【作者】 王慧贤；

【导师】 吕廷杰；

【作者基本信息】 北京邮电大学 ， 管理科学与工程， 2013， 博士

【摘要】 随着互联网全面进入web2.0时代,社交网络(SNS)在世界各国都经历了一个飞速发展的阶段,用户规模和市场规模都呈现出爆发式增长的趋势。越来越多的人通过社交网络构建并维持社会关系,SNS已经成为最受欢迎、增长最快的网络应用之一。作为一个网络时代的新媒体,SNS对于网络社会以及现实社会的影响力是不言而喻的。然而,如同其他大多数网络应用一样,SNS在其发展过程中仍然面临着一系列亟需解决的问题和挑战。首先,社交网络面临“用户流失”问题。从2011年开始,社交网络的发展速度逐步放缓。无论是传统的以“偷菜”等游戏为主导的社交网络还是新兴的微博形态的社交网络都面临着用户规模日益萎缩、用户粘性下降、访问流量降低的挑战。社交网站最大的吸引力就在于其较大的用户规模。用户规模越大,平台在广告投放、资本投资等方面的商业价值就越大。因此,社交网络平台运营商必须深入了解用户需求及影响用户参与行为的影响因素,通过提升网站服务满足用户多样化的需求,从而提升用户的社交体验,留住用户。其次,社交网络面临着用户“搭便车”的问题。社交网络区别于传统媒体最最重要特征就是：用户自生成内容(UGC, User Generated Content)。社交网站“公共资源池”中内容的多少,在一定程度上象征着整个网络用户的活跃度,进而代表着该网站的核心竞争力和潜在发展能力。然而,由于社交网络用户是完全凭借自愿性为网站贡献内容,这就导致了在现实中,很多用户只是被动地浏览其他用户在“公共资源池”中贡献的内容,却从不或者很少向主动向社区贡献内容。这种行为的存在可能导致社交网站公共资源池的内容越来越少,网站用户也随之逐渐流失。为了解决用户的“搭便车”问题,社交网络平台运营商必须探索有效的激励机制来影响这些自私的尾端用户,使其能够有强烈的兴趣贡献内容,从而提升整个平台的效用。再次,社交网络平台面临着大量低质、垃圾信息充斥的问题。社交网络信息发布快速、传播速度快、碎片化表达等特点使用得越来越多的人参与到社交网络媒体平台的信息生成和分享中。由于平台用户的素质良莠不齐,特别是一些“网络水军”的存在,使得社交网络平台成为了冗余信息或垃圾信息滋生、泛滥、扩散的重灾区,严重扰乱了正常的网络秩序。社交网络平台“公共资源池”中大量低质信息的存在与传播,在浪费平台资源的同时,也严重影响了用户参与社交网络的体验过程,为社交网络平台运营商带来巨大损失。因此,如何保证并提升媒体平台公共资源池中的信息质量,已经成为社交网络媒体平台所面临的一个挑战。最后,社交网络媒体平台面临着用户不活跃的问题。用户是社交网站的核心资源,用户之间的互动是社交网站良性发展的重要基础,也是广告商、投资商所看重的关键因素。内容评价,作为一种重要的用户互动行为,对于维持及提升社交网络的活跃度具有非常重要的作用。然而,在实际中,很多用户是“来无声,去无影”,他们登录社交网站只是被动地浏览内容,却并不积极参与互动,或者对浏览过的内容进行评价。在这样的情况下,贡献内容的用户由于得不到其他用户的反馈与支持,其贡献内容的积极性将逐渐降低,从而整个网络的用户活跃度将会越来越低,网站也将变得越来越“沉默”。因此,如何促进社交网络用户积极参与互动,对已浏览内容进行评价,也是平台网络运营商所面临的关键问题之一社交网络的用户是其商业价值的基础。超高的用户人气、频繁的用户互动,高质的平台内容,都能凸显社交网络在广告投放、第三方应用开发、增值服务提供等方面的价值。本文针对以上社交网络发展过程中存在的各个问题,进行了如下研究：(1)为了更加深入地了解社交网络用户的需求,提升用户参与媒体平台的体验,留住用户,本文在第三章采用结构方程建模的方法,研究社交网络媒体用户参与的影响因素。首先基于已有文献,提出了研究变量及研究假设,并在此基础上构建了社交网络媒体平台用户参与影响因素理论模型。然后,通过问卷调研的方法收集数据,并采用SPSS18.0和AMOS17.0统计软件,基于因子分析和结构方程模型等数据分析方法,对本文提出的理论模型进行了验证。研究结果表明：社会联系、娱乐收益、功能收益和直接网络外部性是正向影响用户参与水平的主要因素,而交叉网络外部性对用户参与水平具有负向影响；成本收益、社会联系和声誉地位是正向影响社交网络用户贡献水平的主要因素,而直接网络外部性和交叉网络外部性对用户的贡献水平有负向影响。(2)为了研究社交网络媒体平台用户贡献水平提升激励机制,本文第四章,基于博弈论的框架,构建社交网络用户贡献行为模型,根据用户贡献的动态行为过程,研究个体用户的搭便车行为。并基于委托代理理论,提出了用户贡献的激励机制,根据用户的贡献水平,给予用户一定的虚拟支付激励报酬,从而引导自私型用户贡献内容,提升用户对平台的贡献水平。研究结果表明：社交网络媒体平台用户选择“搭便车”行为是必然的；完全信息条件下,平台无需给予用户额外的激励报酬；不完全信息条件下,用户的贡献产出水平与其激励报酬相关；最优激励系数随着用户贡献产出系数的提升而提升、随贡献成本系数、风险规避系数的提升而降低；用户贡献的虚拟支付激励机制,通过给予用户与其贡献努力水平相关的激励报酬,能够鼓励用户提高贡献概率,在一定程度上解决社交网络平台用户“搭便车”问题。(3)为了提升社交网络媒体平台用户生成内容的质量,抑制用户发布垃圾信息的行为,本文第五章在社交网络媒体平台现有声誉体系的基础上,引入审核机制,提出了一种基于声誉的用户内容质量审核机制,运用数学模型以及博弈论的知识,分析了不同审核概率对用户行为的影响,并据此提出了最优的审核机制,以期为社交网络运营商科学管理平台内容质量提供借鉴。研究结果表明：存在一个引导用户努力的审核概率下界,和一个引致用户不努力的审核概率上界。当审核概率高于引导用户努力的审核概率下界时,所有用户都会选择努力贡献；当审核概率低于引致用户不努力的审核上界时,所有用户都会选择不付出努力；审核概率会显著影响自私型用户的行为,不对称审核机制下,可能出现“声誉振荡”和“逆向声誉”的现象；引入审核机制的声誉系统的整体性能要优于无审核的纯声誉系统；在审核资源有限的情况下,当贡献型用户占比较高,且他们中的大多数拥有高声誉时,应尽量少审核高声誉用户,多审核低声誉用户。(4)为了解决用户不积极参与内容评价的问题,本文在第六章提出了一种内容推荐机制,通过该机制激励用户积极、主动地参与内容评价。通过该机制用户评价的数量和质量与社交网络服务提供方为其推荐内容的相关度休戚相关,用户只有积极主动地参与了内容评论,才能够获得与自己高度相关、符合自己偏好的内容推荐,从而降低自己的浏览社交网络的“公共信息池”中海量信息的时间成本。研究结果表明：社交网络媒体平台能够通过内容推荐的准确度影响用户浏览内容的时间成本,进而影响用户评价的数量和质量；社交网络系统中,社交群内朋友之间的偏好存在着相关性,一个用户可能与其社交群内的朋友具有相似的兴趣和偏好。本文的研究创新之处在于：(1)提出并验证了社交网络媒体平台用户参与影响因素结构方程模型,分别从“参与水平”和“贡献水平”横、纵两个维度对社交网站用户的参与行为进行考察,并通过用户自身、用户关系以及平台环境三个维度考察用户参与社交网络的影响因素。(2)基于博弈论,构建了社交网络用户贡献行为模型,证明了社交网络平台个体用户“搭便车”行为存在的必然性。在此基础上,基于委托代理理论构建了用户贡献的虚拟支付激励模型,提出了平台对用户贡献的最优激励机制,并通过仿真模型验证了该激励机制的有效性。(3)将审核机制引入社交网络媒体平台的声誉系统,构建了基于声誉的内容审核机制,激励用户提升贡献内容的质量。(4)将推荐机制引入到社交网络用户参与评论的激励机制研究中,提出了一种基于声誉值测算的内容推荐机制,通过引导用户以较低的时间成本浏览其感兴趣的内容,激励用户更多地参与社交网络的内容评价。更多还原

【Abstract】 With the arrival of Web2.0, Social Networking Sites (SNS) around the world has experienced a rapid development, both of user scale and market size grow explosively. More and more users build or maintain social relationship through SNS, thus SNS has become one of the most popular web applications. As a new media the in the internet age, SNS has great influence on both of the web and realistic society. However, just as other web applications, SNS is facing a series of problems and challenges need to be addressed.Firstly, SNS is facing the problem of losing customers. Since2011, there has been a slowdown in the development of SNS, the user scale of which has shrank greatly. Whether the traditional SNS oriented by games or the emerging microblogs are faced with the challenges of shrinking subscribers, falling stickiness and declining traffic. The biggest appeal of SNS is their huge user scale. The larger the user scale is, the greater the business value of the SNS platform to advertisement and investment is. Therefore, the operators of SNS have to understand users’demands and influencing factors of user’s behavior in the platform, so as to satisfy users’diverse