在线评论偏差对商家信誉和平台收益的影响及纠偏措施研究

Consumers refer to online reviews and reputation information in the reputation system to make purchase decisions during E-commerce transactions. However, underreporting and the manipulation of reviews make the deviation of review information and reputation information which will directly affect the choice of consumers. Based on the measurement of real reputation of retailers considering reviews bias, we further study the impact of review bias on the revenue of the third party e-commerce platforms. Then, we study how to manage review bias, so as to maintain the effectiveness of reputation system. In view of the silence bias caused by underreporting, a total probability model is constructed to measure the real reputation of retailers, and the influence of different factors on consumers' willingness to evaluate is empirically studied. The Bayesian inference model is constructed to estimate the real reputation of the retailer under review bias caused by manipulation. The third party e-commerce platform is regarded as two-sided-market. Under the framework of two-sided-market theory, game theory is applied to study the impact of different types of review bias on platform revenue. Adopting the incentive measures to solve the deviation in the existing research, this project pays attention to the incentive subjects.Using game theory model to decide who should pay for more truthful online reviews, and an experiment is designed to test the incentive effect by different subjects.

电子商务交易中消费者参考信誉系统中其他购买者发表的在线评论和信誉评分信息进行购买决策。但是自发评论提供不足以及商家为了自身利益操纵评论等问题使得评论信息和信誉信息出现偏差直接影响消费者的选择。课题在基于偏差的商家真实信誉的测度基础上进一步研究偏差对第三方电子商务平台收益的影响，最后回答如何管理偏差，以维护信誉系统的有效性。针对自发评论不足导致的沉默偏差构建评论的全概率模型估计商家真实信誉，实证研究不同因素对消费者评价行为意愿的影响进而对商家信誉的影响。针对评论不真实对应的操纵偏差构建贝叶斯推理模型对商家的真实信誉进行测度。将第三方电子商务平台视为双边市场，在双边市场理论的框架下采用博弈论方法研究不同类型的评论偏差对平台收益的影响。采纳已有研究中解决偏差的激励措施但是从激励主体角度入手，建立博弈模型确定激励主体。设计实验，实证检验不同激励主体下的激励效果。

为减少买卖双方的信息不对称，各大电商平台都提供信誉系统，系统收集了消费者购买产品后对产品或服务的评价然后聚合为商家的信誉评分，在线评论及其信誉信息可以有效的帮助消费者进行购买决策。然而实际中在线评论可能存在各种偏差，本研究探讨了两种形式的偏差：沉默偏差（消费者购买后不提供评论使得信誉系统中评论信息不完整）和操纵偏差（商家操纵评论使得评论信息无法反应商家产品的真实质量）。.以电子商务平台中两种常用的信誉集结模型为例，分析沉默偏差对商家信誉的影响。发现当存在沉默偏差时，信誉评分的有效性取决于消费者评论行为概率分布，给出了信誉评分有效性的条件。.深入分析商家操纵评论的成本和收益，构建博弈模型分析哪些商家以及商家在什么条件下可能对评论进行操纵。.构建平台、卖家、买家三方博弈模型分析评论偏差对平台收益的影响，对平台管理评论偏差提供建议。.本研究为平台管理在线评论提供理论依据，为提高信誉系统有效性提供决策支持。