中国网络赌博人群的精准识别： 赌博非理性认知为核心的多层面因素预测模型
文献类型：学位论文
| 作者 | 肖怡 |
| 答辩日期 | 2024-06 |
| 文献子类 | 硕士 |
| 授予单位 | 中国科学院大学 |
| 授予地点 | 中国科学院心理研究所 |
| 其他责任者 | 梁竹苑 |
| 关键词 | 网络赌博 赌博非理性认知 机器学习 赌博人群预测 中国样本 |
| 学位名称 | 应用心理硕士 |
| 学位专业 | 应用心理 |
| 其他题名 | Precise identification of Chinese internet gambling population: A multi-dimensional factor prediction model focused on gambling-related cognition |
| 中文摘要 | Gambling is an activity prevalent in human society, involving placing bets on the unpredictable outcomes of games or events. With the rise of the internet and mobile devices, internet gambling has gradually replaced traditional physical casinos, emerging as a new form of gambling that poses significant risks to individual physical and mental health, personal and national finances, and social norms. In recent years, internet gambling crimes have been on the rise in China. While targeted crackdowns on internet gambling criminal organizations (such as bookmakers) have shown some success, challenges such as regional jurisdiction limitations persist. Consequently, tackling internet gambling from the perspective of participants, by cutting off the financial and informational flows of internet gambling, is a realistic necessity and feasible strategy for regulating and governing internet gambling issues in China. One prerequisite for implementing this regulatory approach is the precise identification of potential internet gambling participants within the vast population. However, the existing methods for identifying internet gambling participants implemented by mobile payment platforms are relatively limited and lack empirical validation, making it even more challenging to achieve efficient and precise control objectives. The field of psychology has accumulated rich theoretical insights into gambling. Comprehensive theories such as the cognitive-behavioral theory suggest that various psychological factors at multiple levels, including cognitive, personality, and social factors, can predict an individual's gambling intention and behavior. Studies in the field of behavioral decision-making have also found that gambling-related cognition, involving a series of irrational cognitive distortions related to gambling, may play a critical predictive role in gambling. Research from different perspectives suggests that gambling-related cognition, may serve as a core variable in predicting individual gambling behavior. Drawing on existing research in the traditional gambling field, utilizing psychological measurement techniques and machine learning methods provides a methodological foundation for accurately measuring gambling influencing factors and predicting internet gambling populations. Currently, there is a lack of research in academia on predicting the Chinese internet gambling population. Existing studies have several shortcomings: firstly, at the theoretical level, there is a lack of comparative and integrated analysis of the predictive effects of gambling influencing factors and exploration of the relationships among these factors. Secondly, at the methodological level, measurements of internet gambling behavior lack ecological validity, and the accuracy of methods for predicting gambling is insufficient. Thirdly, in terms of applicability, there is a research gap in studies targeting the Chinese internet gambling population. To address the aforementioned shortcomings, this thesis proposes that selecting appropriate internet gambling influencing factors, establishing relationships between these factors and gambling intentions and behaviors, and using machine learning methods to predict gambling populations is a feasible research path. To clarify the influencing factors of internet gambling among Chinese residents and their predictive effects on internet gambling intentions and behaviors, this study proposes the following theoretical framework: 1) Factors at multiple levels (cognitive, personality, social, etc.) can positively predict gambling intentions and behaviors; 2) Cognitive factors related to gambling positively predict internet gambling intentions and