demands through the service improvement, promote users’ social experience, and retain platform user.Secondly, SNS is facing the problem of "free riding". Comparing with the traditional media, the most significant feature of SNS is user generated content (UGC). The amount of the contents in the "public resource pool" symbolizes the liveness of the SNS platform, and represents the core competitiveness and potential development ability in some extent. However, due to the voluntary of user’s contribution, most users choose "free riding", that is, browse the content in the "public resource pool" passively with less or without any contribution,"Free riding" may result in the decrease in the contents of the "public resource pool", which will in turn result in the user churn. To solve the "free riding" problems, operators of SNS have to explore effective incentive mechanism to influence these selfish end users, so as to enhance their content contribution interests, and improve the utility of the whole platform.Thirdly, there are mass low-quality and junk information in the "public resource pool" of the SNS platform. SNS is faced with the problem of mass low-quality and junk information. With the arrival of Web2.0, SNS has experienced a rapid development, and has become one of the main media platforms to get information. Meanwhile, many characteristics of SNS, such as fast information release, quick spread, and fragmented expression, attract more and more user to generate contents and share information. SNS has become a new channel for the public to express interest appeal, and its social influence is enhanced gradually. However, the feature of "self-media" and the widely various users’ quality, especially the existence of "water army", make the SNS platform as the "hardest-hit areas" with many redundant or junk information, which disturbs the normal network order seriously. The existence and spread of mass low-quality information in the "public resource pool" result in the serious waste of network resource, and also influence user experience, lower user’s trust level to the platform, and bring huge loss to the network operators. Therefore, how to guarantee and improve the information quality in the public resource pool has become a challenge for SNS platform.Last but not the least,"inactive" is also a serious problem SNS faced with. Users are the core resource of SNS, interaction between users is the base of the benign development, and the key factor advertisers and investors. The more frequent the interaction between users is, the more business value the SNS has. Content review, as an important interaction behavior, plays a significant role in keeping and improving liveness of SNS. However, in fact, most users are "come soundlessly, and go without any trace". These users log in SNS only to browse content other users generated, but do not interact with others, especially, do not leave comments to the contents they have browsed. Under this situation, users who have contributed content could not get the feedback and support of other users, so their enthusiasm to contribution will be reduced largely, thus the liveness of the whole platform will lower and lower, and the network will become more and more silence. Therefore, how to improve the interaction degree and attract more users to make comments is also a key problem SNS operators faced with.Users are the base of SNS’s business value. High popularity, frequent interaction, high-quality content, can highlight the value of SNS in advertising, the third-party application development, providing value-added services, etc, and also can improve users’stickiness and loyalty. To solve the above mentioned problems SNS faced with, this dissertation makes the following studies.(1) To understand users’demands more clearly and improve user experience, chapter3employs the structure equation model (SEM) to research the influencing factors of SNS users’behavior. Firstly, based on the existing studies, the variables and hypothesizes are proposed, based on which, a theoretical model of influencing factors of SNS users’behavior is established. Then, collects data through questionnaires, employs SPSS18.0and AMOS17.0, and adopts factor analysis and SEM to make an empirical analysis to the theoretical model proposed.Research results show that social relationship, entertainment benefits, functional benefits and direct network externalities are the main factors influencing users’participation level positively, but cross network externality has a negative influence on user’participation level; Additionally, cost-benefit, social relationship, reputation and status are the main factors that affect users’contribution level positively, but direct network externalities and cross network externality have negative effects on user’contribution level.