ABSTRACT. 智慧选品是新零售的核心环节之一，但是现有工作严重依赖人工经验。即使少数研究提出了机器学习辅助方法，但结果的可解释性较差。为此本文致力于提出兼具预测精度与可解释性的基于SHAP模型的选品经验挖掘框架，并基于大型连锁零售超市609件商品的44万余条真实销量数据，检验了该选品经验挖掘框架的预测精度，提取了核心选品经验。该研究对于拓展智慧选品可解释性AI研究方法以及零售企业智慧选品实践具有重要的理论和现实意义。

ABSTRACT. The pervasiveness of multiple types of digital footprints recorded on e-commerce platforms have added fuel to the design of personalized recommender systems. Despite the abundance, consumers’ digital footprints can be confounded with many causes, both internally and externally. To disentangle the causes driving consumers’ behaviors, a causal recommendation method, i.e., DIPC, based on cause disentanglement at various consumption stages is proposed in the paper. Referring to related theories, interest and item popularity are recognized as causes driving consumer behaviors in the need recognition stage, while behaviors in the pre-purchase and purchase stages are assumed to be motivated by interest and conformity. To rigorously evaluate the performance of DIPC, extensive experiments are conducted on a real-world dataset with carefully designed protocol in terms of modeling multiple digital footprints and causality learning. The results show that DIPC outperforms all baselines significantly and possesses good interpretability, demonstrating the superiority of the proposed causal recommendation method.

ABSTRACT. 本文初步探索了航空装备故障知识图谱的构建与应用，面向航空装备故障诊断业务需求构建了故障知识图谱的模式层，基于深度学习为主、启发规则为辅的方法从结构化、非结构化的故障资料中抽取了故障知识，初步构建了某型装备的故障知识图谱。在此基础上开发了基于故障知识图谱的故障问答系统，能够较为精确地回答维修工程师的问题。实践表明，知识图谱可以为装备故障知识管理提供有效手段，进而赋能故障诊断。最后，对仍需解决的难点问题进行了讨论。

ABSTRACT. 随着直播电商蓬勃发展，AI虚拟主播逐渐进入直播间，目前较少研究关注虚拟主播与用户信任关系的建立。本文基于准社会互动理论，探究AI虚拟主播与用户互动中的人际吸引（社会、任务、外观吸引），以及主播的人格化程度如何影响准社会互动的建立，同时纳入情绪满足与认知促进，构建准社会互动视角下虚拟主播与用户信任建立模型。从电商直播情景丰富人与数字人交互信任理论，为直播商家提升用户信任度提供理论参考。

ABSTRACT. 疫情下客户数字化体验场景进一步增多，AI语音助手与不同应用场景的结合，创新了消费者体验。本研究基于认知情感意图框架，构建多阶段理论模型，分析了感知拟人性、感知关系性、感知娱乐性、感知使用成本、感知服务质量和情感变量对消费者采纳意愿和拒绝行为意愿的作用关系。本文有助于从认知和情感层面呈现消费者对AI语音助手的消费者虚拟体验影响因素和消费旅程过程机理，为企业利用AI语音助手增强消费者体验提供参考和借鉴。

ABSTRACT. Artificial intelligence (AI) services have developed rapidly in recent years, and these services have penetrated our daily life. Even though AI can provide better services than humans, people are dissatisfied with it. AI services may be subject to a bias, which refers to speciesism in this study. We propose that the AE design of AI is effective to attenuate people’s speciesism against AI services. We conduct a 2×2 experiment on 676 volunteers and find AE improves satisfaction by weakening speciesism against AI. This process is also moderated by the task types. The research find that AE can effectively weaken users’ speciesism in subjective tasks, thus improving users’ satisfaction, while these influences have little impact for objective tasks. The findings contribute to the research on AI services and the AE design of AI, and they provide a novel explanation about peoples’ unfair treatment of AI services.

ABSTRACT. 在人工智能（AI）辅助决策场景中，AI反应速度是其决策效率的体现，也是消费者在人智协同中最易识别和判断的重要特征。然而，现有基于拟人化视角所探讨AI反应速度与消费者行为偏好的研究产生了不一致的结论，这些结论也难以适用于当前广泛普及的AI辅助决策情境。基于线索利用理论，本研究拟通过3个实验研究探讨AI反应速度及其预期决策质量如何影响消费者的采纳意愿。结果表明，相较于低反应速度，高决策反应速度的AI有更高的消费者采纳意愿；其中，消费者的感知AI智能性和感知AI决策风险在上述影响中发挥了中介效应。此外，本文提出了预期决策质量作为内部线索对AI决策反应速度与消费者采纳意愿起到调节作用，认为预期决策质量对反应速度的影响起到“替代”或“放大”效应。本研究预期发现将从AI反应速度角度为消费者人工智能的算法厌恶或欣赏的有关结论作补充，丰富了AI辅助决策场景下反应速度与消费者采纳意愿的潜在机制和边界条件。

ABSTRACT. 本研究以心理健康咨询为场景，采用情景模拟实验法，探究社交聊天机器人虚拟形象的三种呈现方式（简单文本式、头像式和背景式）对用户自我披露意愿和行为的影响，并检验了社会存在感、私人自我意识和公众自我意识的中介效应。研究表明，背景式虚拟形象呈现方式对自我披露意愿有显著抑制作用，简单文本式和头像式虚拟形象呈现方式没有显著差异。同时，研究发现社会存在感和私人自我意识是两个有效的中介变量。

ABSTRACT. Service robots are gradually being deployed to replace employees and provide certain services. However, service failures inevitably occur due to technology malfunctions and the nature of service. Appropriate design solutions to increase customer tolerance of service failures are vital for service platforms. Based on the computers as social actors paradigm, voice stereotypes, and social cognition theory, we conducted three scenario-based experiments to explore how the human-like voice types of service robots influenced customer tolerance of service failures. The results indicated that robots with adult female or child voices led to higher levels of customer tolerance than those with adult male voices. Also, warmth perceptions, rather than competence perceptions, fully mediated the effects of robot voice types. Furthermore, for robots with adult male voices, increasing the level of robot intelligence enhanced users’ warmth perceptions and diminished the difference between the warmth perceptions of different voices. These findings contribute to the literature on robot design for customer service and offer implications for practitioners regarding service robot design and deployment.

ABSTRACT. Shocks caused by economic competition and the pandemic have made the business environment increasingly volatile, requiring employees to have high adaptive performance and proactive performance. As digital natives enter the workplace, they may utilize digital skills developed in their upbringing to possess an advantage when facing changes. However, studies on digital natives’ change-oriented performance are quite scant. Drawing on social cognitive theory, we hypothesize that digital natives raised under appropriate social guidance achieve adaptive performance and proactive performance using their digital skills of problem-solving and collaboration. We conducted structural equation modeling using data from a multi-source and Three-wave longitudinal survey to validate our model. This study enriches the understanding of the formation of employees’ digital competencies and change-oriented performance and suggests the value of digital natives raised with appropriate social guidance.

ABSTRACT. Video has become an emerging carrier of user-generated content (UGC). As a branch of self-media videos, product review videos posted by influencers have become an influential channel in shaping consumers’ purchase decisions. In the past, studies on influencer marketing have largely focused on textual and pictorial social media posts, while there was a lack of research on unstructured UGC videos, from which we can extract deep cognitive and affective information. In this paper, we constructed a theoretical model to explain the influence of product review videos on viewers' attitudes based on the cognitive-affective framework and authenticity theory. Specifically, we posit that viewers’ attitudes towards the product and the influencer will be influenced by both cognitive cues (i.e., information amount and topic diversity) and affective cues (i.e., vocal and facial emotional intensity) because these cues will affect perceived authenticity on influencers. In addition, such effects will differ across different types of videos (i.e., shopping-haul videos and anti-haul videos). A mixed-method approach was proposed to test the theoretical model. In Study 1, we collected product review videos from Bilibili.com, used text mining, sentiment analytics, audio analytics, and facial expression analytics tools to extract video features, and conducted regression analyses on viewers' attitudes (i.e., viewers' sentiment towards the product and the influencer). We further proposed a lab experiment to establish causality between video features and viewer attitudes, and to elucidate the mediating role of influencer authenticity. The contribution and implication of the research are discussed.

ABSTRACT. Though online retailers have put much efforts to optimize product image to attract consumers and increase revenue, studies about how specific image features promote product sales are rare. This study aims to examine whether image features have impact on product sales, and identify a set of key features essential to promote product sales. Specifically, we collected product images and corresponding sales data from one of the leading online food delivery platform in China. We used image processing technology to obtain image global features and regional features, and conducted a cross-sectional analysis and a difference-in-difference-in-differences framework to examine the effect of image features on sales. Results show that image saturation, colorfulness and the difference of contrast between foreground and background positively influence product sales, while image contrast has negative effect on sales. This study sheds light on the effect of image features and provides practical implications for product presentation

ABSTRACT. Technology-mediated learning (TML), a branch of online learning, has become an important tool to advance teaching and learning during the COVID-19 pandemic in a variety of schools that could not start on schedule. In contrast to face-to-face offline instruction, technology-mediated distance learning lacks actual instructional scenarios and supervision from the teacher, and students' self-regulated learning (SRL) dominates the instruction. There are few coherent studies on the influencing factors and outcomes of SRL. This study adopts the research idea of proposing an understanding of SRL from the perspective of influencing factors and outcomes based on social cognitive theory and anticipates a mixed research approach to determine the personal and technology factors that influence SRL, and the outcomes generated by SRL through semi-structured interviews. In the consideration of personal factors, this study considers emotions as moderators of general factors affecting SRL based on the framework of emotions. A questionnaire will be designed based on the analysis of interview data to validate the theoretical model of the influencing factors and outcomes of SRL in TML and to contribute to the enhancement of the effectiveness of SRL.

