报告题目：Two Hands on the AI Flywheel: Data Collection and Incentive Contract Design for Algorithm-Application Synergy

时间：2026年1月10日上午9：30-11：00

地点：北理工长三院B座514

报告人：刘云川

报告人职称：教授

报告人工作单位：伊利诺伊大学

报告人简介：

    刘云川教授是美国伊利诺伊大学香槟分校(UIUC)终身教授，毕业于美国哥伦比亚大学，是市场营销、运营管理、供应链、信息管理、消费金融等方面的专家，他有很多文章发表在 Marketing Science 和 Management Science 上.刘教授是一名优秀的教师，他每年都名列伊利诺伊大学的优秀教师名单，曾被伊利诺伊大学 MBA 学生评选为年度教授最终候选人。刘教授现任 Decision Sciences 和 Service Sciences 副主编、曾任 Marketing Science 编委，多次获得 Management Science 和 Marketing Science 杰出评审奖，是华人学者营销协会的联合创办人，中国市场营销国际学术年会执行主席。刘教授也是伊利诺伊大学香槟分校商学院中国项目主任，正在推动伊利诺伊大学香槟分校和中国若干大学的合作项目。

报告内容简介：

Firms without in-house AI capabilities face the critical challenge of orchestrating synergy between outsourced algorithmic precision and application integration to activate the AI flywheel effect. Existing research largely adopts a single-task perspective and overlooks the inherent coupling of these dual tasks. To bridge this gap, we develop a dual-task, multi-stage agency model in which firms strategically manage synergy through dynamic data-collection policies and outcome- contingent contracts that balance AI flywheel acceleration with agent moral hazard across evolving data scales. Unlike insights from single-task models, we show that dual-task synergy fundamentally alters data-deviation patterns. Strong synergy amplifies complementarities between algorithmic precision and application integration, accelerating the AI flywheel. As a result, even when additional data weakens the marginal effectiveness of single-task effort, synergistic amplification shifts the firm’s optimal data scale upward. Furthermore, we find that firms' initial data strategies hinge on their AI level: a high level leads to myopic strategies, an intermediate level requires balancing payoffs with incentive costs, and a low level may induce aggressive data expansion when coordination effects are strong. While expanded data enhances task synergy and offsets future incentive costs, insufficient synergy growth can trigger superlinear incentive burdens, forcing firms to scale back. Interestingly, initial dual-task failures can be leveraged to justify paradoxical data expansion, transforming failures into assets of synergy. Our framework provides actionable guidance for AI development in autonomous driving, medical diagnostics, and financial technology, where aligning algorithmic advancement with real-world integration is essential.

（承办：管理工程系、科研与学术交流中心）