behaviors, with gambling-related cognition being a key factor; 3) Gambling-related cognition, as a core variable, can mediate the predictive effects of cognitive factors (risk taking, delay discounting), personality factors (impulsivity, neuroticism), and social factors (social norm) on internet gambling intentions and behaviors. To validate the above theoretical framework, this paper conducted four studies in a sample of Chinese adults based on internet survey questionnaires and actual internet gambling behavior data: Study 1 focused on a sample of general Chinese individuals and distributed internet survey questionnaires in a pilot study (N=136) and a formal study (N=900) to investigate the predictive effect of six predictor factors (gambling-related cognition, delay discounting, risk taking, impulsivity, neuroticism, social norm) across cognitive, personality, and social dimensions on internet gambling intention, and gambling experience, while exploring the mediating role of gambling-related cognition. The regression analysis results indicated that gambling-related cognition, risk taking, impulsivity, and social norm were effective predictors of internet gambling. Moreover, gambling-related cognition mediate the influence of certain predictive factors on internet gambling. Study 2 aimed to validate the predictive effects of internet gambling factors on internet gambling intention and behavior. Selecting effective internet gambling predictors from Study 1, the study expanded the dependent variables to include actual gambling behavior and lottery purchases. In Study 2, an online questionnaire was distributed to a general population (N=866) through an online survey platform, and their actual internet gambling behavior data (such as annual gambling expenditure and frequency) were matched and obtained through Alipay backend. The results of the regression analysis found that gambling prediction factors had weak or non-significant predictive effects on actual gambling behavior, while these factors positively predicted lottery purchases. This suggests that the predictors identified in Study 1 have some predictive power for gambling behavior, yet their effects on actual gambling behavior were weaker in Study 2, potentially due to a smaller internet gambling population among general individuals. Study 3 delved into methodological exploration by utilizing machine learning methods among individuals with high internet gambling experience to explore effective prediction methods for the Chinese internet gambling population and validate the results of Studies 1 and 2. The study sampled individuals with high internet gambling experience (Anti-Gambling Vanguard Platform users) (N=1287), distributed internet surveys through the platform, and obtained actual internet gambling behavior data provided by the platform. The results of the regression and mediation analysis indicated that, consistent with Study 1, gambling prediction factors could predict internet gambling intention and behavior among individuals with high internet gambling experience, with gambling-related cognitions playing a mediating role. Subsequently, clustering algorithms in machine learning categorized the internet gambling population into four groups: low-risk, medium-risk, high-risk, and suspicious-risk individuals. By employing five machine learning algorithms including logistic regression, the AdaBoost algorithm model performed well in predicting internet gambling populations, achieving an overall prediction accuracy of 70% and effectively distinguishing among the four internet gambling groups. Notably, the prediction accuracy was highest for the medium-risk group (81%), with gambling-related cognition being the most crucial predictor in the model (shap value ranging from 0.17 to 0.34). As the sample populations differed across Studies 1 to 3, the predictive effects of each factor also varied. Therefore, Study 4 conducted a meta-analysis to validate and compare the predictive effects of the predictors across Studies 1 to 3. The meta-analysis results revealed significant positive predictive