ABSTRACT. 信息市场是一种基于人群的预测工具，人机融合信息市场受益于人和机器互补的预测能力，提供更准确的预测结果。然而，市场中机器的身份披露可能会产生负面影响，机器亲社会行为的设计会减弱这种负面影响。本文研究机器收益分享设计对参与人活跃度（自愿参与预测任务的数量）的影响，具体关注收益分享的两个设计维度：人机交互目标结构（竞争、合作和无特定目标）和分配方式（按绩效分配和平均分配）。其中，前者表示机器的收益在什么条件或情况下分配，后者表示如何做收益分配。本文基于在线实验发现：第一，与人机合作目标结构相比，参与人在人群与机器竞争情境下活跃度更高。第二，引入机器分享收益设计后，参与人的活跃度降低。机器按绩效分配的收益分配方式会增加用户活跃度。第三，在人群与机器竞争的目标下，机器按绩效分配收益时用户活跃度更高。在人群与机器合作的目标场景中，机器平均分配其收益时参与人活跃度更高。除了上述三点关于收益分享设计对活跃度的影响，本文还发现在相对于人机竞争分享，人机合作时的感知机器威胁和感知机器能力更低，而感知机器温暖更高，结合人机合作分享时参与人活跃度低但决策质量高的结果，说明合作的人机交互目标改善了人机竞争交互带来的负面影响，营造了更有效的人机融合预测环境。本文的发现为人机融合预测中机器亲社会行为的设计提供了理论和实践贡献。

ABSTRACT. 如何优化在线有奖推介活动设计以获取更多新用户是商家和学者关注的重要话题。实践中，许多推介活动中出现了积分等媒介型奖励，然而这种新兴设计对用户感知与决策的影响仍知之甚少。基于动机理论和媒介相关研究提出，媒介型奖励会改变用户的社交动机和外在动机，对推介意愿的影响与其和推介对象的社交关系强度密切相关。三个实验的结果表明，当向弱关系发起推介时，媒介型奖励则会降低用户的感知社会风险，进而提升其参与意愿；当向强关系发起推介时，媒介型奖励会降低用户的感知推介价值，进而降低其参与意愿。研究结论有助于拓展对有奖推介活动、媒介研究和动机理论等多方面的认识，对于商家优化在线有奖推介活动设计具有指导意义。

ABSTRACT. 数智化新跃迁时代，随着人工智能教育的发展，机器人有望成为教师，有效解决教育资源短缺和不均衡的问题。现有对教育机器人接受度的研究只考虑了其智能性和有效性，然而教育不能只注重效率和智能。通过编码学生对教育机器人的接受动因和结构方程验证，结果发现，学生对教育机器人的智能性认可度较高，除有效性外影响学生接受的动因还包括技术相关因素：安全性，情感相关因素：融洽性、外观设计和有趣性，环境相关因素：社会影响。学生拥抱教育机器人的智能性和有趣性，却又因其有效性不足、缺乏情感支持和监管而止步。推动教育机器人的落地需要重视教学设计，关注情感支持，加强监管规范，鼓励人机协同教学。本研究为社交机器人的技术接收度模型提供了新因素——融洽性。

ABSTRACT. 本文通过Web of Science和中国知网数据库检索了2005年至2022年内的国内外关于人工智能服务代理(AISA)的拟人化影响消费者的86篇文献作为研究样本，对其进行了系统梳理。首先，本文回顾了AISA拟人化影响消费者的研究现状并总结了AISA拟人化的定义和类型。其次，本文构建了AISA拟人化影响消费者的研究框架。最后，本文提出未来研究应当关注AISA情感拟人化对消费者的影响、AISA拟人化带来的伦理和隐私问题、AISA拟人化对消费者福祉的影响、AISA拟人化在整个服务阶段中的动态变化机制、AISA拟人化影响消费者的其他应用情境等方向。

ABSTRACT. Multimodality is one of the important directions of AI-powered learning at present and in the future. Improving crystallized intelligence is one of the goals of AI-powered learning. In this study, we develop a theoretical model to explain the relationship between modality of AI and crystallized intelligence, based on the Biggs 3P model. We test the hypotheses using data from an experiment (N = 324) and found that modality of AI has a positive effect on deep learning approach and a negative effect on surface learning approach. The modality of AI can affect crystallized intelligence both directly and indirectly through deep learning approach. Moreover, we examine the moderating effect of technological readiness from learners’ perspectives. Optimism and insecurity have moderating effect on the relationship between modality of AI and two kinds of learning approaches respectively.

ABSTRACT. Deep synthesis can play a positive role in applications such as film production and virtual fitting. It also can generate ethical risks in malicious use such as defamation and fraud. Ethical risks cause public concern, but the concern is not evident in entertaining applications. Therefore, this study aims to understand the role of ethics and entertainment in the acceptance of deep synthesis applications. A mixed-method is used to qualitatively derive the ethical concerns and quantitatively validate the impact mechanism. We confirm that informed consent, privacy protection, traceability, and non-deception have significant impact on ethical acceptability and social acceptance, with privacy protection being the most sensitive. Perceived enjoyment significantly weakens the effect of ethical acceptability on social acceptance. The findings provide the entry point for ethical regulation of deep synthesis applications, and the weakening effect of perceived enjoyment is a wake-up call for regulators to guard against pan-entertainment applications.

ABSTRACT. With the growing popularity of voice assistant, they have revamped how people use technology and connect with brands conveniently. Empirical studies have revealed that factors motivating people to use voice AI assistant include many factors, but a common theme is their capabilities. We believe voice assistance's capabilities cannot be seen as a single construct, we focus specifically on the computational and cognition functional ability of voice AI assistant. According to the analysis of a real interaction data set from voice AI assistant——Xiao Ai classmate voice assistance—— company, we find that the capabilities of computational and cognition is positive to service quality, and higher service quality will make users feel more intelligence. with the expectancy violation, service success as a moderating variable to explain the relationship between computational capability and intelligence preference. This research contributes to the nascent literature on voice AI assistant in customer service and has managerial implications both for how AI assistance should be designed.

ABSTRACT. 在移动互联网盛行的当下，越来越多的广告主采用信息流广告进行广告投放。尽管消费者对信息流广告已经不再陌生，但是鉴于移动用户对隐私的担忧，广告平台还在使用隐性告知的方式进行广告展现。隐性和显性告知对消费者的接受意愿有什么样的影响？本文首次通过 2(推荐方式: 隐性告知/显性告知) X 2（努力差异：有/无）进行了实验室实验，并通过努力感知差异中介变量揭示显性与隐性的广告告知方式对消费者的影响机制。研究 1 揭示了显性告知对消费者的广告接受意愿具有积极的影响，信息流的告知类型（隐性和显性）影响消费者的努力感知， 相对于隐性告知，显性告知提升消费者的努力感知。研究 2 还考察了努力启发对努力感知的影响，从而对消费者广告接受意愿产生有中介的调节作用。

ABSTRACT. 本文采用适应性结构化理论，结合多案例研究，揭示智能信息系统医生使用行为的影响因素。主要研究发现医生对于智能信息系统的使用受到多方面因素的影响，包括算法的准确性和透明度，医生个体因素以及外界社会因素等。为加强医生与智能系统间的合作，目前的医疗AI交互设计应当寻求一种平衡，通过改进交互设计弥补医疗AI算法的部分缺陷，使医生能够普遍提高对于算法的使用度满意度，让AI更好地服务于智慧医疗。

ABSTRACT. 为了分析评论主题与学员满意度决定因素的失验效应，本文基于期望确认理论，以中国大学慕课上5,214个课程的93,679条在线评论为语料，使用潜在狄利克雷分布主题模型进行主题识别，跨学科比较了评论主题与学员满意度决定因素的关系。结果表明，学员对主题的关注程度和其对满意度影响的重要程度显著不同，并非所有从在线评论中挖掘的文本主题都会显著影响他们的总体满意度，技术科学课程评论和负面主题的失验效应更为突出。

ABSTRACT. 中药材新闻推荐是典型的垂直领域新闻推荐问题，为了有效运用领域知识，本文提出一种基于知识图谱的中药材新闻推荐方法。首先，从中药典籍中抽取实体和关系，构建中药材知识图谱；其次，通过基于注意力机制的知识感知卷积神经网络得到新闻嵌入表示；然后，基于用户历史点击新闻及侧面信息，运用基于注意力机制的深度神经网络，生成Top-K推荐列表；最后，通过真实数据集实验，验证了本文模型效果优于基线模型，且具有一定的可解释性。

ABSTRACT. Many organizations have adopted enterprise social media to improve employees' job performance. It has become common for employees to use enterprise social media to deal with work during non-working time. However, the impact of non-working time enterprise social media use remains unclear. Based on conservation of resource theory, this paper constructs a research model to study the relationship between enterprise social media and job performance at different times. Using structural equation model to analyze 323 employee’s data, the results show that: (1) Neither working nor non- working time enterprise social media use has significant effect on emotional exhaustion. (2) Both working time and non-working time enterprise social media use have promoting effect on organizational commitment. (3) Emotional exhaustion has inhibited effect on job performance, and organizational commitment has promoting effect on job performance. Finally, the theoretical and practical implications of this study are explored.