(2) To study the incentive mechanism of improving contribution level of SNS users, chapter4establishes a model of user contribution behavior based on the game theory frame. Through observing the dynamic behavior process of SNS users, study users’"free riding" behavior. Then based on the agent-principle theory, the incentive mechanism of improving contribution level is proposed. According to this incentive mechanism, incentive reward in the form of virtual currency is assigned to users according to their contribution level, thereby, incentive users to contribute content, and improve their time and content contribution level.Research results prove that "free riding" is inevitable. SNS platform doesn’t need to give users additional incentive remuneration under complete information, but need to give them additional incentive remuneration related to their contribution level under incomplete information. The optimal incentive coefficient should increase with the output coefficient of users’contribution, and decrease with the cost coefficient and risk aversion coefficient. The incentive mechanism based on virtual payment, through giving users some incentive rewards related to their contribution level, can encourage them to improve their contribution probability, and solve the problem of "free riding" to some extent.(3) In order to improve the quality of the content users generated in SNS, and prevent users from releasing junk information, chapter5introduces an audit mechanism based on the existing reputation system, thereby puts forward a content quality audit model. With mathematical models and game theory, the influence of different audit probability to users’behavior is analyzed, according to which, the optimal audit mechanism is proposed, so as to provide references for SNS operators to manage the platform scientifically.Research results show that there is a lower audit bound to guide users to exert an effort, and an upper bound to lead users not to exert efforts. When audit probability is higher than the lower bound, all users will choose to contribute, and as audit probability is lower than the upper bound, all users will choose not to work. Audit probability dramatically affects the selfish users’behavior. Under asymmetric review mechanism,"oscillating reputation" and "reverse reputation" will appear. The performance of the reputation system with audit mechanism is superior to the pure reputation without audit system. Under the condition of limited audit resources, and user with high reputation account for a relatively high proportion, SNS platform should review users with low reputation more frequently.(4) To promote users to make comments to the contents they have browsed, chapter6proposes a content recommendation mechanism, through which, users are encouraged to make comments positively. According to this mechanism, the amount and quality of comments are related to the potential content recommendation offered by SNS platform. Only by providing more content comments, users can get content recommendations fit their preferences, thereby lower the time costs consumed in searching information complying with their own interests among the mass information of the "public information pool".The research results show that SNS platform can affect users’ browsing costs through the accuracy of the content recommended, so as to incentive them to provide more comments; Within the SNS system, there is a correlation between users and their friends in a certain social group, and it is possible that users within a social group possess similar interests and preferences.The innovations of this dissertation are as follows:(1) A structure equation model of influencing factors of SNS users’ behavior is proposed and verified. In the process of model establishment, users’participation behaviors are investigated from2dimensions, that is, participation level and contribution level. Furthermore, online time and visit frequency are employed to measure users’participation level, the quality and quantity of content generated are adopted to measure users’ contribution level. Meanwhile, influencing factors of users’participation behaviors are investigated through3dimensions, namely, users themselves, relationship between users and platform environment. Additionally, direct network externality and cross-network externality are introduced into the model to investigate the influence of user amount and advertisements to users’behavior.