刘云川是美国伊利诺伊大学香槟分校商学院(UIUC)终身教授，毕业于美国哥伦比亚大学，获市场营销学博士学位。他在管理学顶级期刊Management Science和Marketing Science上发表过多篇论文。他是华人学者营销协会的联合创办人，是海外华人会计、金融、管科、信管和营销五大协会的总协调人，是第七届中国市场营销国际学术年会暨中国创造展执行主席。
形式为市场和运营文献阅读方法和论文写作兴趣小组。帮助博士及硕士同学阅读、分析和研讨营销与运营领域的国际顶级期刊（例如Management Science, Marketing Science等）论文，并以这一等级期刊为目标，培养研究生如何发现管理中的新问题，并撰写国际顶级期刊论文的能力。同时也为已完成working paper的同学提供交流和讨论的机会。
Affiliating the Influencer with Influencer Encroachment，with 王璐莹（澳门挂牌之最完整挂牌2021级管理科学与工程专业博士生）
We study the effects of influencer encroachment on marketer management through word of mouth. An influencer encroaches on a seller's market by selling a substitutable product to followers. Influencer also posts the seller's product test resultsto followers and allows sponsorship from the seller for the seller's product promotion. We use Bayesian Persuasion to formulate the effect of the influencer's review on consumers' belief updating when the influencer is the seller's competitor, and therefore a game theory model to formulate the interaction between influencer and seller on influencing management. Our results suggest that influencer encroachment plays an important role in the seller's influencing marketing on sponsorship volume and product sales. Improved persuasion efficiency is achieved through the closer position of the influencer's product to the seller's product and therefore more intensive competition between the influencer and seller. Interestingly, the seller may prefer to sponsor the influencerwho has encroached on the market than the one who has not. In addition,compared with previous literature, the seller may benefit from a lower level of affiliation in the contract signed with the encroached influencer, which is because the influencer’s lossof conducting a sponsored test is decreased with the affiliation level, and it is more profitable for the seller to save promotional cost by specifying a lower level of affiliation even though the sales revenue decreases.
(“Un”)Fair Machine Learning Algorithms in Platforms, with 李亚萍（澳门挂牌之最完整挂牌2021级管理科学与工程专业博士生）
To address algorithmic bias, several fair machine learning algorithms that require equal impact (e.g.,equal opportunity) have been proposed. Compared with the prevailing algorithms that require equal treatment, fair ML algorithms are commonly perceived as “un”fair because they erode profits or welfare of certain parties and can never make everyone better off. Nevertheless, practical evidence from platform algorithms, such as LinkedIn Recruiter and Airbnb's smart pricing, counters the accusation of “un”fairness. To figure out why fair algorithms in platforms can make everyone better off, we develop a game-theoretic model where a platform, acting as an algorithm designer, provides algorithm services to a firm, functioning as the decision maker. The decentralization of algorithm designing and decision-making we considered has been overlooked in the existing literature on algorithmic decision-making. We find that whether fair algorithms are (“un”)fair (fair or “un”fair) in platforms depends on the misclassification cost. When the misclassification cost is low, the fair ML algorithms are indeed fair that make everyone better off. When the misclassification cost is high, the fair ML algorithms are “un” fair that they make certain parties worse off.