ABSTRACT. 为捕捉更全面的用户偏好、景点特征及评级与在线评论的交互信息以解决数据稀疏和特征融合问题，论文提出了一种混合3D卷积融合评级和在线评论的景点推荐方法：首先采用深度矩阵分解和细粒度情感分析分别从评级和在线评论中提取用户和景点特征，其次通过特征堆叠实现交互后输入 3D 卷积网络提取非线性信号，然后利用相似度计算，为用户推荐排名前N的景点，最后采用去哪儿网数据集进行验证。结果表明，该方法具有较优的推荐性能。

ABSTRACT. 当前地理标记照片广泛应用于个性化旅游路线推荐，但大多数研究尚未考虑这些照片的视觉内容。为此论文提出了一个考虑景点视觉信息的深度学习路线推荐模型，该模型利用卷积神经网络ResNet50提取视觉特征并构建用户视觉偏好，基于包含双向LSTM编码器和注意力机制LSTM解码器的改进Seq2Seq模型，采用beam search算法生成推荐路线。使用真实数据集对论文所提模型进行了验证，结果显示其表现优于其它基线方法，表明利用景点视觉信息可有效提高推荐性能。

ABSTRACT. Based on a natural experiment of online advertising in an e-commerce platform, this study empirically investigated the impact of display advertising introduction on the performance of native advertising, as well as the moderating role of merchant size and reputation. We found that after the introduction of display advertising, native advertising clicks, bidding and investment decreased significantly by 6.6%, 6.6% and 11.9%, respectively, while total advertising costs increased by 18.4%, indicating that display advertising and native advertising are partially substitutable. In addition, the negative impact of display advertising introduction on the effectiveness of native advertising increased as the reputation of the merchant declined, and this effect was strongest among regional chain stores, but not significant among national chain retailers.

ABSTRACT. Content creators, such as YouTube and TikTok influencers, attract growing attention from advertisers, which presents a novel tradeoff for the hosting platform. On one hand, allowing content creators to embed sponsored ads (CADS) in the content undermines the platform's own ad sales (PADS); On the other hand, the platform might benefit from CADS through commissions. We develop a game-theoretical model to examine this tradeoff, in which a group of content creators and their hosting platform compete with each other on ad sales. We find that the dual mode (i.e., allowing CADS) might be optimal for the platform, depending on the qualities of PADS and CADS. Unlike the literature where the relationship between the platform offerings and the third-party participation is exogenous, we show that the strategic relationship between PADS and CADS can be substitutable, complementary, or independent from each other, which is endogenously determined by the platform.

ABSTRACT. 本文从霍夫兰说服模型中信息内容的角度通过文本分析的方法对直播带货中主播的各种表达风格进行分类，然后基于不同类别的表达风格探究其对消费者购买意愿和社会临场感的影响，并探讨了社会临场感的中介效应及情境因素和商品类型的调节作用。本文采用真实直播间下相关数据来计算消费者社会临场，数据更加真实全面可信度更高。本文得到的结果可以完善传播说服理论在电商直播背景下的应用，从而更好地为电商平台和商家主播提供了实践指导。

ABSTRACT. 基于微信公众号研究突发公共事件各类利益相关者信息发布动因及情感极性，为相关部门有效引导舆情，构建良好舆论氛围提供对策建议。首先将突发公共事件微信舆情中的利益相关者进行分类，然后利用扎根理论对爬取及访谈获得的相关资料进行编码，基于社会交换理论与使用和满足理论对各类利益相关者信息发布动因进行分析，构建了突发公共事件微信舆情利益相关者信息发布动因模型。以“山东烟台栖霞金矿事故”为例对上述模型进行了验证，并进一步分析了利益相关者信息发布的情感极性，最后提出舆情治理建议。

ABSTRACT. 本文基于FISHBEIN合理行为理论建立研究模型，探讨了新冠疫情防控常态化背景下公众产生倦怠情绪的主要原因，以及公众倦态情绪对疫情防控行为的影响作用，并进行实证检验。研究结论表明：信息疫情对公众倦怠情绪有正向影响，规范性社会影响和信息性社会影响对倦怠情绪有负向影响；信息疫情负向影响防控行为，且全部通过倦怠情绪产生；规范性社会影响和信息性社会影响正向影响防控行为，且部分通过倦怠情绪产生；倦怠情绪对防控行为有负向影响。

ABSTRACT. 2022年3月，由新冠病毒奥密克戎变异株引发的疫情突袭上海。封控期间，上海接连发生生活资源不足、医疗资源稀缺、社区功能瘫痪等问题。传统求助热线无法正常运行，社交媒体就成为人们寻求帮助的重要渠道。本文以上海新冠疫情为例，以微博数据为研究对象，对新冠疫情爆发期间公众线上求助信息进行时空分析，探究其对危机治理的影响和作用，分析问题并提出针对性建议，为相关管理部门处理类似事件时提供科学的决策依据。

ABSTRACT. Social media is used as a front for online mapping of offline public opinion on public health emergencies, and multimodal information with images and texts becomes the main way of public sentiment expression. To make full use of the relevance and complementarity among different modalities and improve the accuracy of multimodal negative sentiment recognition of online public opinion for public health emergencies, we construct a two-stage, hybrid fusion strategy-driven multimodal fine-grained negative sentiment recognition model (THFMFNSR). This model comprises BERT and ViT-based multimodal feature representation, GCN-based feature fusion, multiple classifiers, and rule-based decision fusion. By collecting image-text data about COVID-19 from Sina Weibo, this paper verifies the effectiveness of the model and extracts the best sentiment decision fusion rules and classifier configurations. The results show that compared with the optimal sentiment recognition model with text, image, and image-text feature fusion, the accuracy of this model in sentiment recognition and fine-grained negative sentiment recognition is improved.

ABSTRACT. 本文选取中国大陆31个省级行政区的数据构建关系矩阵，利用UCINET软件进行网络图谱、中心性、块模型、差异性和相关性分析，借助二次指派程序分析地域及疫情的调节作用。研究发现区域民间借贷风险网络存在风险中心，网络结构和位置特征均影响借贷风险传导。P2P平台表现出更明显的集聚效应，和实体借贷协同加剧民间借贷风险传导。东部省级行政区位于更核心的位置，疫情的调节作用不明显。相关部门应提前防范借贷风险，以进行公共危机管理。 We selected the data of 31 provincial-level administrative regions in mainland China to construct a relationship matrix, analyzed the network map, centrality, block model, difference and correlation by UCINET. Through the quadratic assignment procedure, we verified the moderating effects of regions and the COVID-19. Major findings include that there are risk centers in the regional private lending risk network and the network structure characteristics affect the transmission of lending risk; P2P (Peer-to-Peer) platforms show a more obvious agglomeration effect, and synergize with physical lending to exacerbate private lending risk transmission; the eastern provincial administrative regions are located in a more central position, and the moderating effect of the COVID-19 is not obvious. Relevant departments should prevent private lending risks in advance for public crisis management.

ABSTRACT. 本文基于农村亲族社团特性构建了农村社会网络，揭示了社会网络结构对农村传染病传播行为的影响。研究发现：不同的村庄社会结构下，传染病的爆发程度度不同。经过对比，分裂型村庄结果能够减小传染病的爆发程度。基于此，提出提升农村重大疫情应急管理的建议，为农村疫情防控提供理论支持。

ABSTRACT. This paper uses quasi natural experiments and breakpoint regression methods to conduct empirical research based on Weibo behavior data, aiming to explore the impact of major government prevention and control policies on public sentiment and risk perception. The study found that the city closure policy exacerbated the negative, anxiety, fear and anger of the public in Wuhan and Beijing, and improved the risk perception level of the public in both places; The policy of blocking residential areas not only alleviates the anxiety and fear of the public in both places, but also reduces the level of risk perception in Wuhan; The unsealing policy did not have a significant impact on the public sentiment and risk perception of the two places. The research results provide reference for the government to formulate prevention and control policies to stabilize public sentiment.

ABSTRACT. In this paper, we investigate a two-echelon fresh product supply chain composed of a supplier and a distributor. The expression of preservation effort and order quantity are obtained under three kinds of preservation investment methods, as well as the intrinsic correlation mechanism among these decisions. With the aim of maximizing profit, the optimal preservation decision is made. This study shows that, joint investment in preservation is always the best way for both parties. When the gap between wholesale price and produce cost is small, the joint investment of supplier and distributor can be approximately equivalent to the investment of distributor alone. Therefore, distributors should take the initiative to undertake the preservation investment in order to maximize profits, while suppliers should avoid undertaking the investment alone.

ABSTRACT. The liberalization of trade and globalization bring challenges as well as opportunities to companies, especially for the Small and Medium Enterprises (SMEs). In order to reduce time and costs, while increasing the quality of products, an efficient implementation of systems engineering process is necessary. However, even if a few systems engineering standards especially address SMEs, surveys show that SMEs hardly use formalized processes, resulting in poor process reproducibility and monitoring, among other drawbacks. Indeed, because of strong resource limitation, SMEs have difficulty to estimate any short-term benefit in spending time formalizing their processes. Against this background, this paper explores a methodology to support project managers in fostering the progressive deployment and standardization of engineering processes in SMEs starting from the perspective of their own experience and referring to the international systems engineering standards. Our proposal stands by the companies’ side to help companies to make their engineering processes evolve by progressively aligning them with the systems engineering standards without disrupting the entire organization and practices of the company.