(2) Based on the frame of game theory, a contribution behavior model of SNS users is established to invested users’behavior. And a virtual payment incentive model is constructed with the agent-principal theory, and the validity of the incentive mechanism is verified by simulation, Many scholars have studied the "free riding" problem and its restrain mechanism in virtual communities. However, there is seldom researches study the reason and probability of users’contribution in SNS with mathematical models. This dissertation constructs a contribution behavior model of SNS users based on the frame of game theory, though investigating users’dynamic behavior process, it is proved that "free riding" are inevitable in the platform of SNS. Research results show that both of the sharing benefits and independent costs are the main reasons for free riding. Meanwhile, agent-principal theory is applied in SNS area for the first time in this dissertation. The virtual payment incentive model for users’contribution is established to study the optimal incentive mechanism for both risk-neutral and risk-aversion users under complete information and incomplete information. Additionally, through dynamical simulation model, the optimal incentive level is obtained, and the individual and overall completion rates of the contribution tasks of SNS are investigated, so that the validity of the proposed incentive mechanism is verified.(3) Introduce audit mechanism into existing reputation system of SNS, and establish an audit mechanism based on users’reputation to incentive users to improve their generated content quality. Existing studies on information quality in virtual communities are focused on the reality of users’comments and the reliability of merchants’commodities information in E-commerce sites, or the behavior of malicious nodes in P2P networks. Generally, existing studies, on the basis of game theory, restrain false information release through investigating the users’ historical behaviors to influence their reputation in future. Reputation system has been applied in SNS widely, however, generally, there is no audit mechanism introduced. The dissertation introduces the audit mechanism into existing reputation system of SNS, and proposes an audit mechanism based on users’reputation to incentive users to improve their generated content quality. Additionally, based on mathematical models and game theory, analyze the influence of different audit probability on users’behavior, according to which, propose the optimal audit mechanism to study the optimal audit resource allocation problem under limited audit resource.(4) Propose a content recommendation mechanism based on users’ reputation to incentive review to browsed contents positively. Interaction between users is the key factor influencing the success of SNS, in which, review of individual users to the contents they have reviewed is a main form of interaction. However, there is no literature to study the incentive mechanism of users’comments. This dissertation introduces content recommendation mechanism to the study of user review incentive mechanism in SNS, and calculates users’reputation value and relevancy between users with users’historical comments and existing comments of the potential contents. A content recommendation mechanism based on reputation is proposed to guide users browse contents they interested with lower costs, so as to incentive users to make more content comments. Through connecting the quality and quantity of users’content comments with the accuracy of the recommended content they received and the time costs they browsed, incentive users to provide more comments, so as to improve the liveness of the platform. Additionally, considering users in a social group may have the similar interests and preferences, the dissertation introduce influence of one’s friends into the reputation value calculation process, and extend the base reputation model to a social reputation model. Taking review records of one’s friends into consideration, a reliable reputation value of a potential content is obtained based on the relevancy between oneself and his/her friends, so as to incentive users to comment more.更多还原