，重大项目）; 中国工程院重大咨询项目 |
| 投稿方向 群体决策，人机协同 |
| 作者 | 单位 | 邮编 | | 孙妍妍 | 中南大学商学院 | 410083 | | 周艳菊 | 中南大学商学院 | | | 徐选华* | 中南大学商学院 | 410083 | | 郭栋炜 | 中南大学商学院 | |
|
| 中文摘要 |
| 针对人机偏好特性差异融合难问题,提出一种基于人机双主体协同决策驱动的复杂偏好融合方法. 首先, 构建人机对等协作决策框架, 设计包含人类和机器决策属性的多层混合属性结构. 其次, 通过机器学习技术实现机器偏好的多层次表达, 基于群体决策理论实现人类偏好的聚合, 并引入人机信任度建立混合偏好融合模型. 然后, 提出一种考虑人机独立和整体权重的混合属性权重优化方法, 实现人机复杂偏好集结. 最后, 以美国实体清单制裁风险评估为例进行验证, 结果表明, 方法能够保持人机决策优势互补, 准确反映卡脖子企业风险等级, 相较单一智能模型结果更优, 能为提升混合智能决策系统的科学性和可靠性提供有效的方法支撑. |
| 英文摘要 |
| To address the challenge of integrating human-machine heterogeneous preferences, this paper proposes a complex preference fusion method driven by dual-agent collaborative decision-making. First, a peer-level cooperative decision-making framework is established, featuring a multi-layered hybrid structure that integrates decision attributes from both human judgment and machine intelligence. Then, machine preferences are hierarchically modeled via machine learning, while human preferences are aggregated using group decision-making theory. A hybrid preference fusion model is further developed by incorporating a human-machine trust metric. Additionally, a hybrid attribute weighting method is proposed that considers both the independent importance of human and machine agents and the global significance, thereby enabling the integration of complex preferences. Finally, a U.S. Entity List risk assessment case study shows that the proposed method leverages the complementary strengths of human and machine intelligence, accurately reflects the risk levels of bottleneck enterprises, and outperforms single-agent models. This approach provides effective support for enhancing the scientific rigor and reliability of hybrid intelligent decision-making systems. |
| 关闭 |
|
|
|
|
|