ABSTRACT. Although mobile promotions are still a major marketing method in recent years, conversion rates for mobile promotions are not very high. How to improve the effectiveness of mobile promotion is still a challenge for company. Based on probability distance and construal level theory, this paper explores the influence of construal-matching effect on consumer repurchase behavior by using the large-scale field experiment in a low-cost airlines. The results show that the matching between low probability distance and high construal level has a significant negative effect on consumer repurchase behavior, while the matching between high probability distance and low construal level has a significant positive effect on consumer repurchase behavior. The differences of this result and the construal-matching effect are discussed. Furthermore, the moderating effects of probability distance and individual characteristics on matching effect also are discussed. This research has guiding significance for marketing, consumer relationship management, and customized service strategy.

ABSTRACT. 为了探究国际联合资助与国际合作发文间的网络演化关联关系，本文根据WoS数据库SCIE核心集中2009-2018年的论文数据构建国际联合资助网络和国际合作发文网络，使用QAP、ERGM和TERGM方法进行分析。研究发现，国际联合资助网络与国际合作发文网络具有较强的关联性；处于中心位置的资助机构与科研机构在网络演化中更容易形成新的合作关系；一个节点在国际合作发文网络中处于高度闭合的聚类中时，在资助网络中同样难以形成新的合作关系等结论。

ABSTRACT. 本文探究数字服务化中制造企业所需的数字化能力、数字化能力支持的服务化路径和影响结果。研究以动态能力为理论框架，从三个维度测量数字化能力：数字采集能力、数字分析能力和数字应用能力。来自246家制造企业的问卷调查数据支持：数字化能力通过影响支持产品的服务（SSP）和支持客户的服务（SSC）提升企业竞争力，且数字化能力通过提升SSP间接提升SSC。制造企业应运用数字化能力构建SSP，在该服务过程中进一步学习以构建客户服务方案。

ABSTRACT. 智慧品类规划是新零售的核心竞争力之一。但是现有规划方法往往关注单品类内商品需求的替代关系，却忽略了跨品类商品间的关联条件。为此本文结合机器学习和运筹建模方法，提出了考虑替代和关联条件的品类规划模型，通过18万实际门店小票数据，利用差分进化算法求解出包含已有商品和新品在内的各属性组合商品的实际需求，检验了模型的预测精准度和适用性。本研究对丰富智慧品类规划方法，赋能零售品类规划实践具有重要的意义。

ABSTRACT. 数字化能力是企业实现数字技术赋能数字化转型的关键。本研究深化了数字化能力的概念，提出了数字化能力的具体构成，分析了数字化能力之间的内在逻辑及其对企业绩效的影响。通过理论分析，数字化能力可以解构为四个重要维度：数据采集能力、数据集成能力、数据分析能力和数据应用能力。通过问卷数据分析，我们阐明了数字化能力的内在逻辑。研究发现，数字化能力可以帮助企业成功实施数字化转型，提高企业绩效。

ABSTRACT. A large number of players purchase superior virtual gear to increase their chances of winning a battle. However, the design of PvP games, particularly the design of the competitive balance among players, remains largely unexplored. We consider an optimal product line design problem that allows the developer to design the competitive balance. Results indicate that tournament design enables the developer to alleviate the well-known cannibalization effect caused by the release of free versions through multi-dimensional versioning, which renders the freemium strategy optimal for the game developer. Second, tournament design allows the developer to overcome players' budget constraints. Specifically, the developer can effectively monetize the participation of players who play the game for free by handicapping them through tournament design and charging the paying players a higher price. Counterintuitively, tournament design results in a Pareto improvement in the sense that it increases the developer's profits and benefits every player.

ABSTRACT. 现有研究大多集中分析同边和跨边网络效应单个因素对平台发展的影响机制，却忽视其协同作用。本文利用 24149 条 Airbnb 的房源数据，构建网络效应与房东质量对房东响应行为的影响机制模型。实证结果发现，跨边网络效应和房东质量均正向影响房东响应行为；同边网络效应对房东响应行为具有负向作用；房东质量对网络效应影响房东响应行为具有差异性的调节作用。本文丰富了网络效应的相关研究，同时有助于平台管理者针对性地优化治理策略。

ABSTRACT. A substantial body of platforms literature has primarily characterized two-sided platforms as markets with cross-side network effects (CNEs). Most literature investigates how to attract more suppliers to leverage CNEs but ignores that CNEs driven by different suppliers may be heterogenous. To fill this gap, we identify two distinct suppliers i.e., individual hosts and professional hosts with three key differences in a set of housing rental platforms and examine how CNEs are influenced by different suppliers. Our results suggest that the quantity of individual hosts (vs. professional hosts) has larger CNEs on the renter size. In turn, renter size has larger CNEs on the quantity of individual hosts (vs. professional hosts), while renter size has weaker CNEs on the quality of individual hosts (vs. professional hosts). These findings generate important implications regarding how two-sided platforms may better govern ecosystems with different participants.

ABSTRACT. 数字政府建设以来，政务界面渐成公众获取公共服务主要渠道，但政务界面的公共服务供给特征如何影响公众的感知绩效尚未明晰。本文从界面治理理论视角切入，结合感性工学理论，采用交叉学科方法分析28个省级微信政务小程序界面。研究发现，服务形象、服务供给和服务创新是影响政务界面公共服务感知绩效三个维度；服务形象和服务供给分别受政务界面“外观形态要素”与“功能形态要素”两类供给特征影响，服务创新则是受界面整体影响；地域风景版面较小、证照卡包按需添加、专题服务列表导航且设置标题列举服务、小标题长度适中、整体字体适中、不设置其他信息区、设置底部导航区可以提升服务形象和服务供给维度的公共服务感知绩效。本文加深了学界对政务界面的理解，深化了界面治理理论内涵，并提出政府应当构建清爽、弹性、人文的政务界面。

ABSTRACT. 日益增加的版权成本使得数字内容平台迫切需要对作品的用户留存能力进行进行更准确的评价，但由于内 容平台大多采用聚合订阅，导致单个内容的评价非常困难。本文结合某内容平台的真实数据，通过构建用户访问 的轨迹网络，得到作品的留存贡献分数及相应排名并开展留存预测，利用 XGBoost 模型结合可解释性机器学习领 域的 SHAP 方法揭示了特征与留存、特征与特征之间的非线性作用，相关结果对于数字内容平台的管理具有一定 的参考意义。

ABSTRACT. 本文采用多层线性模型，通过对Indiegogo数据的研究，探索了在“连续发起人层-项目层”的层次特征下众筹绩效的影响因素。主要研究发现是：项目质量信号（视频数、图片数）、项目交互行为（更新数、评论数）正向影响着众筹绩效；连续发起人经验（发起项目数、支持项目数）对众筹绩效存在着正向的跨层作用；连续发起人社会资本（总评论数）积极地跨层调节着图片数、项目评论数和绩效之间的关系，但在调节更新数和绩效之间的关系时是消极的。

ABSTRACT. 按需服务平台将供给和需求进行精准及时的匹配，而零工经济的发展改变了按需服务企业的用工模式。本研究基于双边市场理论，构建了垄断市场中，在雇员模式、混合模式和零工模式下的博弈模型，对零工经济下按需服务平台定价及用工模式选择进行研究，不仅明确了同边和跨边网络效应对平台的最优价格、需求规模、零工从业者数量和最大利润的影响，也发现了在不同情况下平台的最优用工模式，为按需服务平台的发展带来一些管理启示。

ABSTRACT. 内容平台通过短视频带货、直播带货等形式开启电商业务，通过佣金获得利润。在开展电商业务的过程中，内容平台会与电商平台合作构建联合渠道，也会自主构建自营渠道（如：抖音小店）以及同时运营两种渠道。本文构建主从博弈模型，分析内容平台在三种渠道模式间的决策以及决策对电商平台的影响。研究发现内容平台的决策与商家流量变现能力有关，且会对电商平台利润产生非连续的影响。

ABSTRACT. Cloud migration is a crucial step in enterprise digitalization. Enterprises migrate applications into the cloud for faster processing time, more access and lower energy consumption than deploying locally. However, the existing works do not consider the efficient utilization of cloud resources when maximizing the enterprises’ utility. In this paper, we propose a cross-layer cloud migration framework for efficiently allocating cloud resources to multiple competing enterprises. We divide the cloud migration system into two layers: enterprise layer and cloud layer. We introduce a Fisher market in economics to obtain optimal service allocation strategy and equilibrium price at the enterprise layer. In the cloud layer, we consider the optimal resource provisioning and formulate the problem as a convex function. Then, we propose a gradient-based algorithm for cross-layer resource allocation. Finally, we verify the effectiveness of the algorithm by simulation.

ABSTRACT. 本文研究大数据环境下突发事件应急管理的技术赋能与使能价值，揭示突发事件应急管理技术赋能与使能价值创造之间的关系，以提升突发事件应急管理绩效。在使用文献计量学对相关领域文献进行元分析的基础上，首先界定突发事件应急管理技术赋能与使能价值创造的概念，阐述大数据环境下应急管理决策范式，进而构建突发事件智慧应急管理中从智慧应急基础“输入”到应急管理绩效提升“输出”的过程模型，并以地震灾害为例，对智慧应急的赋能与使能过程进行具体分析。本文通过探究突发事件应急管理从技术赋能到使能价值创造的过程，为优化突发事件智慧应急应用效果的研究提供了新视角。

ABSTRACT. 在线问答社区的持续发展取决于用户的持续知识贡献。然而，用户的知识贡献行为并非一成不变。本研究利用潜类别增长模型对在社区用户的知识贡献行为进行分析，刻画用户知识贡献行为的发展轨迹，并根据发展轨迹的异质性识别社区中不同类型的用户。研究结果有助于管理者了解问答社区中不同类型用户的特点，更有针对性地采取措施促进用户持续贡献知识。

ABSTRACT. 数字化时代，开放式创新平台为企业收集用户创意提供了新的渠道，领先用户能比一般用户更早的预见创新，因此领先用户识别是一种寻找有效创意的合理解决方案。本文以网络志为研究方法，通过对美创平台的深度观察，研究平台用户的特性，对领先用户进行了探索，并识别出了三种类似用户。本研究为企业识别领先用户，帮助企业快速识别有效创意提供了精准有效的方案。

ABSTRACT. Accurately excavate and identify the potential research topics and current research hotspots of the library and information subject, so as to grasp its development trend, dig out the core research hotspots, point out the characteristic research direction at this stage, and then promote the development of the library and information field research. This paper proposes a topic mining analysis model based on social network analysis and LDA. First, combine the co-word analysis algorithm to analyze the relationship between the subject words of different documents; combine the social network analysis algorithm to improve the coupling degree of the co-word network subject words; use the implicit space model to reduce the dimensionality of the co-word network to improve the coupling of social networks; Finally, the hidden location clustering algorithm is used to explore potential communities and improve the topic recognition effect. The method proposed in this paper optimizes the effect of the topic mining algorithm in identifying short text topics to a certain extent, eliminates subjective wishes, and is classified and predicted by the computer itself. Reveal the research hotspots in frontier fields, and provide a certain degree of reference value for emerging scholars who are committed to researching frontier disciplines.

ABSTRACT. [目的/意义] 从用户兴趣建模、知识服务机制、知识资源管理三方面揭示国内知识服务系统核心业务研究进展，以支持知识服务系统场景化用户兴趣建模、智能化知识服务机制混合、规范化知识资源管理，进而供数智化技术加持、嵌入式服务驱动的智慧服务系统模型构建参考。[方法/过程] 文章用内容分析法归纳了155篇文献内容，从用户兴趣建模、知识服务机制、知识资源管理三方面阐述了国内知识服务系统核心业务研究进展。[结果/结论] 用户兴趣建模研究最薄弱，多按照标签网络及其拓展的建模思路，选择模型表示方法揭示形式化用户兴趣框架但相关研究较单一，常混用显隐式方法采集用户基本、行为信息并通过数据预处理、兴趣度量化初始化用户兴趣模型，并用数据挖掘技术自适应进化用户兴趣模型以提升系统业务性能。知识服务机制研究亟待优化创新，多依托用户自主、浏览、提问、交互、检索行为提供知识定制、导航、问答、推荐、检索服务，集中于通过集成、优化服务机制实现知识服务定制化、导航可视化、问答知识化、推荐智能化、检索语义化以提升系统业务功能，尤以知识问答、检索研究居多。知识资源管理研究参差不齐，多依托业务需求采集、表示、组织、存储、推理、更新知识资源以创新集成、应用知识资源以提升系统业务效能，尤以知识采集、表示、存储研究居多。

ABSTRACT. [目的/意义] 系统分析国内基于知识图谱的用户画像研究成果以揭示其构建及管理、应用场景，以智能计算、语义挖掘用户特征进而支撑用户驱动型智慧服务 [方法/过程] 文章用内容分析法归纳了90篇文献内容，从架构体系、核心内容、实践应用三方面阐述了国内基于知识图谱的用户画像研究进展。[结果/结论] 知识图谱赋能的用户画像通过创新改进用户画像构建技术、机制以表示、识别、量化用户兴趣进而提升用户画像构建效能、性能。

ABSTRACT. 随着用户个性化需求的增加及付费观念的养成，社会化问答社区逐渐从免费走向付费，知识分享者给出见解的行为开始受到金钱激励的影响，并产生可以量化的商业效果。本研究基于面子理论分析线上知识分享场景中主讲人使用的礼貌策略对知识产品商业价值的影响作用。研究结果表明，在沟通中采用模糊限制语和施加压力的礼貌策略会对在线知识分享产品的商业价值产生积极影响，上述影响受到了粉丝数量的负向调节。

ABSTRACT. 作为当前最具影响力的数字加密货币，比特币的巨大价格波动一直是国内外学者关注的焦点。本文基于行为金融学理论，研究在比特币交易中用户行为变化对比特币价格的影响。为刻画比特币的用户行为，本文进一步运用复杂网络的相关理论，通过比特币用户交易网络结构特征演化反映用户交易行为演化。比特币用户交易网络是将实际用户之间的比特币交易映射成异构网络，它不仅是用户交易的网络化体现，而且通过交易网络结构特征可以发现在现实中不易观察到的用户行为特征的演化。最后，通过分析网络结构指标对比特币价格的影响，进而发现用户交易行为对比特币价格的影响机制。利用2015年1月至2019年7月31日间的比特币真实交易数据的实证研究发现：用户活跃度、用户囤币行为、用户交易强度、用户紧密度和机构投资者交易频率的变化对比特币价格波动具有正向影响。尤其当比特币价格处于波动期时，这种正向的影响更加显著。本文的研究也发现用户行为特征变化对比特币价格波动影响存在延迟效应，这使得利用用户行为特征来预测比特币价格波动具有一定合理性。本文的研究对数字货币价格机制具有一定的管理启示。

ABSTRACT. Academic social networking sites provide a simple avenue for researchers to exchange information. However, the mechanism by which question-related characteristics affect the answer outcome in academic social Q&A (SQA) is still unclear. This study focuses on how different combinations of question-related characteristics affect answer outcomes in ResearchGate based on a configurational perspective. The study adopts fuzzy-set qualitative comparative analysis (fsQCA) to analyze complex causal relationships among variables of question-related characteristics and answer outcomes extracted from the log under the “Artificial Intelligence” topic in ResearchGate. The results show that, when questions contain a high degree of emotion and the questioner is often involved in SQA activities, even if the question text is not highly readable, they will receive high-quality answers; while questions with low answer quantity and quality are often difficult-to-read or from novice academics. In addition, the study proposed a particular pattern of communication in academic SQA: emotion-led communication.

ABSTRACT. With increasing online review manipulation on various transactional platforms, a new merchant manipulation strategy arises – manipulating the reviews by inducing consumers to praise employees’ names in reviews and give high ratings. To explore the effect of this new manipulation strategy on eWOM, this study constructs machine learning models for average rating and rating distribution prediction, using review text related features and name-related feature reflecting the manipulation. Our results indicate that the manipulation of reviews does improve average rating of merchants but leads to sharper J-shaped distribution and low-quality reviews, which will reduce reference value and reliability of reviews, further weakening the positive impact of average rating.

ABSTRACT. 短视频快速兴起与发展的背景下，从顾客体验视角理解短视频场景营销对消费者购买意愿的影响因素与作用路径，是提升短视频质量，促进消费者购买意愿，发挥短视频场景营销价值最大化的重要前提。以S-O-R模型为基础，引入时效性加入自变量，构建短视频场景营销消费者购买意愿模型。依据问卷调查方法获取的475份数据，采用结构方程来验证假设及变量的净效应，采用模糊集定性比较方法（fsQCA）来探索整体视角下形成高消费者购买意愿的条件组态。

ABSTRACT. 基于识别复杂关联和区分点击事件考虑应用内具体行为类型的需要，本文提出基于希尔伯特-施密特独立性指标得到行为类型间有向无环图骨架，再根据行为类型所属的移动应用配合错误发现率(FDR)控制等策略得到应用间关联的组有向无环图骨架算法。本研究基于真实数据设计了一套模拟数据生成机制，比较了新方法相比基于聚合数据的相关系数和有向无环图骨架方法的效果差异，结果显示新方法可以很好地考虑组内细粒度变量的信息，识别出原本被掩盖或忽视的关联。数据分析结果发现，应用品类间紧密关联程度差异大; 音乐、视频等品类内应用间关联更松散; 行为类型层面，消耗用户更多时间与金钱的深度行为间更容易产生相互促进或竞争效应。

ABSTRACT. 本文采用文本挖掘的方法，以“携程问答”平台有关新冠肺炎疫情的旅游提问数据为例，探究疫情背景下问答平台用户的旅游信息需求类型，并构建旅游信息需求分类体系。研究发现，用户的旅游信息需求可以分为知识功能型、建议决策型以及娱乐休闲型三类，其中知识功能型主要关注受到疫情影响的旅游目的地运营情况、防疫政策以及疫情发展等信息，建议决策型则大多出于对旅行效率以及安全因素的考虑，娱乐休闲型更注重旅行的休闲趣味性。

ABSTRACT. Based on the Stimulus-Organism-Response (S-O-R) theoretical framework, this paper develops a theoretical model of how live knowledge features affect users' willingness to reward by influencing their viewing experience. The main findings are that the live knowledge feature variables of host professionalism, richness of materials, responsiveness, and quality of knowledge all have significant and positive direct effects on the reward willingness. The mediating variable social presence plays a significant and positive mediating role in the effect of all independent variables on the reward willingness. Cognitive load only mediates significantly and negatively in the effect of anchor professionalism and responsiveness on the reward willingness.

ABSTRACT. 2021年中国直播电商行业的总规模达到12012亿元，直播自带娱乐属性，通过全方位商品的展示，降低了消费者网上购物的信息搜寻成本，提高了购物效率，但其引发的冲动性购买也会导致消费者的购后不适。冲动性购买行为占据了网络购物份额的40%以上，在直播的过程中究竟哪些因素会影响消费者的冲动性购买行为，本文从技术可供性视角，通过实证研究，发现元发声、可视性、购物指导性对冲动性购买均有显著影响，并通过对消费者情绪的唤醒与愉悦促进冲动性购买，本研究将有助于指导商家更好开展电商直播。

ABSTRACT. There exists an algorithm paradox which scholars have still not paid attention to in the gig economy. This paper builds a moderated-mediation model to explore and validate the specific boundary condition and impact mechanism of such an algorithm paradox. Results based on 314 data collected from Chinese gig workers find that, algorithmic monitoring can enhance gig workers’ organizational identification, but through the mediating effect of job flexibility, algorithmic monitoring can indirectly decrease gig workers’ organizational identification. Criteria control can not only mitigate the negative relationship between algorithmic monitoring and job flexibility, but also mitigate the negative relationship between algorithmic monitoring and organizational identification mediated by job flexibility. These findings prove the reality of the algorithm paradox and discuss theoretical and practical implications about how to enhance gig workers to identify with the platform.

ABSTRACT. Consumer purchase prediction is very important for strategy formulation and consumer segmentation. In particular, airlines have to target groups before implementing the operational strategy with the competition increases. However, due to the higher dimensions and a large number data missing in airlines, how to effectively predict the repurchase of consumers has become an urgent problem. This paper proposes a novel ensemble algorithm framework for predicting the binary classification based on evidence theory. The results show that the accuracy of the ensemble algorithm based on evidence theory is more than 90% using the public datasets. Further, this paper uses this algorithm model to predict the repurchase possibility of consumers in airlines. The results show that the performance and accuracy of this algorithm are improved compared with common used ensemble algorithms. This study has guiding and theoretical significance for consumer purchase prediction, the application of evidence theory and precision marketing

ABSTRACT. 求职市场信息不对称以及低效的现状对职位推荐方法提出了更高的要求。一般商品推荐中最大化用户点击的目标并不适应职位推荐情景下互惠和竞争的特征，同时求职平台本身数据稀疏的特征进一步限制了现有职位推荐方法的效率。进一步的，职位推荐方法的可解释性决定了使用者对其的信任和理解，进一步影响了职位推荐方法在实践中的应用范围。为了解决以上问题，本文提出了一个可解释的双边异构图竞争迭代模型。首先，本文设计双边异构图整合来自求职者和职位的多源信息，通过图结点的连接和信息传递可以有效应对求职情境下数据稀疏的问题。其次，迭代模型以HR 的点击为目标，充分考虑求职者和职位的双边偏好，推荐满足互惠特性的职位。再者，模型中加入了一种竞争增强的策略，显式化地建模求职者在职位上的竞争热度和求职者的竞争偏好，通过两阶段优化实现对竞争热度的分散。 最后，本文在图网络中设计了多种粒度的注意力机制用于挖掘求职者的个性化偏好并实现具备解释性的职位推荐。本文在真实招聘平台数据集中进行了大量的实验，验证了推荐模型的有效性，鲁棒性和可解释性。

ABSTRACT. Deep learning (DL) models have significantly improved the performance of text classification and text regression tasks. However, DL models are often strikingly vulnerable to adversarial attacks. Many researchers have aimed to develop adversarial attacks against DL models in realistic black-box settings (i.e., assumes no model knowledge is accessible to attackers) that typically operate with a two-phase framework: (1) sensitivity estimation through gradient-based or deletion-based methods to evaluate the sensitivity of each token to the prediction of the target model, and (2) perturbation execution to craft adversarial examples based on the estimated token sensitivity. However, gradient-based and deletion-based methods used to estimate sensitivity often face issues of capturing the directionality of tokens and overlapping token sensitivities, respectively. In this study, we propose a novel eXplanation-based method for Adversarial Text Attacks (XATA) that leverages additive feature attribution explainable methods, namely LIME or SHAP, to measure the sensitivity of input tokens when crafting black-box adversarial attacks on DL models performing text classification or text regression. We evaluated XATA’s attack performance on DL models executing text classification on three datasets (IMDB Movie Review, Yelp Reviews-Polarity, and Amazon Reviews-Polarity) and DL models conducting text regression on three datasets (My Personality, Drug Review, and CommonLit Readability). The proposed XATA outperformed the existing gradient-based and deletion-based adversarial attack baselines in both tasks. These findings indicate that the ever-growing research focused on improving the explainability of DL models with additive feature attribution explainable methods can provide attackers with weapons to launch targeted adversarial attacks.

ABSTRACT. The openness and convenience of intelligent media have led to the disturbance or "noise" of the network environment during the spreading of Internet rumors, which adds uncertainties to the spreading. In previous studies, the Wiener process in probability theory is often used to describe the disturbances in the spreading of rumors. However, it has been shown that there are shortcomings in the random rumor spreading equation. To solve the problem, the uncertainty theory is introduced in this study. Specifically, we construct an uncertain rumor spreading model, and then calculate the analytical solution and the inverse uncertainty distribution of the solution. Finally, a numerical simulation is carried out with the rumor spreading data from WeChat, Weibo and other intelligent media platforms to verify the validity of the uncertain rumor spreading equation.

ABSTRACT. 消费是推动经济增长的三驾马车之一，社会消费品零售总额是反映社会消费总需求的关键指标。然而，传统统计预测方法尚未达到对社会消费品零售总额预测的理想效果。为弥补传统预测变量及预测技术的不足，本文基于深度学习长期和短期时间序列网络(LSTNet)，结合网络搜索数据与政府统计数据，构建LSTNet&BI模型开展浙江省及地级市社会消费品零售总额的预测研究，同时考虑多种基准预测模型进行对比分析。研究发现：(1)引入网络搜索数据能够有效提高LSTNet模型的预测性能与预测精度；(2) LSTNet&BI模型具有较好的泛化能力，对浙江省社会消费品零售总额的短期和长期预测效果都较稳定，其预测性能与预测精度均优于其余五种基准模型(LSTNet、LSTM&BI、SVR&BI、XGB&BI和ARIMA)；(3) LSTNet&BI模型具备较强的稳健性，其对杭州市、绍兴市和衢州市社会消费品零售总额的预测效果也较好。研究结果表明LSTNet&BI模型具有一定的实用价值，该方法为社会消费品零售总额预测提供了一种新思路，丰富了机器学习在宏观经济指标预测领域的应用研究。

ABSTRACT. The referral reward program (RRP) is a social marketing method by which firms reward existing users and encourage them to recommend products or services to their friends. Prior research has primarily focused on the impact of RRP design on users’ participation, however, the role of individuals’ characteristics is unexplored. Based on regulatory focus theory and self-efficacy theory, this research proposes and investigates the effect of users’ regulatory focus (promotion-focused vs. prevention-focused) on their referral intention and explores its mechanism and boundary condition. The results of three experiments show that compared to the prevention-focused user, the promotion-focused user has higher self-efficacy to complete the referral task; and thus, has higher referral intention. This effect will be attenuated when the tie strength between inviter and invitee is strong. The findings not only contribute to the research on RRP and regulatory focus but also can provide guidance for firms to optimize their RRPs.

ABSTRACT. Live commerce differs from the traditional social commerce in that influencers can engage potential buyers with instant impression management via the live channel. Our study draws on impression management theory and conceptualize influencer-product congruence as a novel construct that characterizes influencers’ impression management strategy in live commerce. We propose that perceived influencer-product congruence positively impact both functional and emotional value of the product, which consequently lead to purchase intention. In addition, male and female buyers display different purchase tendencies in response to different value perceptions of the product. We conducted a scenario-based survey with 239 participants who had recent live shopping experience on Douyin’s live commerce. The data analysis results supported our research hypotheses. Our study extends impression management theory in the live commerce context. The findings also yield practical insights for influencers and live commerce managers.

ABSTRACT. We aim to explore how listings owned by a multi-listing host affect each other, and the spillover effects of other listings’ online reviews on focal listing. Using a unique dataset of 1478 listings managed by 542 hosts ranging from October 2012 to June 2019 on Xiaozhu.com, we constructed two empirical models to examine the main effect of the number of other listings owned by the same host and the moderating effect of other listings’ online review on the focal listing performance, i.e., date difference between listing’s first sales and online date, and listing’s monthly sales. The results show that the number of other listings owned by the same host negatively affects listing performances, and online reviews have positive spillover effects. Our research contributes to both the home-sharing and spillover effect of online review literature and has substantial implications for the operators of home-sharing platform and multi-listing hosts on the platform.

ABSTRACT. This paper seeks to investigate the predictive value of investors' intra-day social media discussions on firm risk. We conduct a textual analysis of user-generated content (UGC) to extract the volume (e.g., number of posts, comments, views) and valence (e.g., helpfulness, sentiment, opinion, topic dispersion) characteristics. Various predictive models based on machine learning methods are utilized to evaluate the predictive performance of UGC. The results indicate that both UGC valence and volume can provide additional predictive gains over the baseline model. More importantly, UGC valence can significantly outperform UGC volume in predicting firm risk. Besides, we find that UGC during the non-trading period of a day shows significantly better predictive performance as compared with UGC generated during the trading time. This suggests that social media discussions generated in the non-trading hours are more informative and value-relevant to investors, whereas there is excessive noise on social media during the trading hours.

ABSTRACT. The booming of online platforms has attracted academia’s increasing interest in cross-platform spillover of product consumption. This study investigates how physicians’ content creation in Tik Tok influences patients’ demands, comments and satisfaction towards the physicians on online health communities (OHCs). Using the difference-in-differences approach, we uncover asymmetric influences of cross-platform spillovers for high- and low-awareness physicians in Tik Tok. Specifically, low-awareness physicians do not enjoy the benefits (i.e., the increased volume of orders and comments on OHC) from content creation in Tik Tok, but their ratings turn to decline due to attention distraction caused by cross-platform activities. Conversely, for high-awareness physicians, we find a positive cross-platform spillover effect for orders and comments on OHC without decreasing their ratings. Despite the existence of attention distraction from cross-platform services for high-awareness physicians, the negative impact on feedbacks is offset by higher ratings from their cross-platform consumers.

ABSTRACT. 本研究以日记法和访谈法收集数据，采用混合式主题分析法进行数据分析，深入探索了消费者对偶遇商品信息的利用行为。研究发现消费者对偶遇商品信息的利用是一种先以认知判断指导实践活动、后借实践活动修正认知判断、并最终实现价值创造的信息行为。该行为分为信息需求感知、信息利用方案设计、信息利用方案实施和信息利用后行为四个的阶段。充分利用偶遇商品信息不仅可提升消费者的单次购物收益，亦有助于拓展商家的利润空间。

ABSTRACT. Despite members is high social identity and more loyalty to a product or brand, the existing marketing literatures offers little guidance on how to promote members’ actual purchase behavior based on the matching consistency and construal level theory. To address this gap, the current research examines that the matching consistency effect between predicted purchase probability and the construal level of promotions advertising on the repurchase behavior of members based on a large field experiment in airlines. The findings show that the matching effect between promotional messages with concrete or feasibility information and high purchase probability has more likely to promote members' repurchase behavior, while the matching effect between promotional messages with abstract or desirability information and low purchase probability has not any influence on members' repurchase behavior. This study has theoretical and practical guiding for the researches on marketing and customer management.

ABSTRACT. 随着信息传递效率的提高，社会事件的影响迅速扩散，对市场的冲击也愈发强烈，把握住事件冲击带来的市场交易机会对市场参与者来说显得至关重要。为了在信息时代背景下更好的识别市场上产品的交易信号，本文从在线开放社区中识别热门产品，并结合产品序列间的格兰杰因果检验和滞后相关性分析，提出一个基于热门产品分析的交易信号识别方法，为企业管理者更好地识别产品的市场介入时机与运营策略构建提供重要的决策支持。

ABSTRACT. Emotional chatbot is a kind of chat robot with the ability of emotional analysis and expression. At present, the technology of emotional chatbot is still in its infancy, and the academia lacks systematic and targeted theoretical re-search on the factors affecting the acceptance willingness of emotional chatbot. Based on the expanded valence the-oretical model, this paper takes users' perceived risk of emotional chatbot services and system quality defects as users' negative potency factors, human interaction, emotional competence and perceived pleasure as positive potency fac-tors, and personal innovation, efficiency expectation and self-efficacy as subjective psychological factors. We discuss the factors affecting users' behavior intentions of emotional chatbot, and takes gender, occupation, education level as adjustment variables to explore the differences of behavior intentions of different user groups for emotional chatbot. The research results provide suggestions for improving and promoting emotional chatbot, and lay a foundation for related research in the academia.

ABSTRACT. Although AI-based chatbots have burgeoned in multiple business practices and demonstrated emerging potential in emotional support, little knowledge about the mechanism underlying the relationship between AI-based relational chatbots’ anthropomorphic attributes and user engagement. Drawing from the social judgment theory, we integrated agentic-style and communal-style to build relational chatbots’ conversational styles and proposed that emotional judgement mediated the relationship between conversational styles and user engagement with relational chatbots. Two types of AI avatars (humanlike vs. cartoonlike) are employed to examine the possible boundary effect of the above relationship. Two studies involving several online and lab experiments are planned to conduct to examine our hypotheses further. It is expected that we could extend the existing theories involving social judgment theory and human reactivity to AI-based chatbots in the setting of emotional care, and identify the potential joint effects of different types of anthropomorphic attributes of AI-based relational chatbots on user behaviors.

ABSTRACT. User-friendliness and intuitive interface fostered user participation in online health platforms. People know their health status anytime and anywhere and consult doctors online instead of going to physical hospitals. This study takes an online health platform as the research object, proposes and tests a model to explain users’ continuance intention to participate in online health communities through user engagement and perceived benefit, and tests the moderating effect of perceived vulnerability and perceived severity. The results show that the technology affordance characteristics of online health platforms affect users’ perceived benefit and online engagement, which further influence users’ continuance intention. Perceived vulnerability and perceived severity moderate the relationship between perceived benefits, online engagement and users’ continuance intention. This study has theoretical and practical implications for the online health discipline.

ABSTRACT. We examine the effect of social capital on participant retention in online mental health community, and disentangle the effect of bridging and bonding social capital on participant retention in this paper. Specifically, we derive participant profile data and activity data for 15 years from a Chinese online mental health community and construct social networks based on reply relationship for every half year. Following prior studies, bridging social capital and bonding social capital are measured by structural holes and network closure respectively. We conduct survival analysis to examine whether social capital has effect on participant retention, and use panel Logit model to capture the efficacy of different types of social capital. Results show that social capital significantly improves participant retention rate; bridging social capital has positive effect on participant retention, whereas bonding social capital has negative effect on participant retention

ABSTRACT. In the field of public health, suicide risk prediction is an urgent problem to be solved. According to psychological characteristics, it is valuable to consider users’ historical post in addition to current post for predicting suicide risk. Based on this rationale, this paper proposes a deep learning-based suicide risk prediction framework (DLDHI) considering dynamic historical information. To capture the dynamic and complicated information of historical post, this paper designs an improved unit based on long short-term memory. The importance and effectiveness of the prediction framework and its components are verified by comparison with the benchmark model and ablation ex-periments.

ABSTRACT. Covid 19 has killed people and significantly increased mental health conditions such as depression and anxiety. However, the cost of seeking professional help, limited availability of mental health service, and stigma associated with the disease is considered the most significant barrier to seeking help. Digital technologies such as social networking sites could bridge the gap by providing opportunities for patients to disclose their condition and seek social support. Nonetheless, Privacy risk perception has greatly hampered patients’ intention to seek help. Utilizing the privacy calculus theory, the study aims to fill this gap by providing the antecedents and their impact on patients’ risk perception of disclosing mental health information on social networking sites. The survey method will be adopted to conduct the study. Participants will be recruited through an online questionnaire on Facebook. The result will be analyzed using Amos software; afterward, theoretical and practical implications will be provided.

ABSTRACT. 本文使用中国领先的心理健康平台数据，实证研究不同自杀危机下的社会支持质量和社会支持数量对在线心理健康患者主动社会互动的影响。研究结果表明：支持质量会正向影响在线心理健康患者主动社会互动行为的产生，而支持数量会负向影响主动社会互动行为的产生。随着自杀危机的提升，社会支持质量的正向影响作用逐渐消失，而社会支持数量的负向影响逐渐消失，并最终对主动社会互动行为产生正向影响。

ABSTRACT. 本文以大学生群体为研究对象，基于自我调节理论与社会支持理论构建手机使用习惯对焦虑程度影响的研究框架，采用多层线性回归的数据分析方法，通过实证分析揭示大学生的手机使用习惯中关键的时间维度与功能维度指标对自身焦虑程度的影响，及其通过内外部因素进行自我调节的作用机制。主要研究发现是：手机使用习惯的时间维度与功能维度指标均会显著影响大学生的焦虑程度，其中手机使用时长对焦虑程度的影响呈正U型，具有非工作时段使用手机习惯的大学生个体焦虑程度更高，使用手机听音乐、运动等放松习惯有利于缓解个体的焦虑程度；社会支持中来自朋友的社会支持会对手机使用习惯与焦虑程度之间的关系有显著的正向调节作用，但来自家人的社会支持对此关系的调节作用不显著。此研究基于手机使用习惯视角为从内外部因素调整大学生的焦虑程度提供参考。

ABSTRACT. Understanding how government short videos promote citizens’ policy compliance in times of epidemic outbreaks is worthy of attention. Based on grounded theory, this study constructs a theoretical model of government short videos influencing policy compliance taking China's interview data as evidence. Results show that government short videos have significant effects on citizens’ policy compliance through four cognitive appraisals: epidemic, policy, government-public relationship, and social influence. In addition, resilience as an individual feature moderate the relationship between these cognitive appraisals and compliance. Findings contribute to the study of COVID-19 policy compliance and enrich the literature on policy compliance antecedents from the perspective of information sources and cognitive appraisals. In addition, this study provides targeted theoretical guidance and increases the understanding of how government short videos may be used to promote policy compliance during pandemics in and beyond COVID-19.

ABSTRACT. 政务数据是政府重要且有价值的资产，在保证数据安全条件下，充分挖掘政务数据这座“钻石矿”，有利于实现政务数据的经济价值和社会价值。笔者以政务数据价值化为研究对象，构建包括政务数据生产者、数据传递者、数据开发者、数据消费者四大节点的环形价值生态链，从数据价值和数据需求两个层面探讨政务数据生态链动力机制，并结合系统动力学模型，对政务数据价值化过程进行仿真模拟和灵敏度分析，系统地还原政务数据从生产、传递、开发、消费，最后到数据价值实现的全过程。研究发现，“数据价值”和“数据需求”构成驱动政务数据生态链良性循环的两大动力机制，政务数据价值化通过“数据需求—数据流转—生态链核心节点协同创造价值—数据价值变现”路径实现。其中，数据流转率、数据吸收整合能力、数据价值实现风险、数据需求程度等对政务数据价值实现具有显著作用。针对性提出以数据安全为前提，以数据需求为导向，发挥生态链核心节点主导作用，协同创造数据价值，构建政务数据价值化生态系统等建议，为打开政务数据价值化过程“黑箱”，释放政务数据价值提供理论指导和实践启示。

ABSTRACT. [研究目的]在数字治理快速发展的背景下，从多视角下理解与分析数字化工具服务数字治理过程中的负担水平，是提升政府数字治理能力与精度，促进公众可持续参与，发挥新时代下数字价值最大化的重要前提。[研究方法]通过基于TOE框架，构建技术、组织、环境因素下造成公众数字化工具负担模型，基于275份调研问卷，采用SPSS进行数据检验，并使用模糊集定性比较分析方法（fsQCA）来探索整体性视角下数字化工具增负效应的联动方式与路径。[研究结论]在单一条件下并不能构成数字化工具高负担水平的必要条件，但提升使用者的数字化素养在优化数字化工具使用效率发挥较为普适作用。根据结果，有“技术”“技术-环境”“技术-组织”“组织-环境”“技术-组织-环境”五种主导型模式发挥了显著性影响，即导致高数字化工具负担水平的驱动路径具有“殊途同归”的特点。在依据数字化工具掌握程度不同的跨组别研究中，发现不同组别对于高数字化工具负担水平驱动路径具有显著性差异。本文提出希望政府可以持续关注负担水平情况，合理规划财政资源使用，提升数字工具适配程度，创新现有数字治理指标绩效评价体系。

ABSTRACT. 综合分析我国适老化改造政策，梳理我国适老化改造政策演化历史，为不同学科研究人员、不同行业适老化研究改造以及未来的政策制定提供参考。本研究筛选出我国2002年至今共285份适老化相关政策，使用动态主题模型、共现分析和引文分析法分别用于主题演化、府际合作和关键政策识别，分析政策的主题、发文主体演化特征。我国适老化改造政策从阶段上可以分为无障碍建设、专项适老化、智慧适老化三个阶段，主要有无障碍公共设施改造、居家适老化、社区适老化、公共服务适老化、智能技术适老化、医疗机构适老化6个主题，目前重点为后三者。适老化改造政策的发文机构数量不断增加，协作体系多元化，改造面向的对象和场景从少数特殊向全面普及转变。

ABSTRACT. Older adults use digital intelligence to generate rich information practices during their travels, yet the specific shaping process of this information practice remains understudied. In current paper, we conducted in-depth interviews with 24 older travelers, using activity theory as a framework and grounded theory for exploratory coding of interview content. The results of the study show that the information practices of older adults involve eight core elements such as technology affordance, social support, and digital literacy; meanwhile, according to the differentiated information behaviors of older adults in different stages of travel, older adults' travel information practices can be divided into two distinctive patterns, information input and information output.

ABSTRACT. 城市社区是国家基层治理的基本单元，在公共卫生事件应急处置中发挥重要作用。为探索城市社区突发公共卫生事件应急处置能力影响因素，以结构功能主义为研究视角，使用扎根理论编码方法，借助质性分析软件Nvivo11 plus，从经济、政治、社会和文化四方面构建理论分析框架。为验证模型合理性，从全国28个省份获取到530份城市社区应急处置能力调查问卷，利用SPSS进行信度效度分析基础上，深入进行探索性和验证性因子分析，对模型构建的科学性进行验证。

ABSTRACT. Blockchain has been touted as a game changer in supply chains. While the technology can greatly improve the overall supply chain management, there are ethical challenges, and its ethical nature fundamentally implicates issues associated with sustainable development goals (SDGs). We are therefore yet to know the underlying ethical dilemmas as well as how to navigate the varied ethical complexities while creating value through blockchain-powered supply chains. Using the theoretical lens of normative ethics, this study detects various ethical complexities faced by blockchain in a developing country. The study identifies four types of ethical challenges, and proposes a proper balance between four types of ethical dualisms: balancing design efficiency with ease of use, balancing transparency with privacy, balancing open sharing with adequate data protection, and balancing user concerns with environmental concerns. The study offers rich contextual insights for developing countries and emerging markets, and has both theoretical and practical implications.

ABSTRACT. 大数据发展带来数据泄露等问题，用户信息披露担忧及隐私监管问题有待解决。本文基于隐私 关注视角，构建用户、APP 服务商和政府监管部门的三方博弈模型，进行演化稳定策略与数值仿真模拟， 探究各方参与隐私保护机制的倾向。研究发现：用户的信息披露行为是重要环节。用户披露与服务收益 及政府的信任收益有关，APP 服务商选择积极保护信息受政府部门补贴、成本的影响；政府监管隐私保 护带来的正面影响及用户信任、APP 惩罚金额影响其决策。

ABSTRACT. 在线旅行社(OTA)越来越被认为是扩大消费者基础的重要渠道。然而，目前尚不明确OTA渠道的引入对航空公司收入有何影响。在本研究中，我们基于一个准实验，结合倾向得分匹配和双重差分方法，从航空公司提供的同质和异构服务角度实证验证了OTA平台的引入对其收益的影响。结果表明，OTA的引入导致自有平台的同质化服务收益显著下降，同时对异构化(辅助)服务收益产生积极影响。我们的研究结果对平台开发、多渠道战略和收益管理具有启示意义。

ABSTRACT. 本文通过质性研究方法，通过挖掘2021年野性消费鸿星尔克事件后社交媒介舆论和电商平台的评论数据，证实了劣势者效应是野性消费鸿星尔克事件的内在机理。随机梯度下降模型的机器分类表明，共有54%的评论可以被正确分类到5个劣势者的情感动机内。其中，同情所占动机比例最大，其次为个人主义动机和激励动机，选择自由与怀旧动机占比最少。因此，中国消费者在国货意识增强的背景下，消费者识别出了鸿星尔克劣势者定位，且在多种情感动机驱动下产生支持鸿星尔克，并最终导致了此次野性消费事件。

ABSTRACT. 在区块链驱动的制造产业链数据交易背景下，针对企业主观因素导致的上链数据质量低下的问题，基于契约理论设计提高数据质量的激励机制。本文以数据隐私保护程度为企业私有信息，考虑企业数据上链的能耗成本和隐私损失，通过算例证明，不同类型的企业存在一组最优契约，在满足企业数据隐私需求的同时，使得制造产业链收益最大化。研究证明，设计的激励机制可以有效激励企业将更多的高质量数据上链，从而推进制造产业链协同创新发展。

ABSTRACT. 互联网时代，电子口碑主导着消费者购买决策，但商品溯源信息开通如何调节在线评论的影响呢？为弄清这个问题，基于京东商城中50款五常大米连续6个星期的短平衡面板数据，探究商品溯源信息开通、在线评论与产品销量之间的作用机制。实证分析表明：（1）在线评论数量及效价均显著提升网购农产品销量，且积极评论的促进作用更为明显，不同评论特征之间存在正向交互效应。（2）开通商品溯源信息的农产品热度较高，但其销量受评论效价影响并不显著，这意味着追溯技术应用有助于缓解口碑依赖。研究结论不仅可以为销售企业与电子商务平台有效管理农产品评论信息提供管理建议，而且对涉农龙头企业积极建立农产品质量安全追溯监管平台有指导作用。

ABSTRACT. 在大数据、人工智能蓬勃兴起的时代，平台经济应运而生，已成为助力商品生产、提升流通效率、集聚行业资源等的新型经济模式，同时针对其的监管更为重要。分析中国平台经济的运转模式，并对平台监管的问题进行剖析，发现仍存在理念滞后、体系分散、监管政策不适应等问题。最后在监管体系、强化规章主动性和前瞻性以及动态推动完整监管体系和创新监管工具等方面提出对应改进建议，能够更好地规范平台运营秩序。

ABSTRACT. With the development of online retailing on C2C e-commerce platforms such as Taobao, the loss of individual shops is becoming increasingly severe. How to retain high-quality sellers and enhance sellers' e-commerce platforms trust becomes an eager to learn for platform operators. Based on institutional trust theory, this study investigates how functional value, security value, and institutional value in structural assurance, as well as social media performance, perceived behavioral control, interactive value in situational normality can affect sellers' institutional trust. To test the hypotheses suggested by our conceptual framework, we collected 2970 valid questionnaires from individual sellers in Taobao, which is the most representative C2C e-commerce platform in China. The results show that functional value, security value, institutional value, perceived behavioral control and interactive value have a significant positive effect on platform trust, while social media performance has no significant impact on platform trust.

ABSTRACT. 移动健康平台已允许竞争对手在其平台销售同类型产品,但会优先为用户推荐其自营产品。现有文献忽略了平台具有市场支配地位，会采取优先推荐自营商品的销售策略。本文通过博弈模型分析了优先推荐自营产品的推荐策略对两者间价格竞争和合作关系的影响。研究表明，当自营和第三方产品的用户认知差异性较小时，合作后的二者均可从轻度的优先推荐策略中获益。但当认知差异性较大时，过度的推荐策略会导致第三方零售商拒绝合作。