effects of gambling-related cognition, risk taking, and social norm on internet gambling. Moreover, the type of dependent variable (gambling intention/gambling behavior) and sample type (general population/special population) moderate these effects. For gambling intention, gambling-related cognition and social norm have a positive predictive effect; for gambling behavior, risk taking, social norm, gender (male), and debt have a positive predictive effect. In the generall population, gambling-related cognition, social norm, gender (male), and debt have a positive predictive effect; in the special population, gender (male) and debt have a positive predictive effect. In conclusion, this research found that: 1) in the Chinese social context, gamblingrelated cognition and social norm can predict internet gambling intention, while risk taking and social norm can predict gambling behavior; 2) gambling-related cognition mediate the predictive effects of other factors on gambling intention and behavior; 3) among multiple gambling influencing factors at the cognitive, personality, and social levels, gambling-related cognition are the most effective predictive factor for precisely identifying internet gambling population; 4) utilizing machine learning methods can effectively predict and identify the Chinese internet gambling population, with AdaBoost being the optimal algorithm. This research proposed and validated a multi-dimensional internet gambling predictive model centered around gambling-related cognition, enriching and enhancing the relevant foundational theories of internet gambling psychology. It can provide evidence-based support for governing internet gambling issues based on psychological theories, offer empirical suggestions for accurately identifying internet gambling populations in China, and propose potential operational strategies from a psychological perspective for addressing the prevalent issue of internet gambling in China. |
| 英文摘要 | 赌博（gambling）是一种在游戏或事件的不可预知结果上下注的行为，是人 类社会中普遍存在的活动。随着互联网和移动设备的兴起，网络赌博逐渐取代传 统实体赌场，成为了新兴赌博形式，对个体身心健康、个人和国家财产、社会风 气等均存在巨大危害。近年来，我国的网络赌博犯罪问题高发。针对网络赌博犯 罪团伙（庄家）开展的专项治理活动虽取得一定成效，但也面临着地域管辖限制 等难题。因此，从网络赌博参赌人员下手，切断网络赌博的资金流和信息流，是 当前我国监管治理网络赌博问题的现实需求和一个可行对策。实现这一监管方法 的前提之一，是可以在庞大的人群中精准识别潜在网络参赌人员。但现有的移动 支付平台实行的网络参赌人员识别方法依据较为单一，缺乏实证依据的验证，更 难以实现高效和精准兼备的管控目标。 心理学领域对于赌博的研究积累了丰富的理论成果。如认知-行为理论为代 表的综合理论认为，多个层面心理因素，如认知、人格和社会因素，均可以预测 个体的赌博意愿与行为。行为决策领域对赌博的研究也发现，赌博非理性认知， 即关于赌博的一系列非理性的认知扭曲，对于赌博预测可能具有关键的预测作用。 两派不同取向的研究均提示，赌博非理性认知可能是预测个体赌博行为的核心变 量。借鉴传统赌博研究领域已有结果，利用心理测量技术与机器学习方法，可为 准确测量赌博影响因素及预测网络赌博人群提供方法学的基础。 目前，学界尚未见开展对中国网络赌博人群的预测研究。现有研究存在多处 不足。其一，在理论层面，缺少对赌博预测因素的预测作用的比较和整合及各赌 博预测因素间关系的探索。其二，在方法层面，网络赌博行为的测量缺乏生态效 度，且赌博预测的方法准确性不足。其三，在适用性层面，针对中国网络赌博人 群的研究存在空白。 为克服上述不足，本研究认为，选取合适的网络赌博影响因素，在网络赌博 影响因素与赌博意图及行为之间建立关系，并使用机器学习方法对赌博人群进行 预测，是一条可行的研究路径。因此，本研究提出以下理论框架：1）多个层面 的影响因素（如认知、人格、社会等）均能正向预测赌博意图和行为；2）赌博 相关的认知因素正向预测网络赌博意图和行为，且赌博非理性认知是其中的关键 因素；3）赌博非理性认知作为核心变量，可以中介认知层面（风险寻求、延迟 折扣）、人格层面（冲动性、神经质）和社会层面（社会规范）各变量对网络赌 博意图和行为的预测作用。 为验证上述理论框架，本论文在中国成人样本中，基于网络调查问卷和网络 赌博实际行为数据，开展四个研究： 研究一以中国一般人群为样本，通过一个预研究（N=136）和一个正式研究 （N=900），发放网络调查问卷，初步考察了认知、人格、社会三个层面的六个预 测因素（赌博非理性认知、延迟折扣、风险寻求、冲动性、神经质、社会规范） 对网络赌博意图及参赌经验的预测作用并探索赌博非理性认知的中介机制。回归 分析结果发现，对于赌博意图，赌博非理性认知是最关键的预测变量。而对于参 赌经验，社会规范是最关键的预测因素。中介分析进一步发现，赌博非理性认知 中介了风险寻求、冲动性和社会规范对赌博行为意图和参赌经验的影响。研究一 结果表明，赌博非理性认知、风险寻求、冲动性和社会规范是有效的网络赌博预 测因素，且赌博非理性认知中介了部分赌博预测因素对网络赌博的预测作用。 研究二旨在验证网络赌博预测因素对网络赌博意图及行为的预测作用。研究 二筛选了研究一中的有效网络赌博预测因素，将因变量拓展到实际赌博行为和购 买彩票行为。研究二在一般人群中（N=866）通过线上问卷平台发放问卷，并通 过支付宝后台，匹配并获得他们的实际网络赌博行为数据（如年均赌博支出和次 数）。回归分析的结果发现，对于实际赌博行为，赌博预测因素的预测作用较弱 或不显著；而对于彩票购买行为，赌博非理性认知、风险寻求和社会规范均具有 显著正向预测作用。这表明研究一筛选出的赌博预测因素对网络赌博行为具有一 定预测力，但可能由于一般人群中网络赌博人群较少，研究二中赌博非理性认知 等因素对实际赌博行为的预测作用较弱。 研究三进行了方法层面的探索，在高网络赌博经验人群中使用机器学习方法， 探索中国网络赌博人群的有效预测方法，并验证研究一、二的结果。研究三以高 网络赌博经验人群（支付宝反赌先锋平台用户）为样本（N=1287），通过该平台 发放网络调查问卷，并获取平台提供的用户网络赌博实际行为数据。回归和中介 分析的结果表明，与研究一一致，在高网络赌博经验人群中，赌博预测因素可以 预测网络赌博意图及行为，且赌博非理性认知具有中介作用。随后，通过机器学 习中的聚类算法将网络赌博人群分为以下四类：低风险、中风险、高风险和可疑 风险人群。通过逻辑回归等五种机器学习算法构建预测模型，发现 AdaBoost 算 法模型在网络赌博人群预测中表现良好，最优模型的总体预测准确率达到 70%以 上，可以有效区分出上述四类网络赌博风险人群。其中，该算法对于中风险人群 的预测准确率最佳(81%)，且赌博非理性认知是各预测模型中最重要的预测变量 (shap 值为 0.17-0.34)。 由于研究一至三的样本人群不同，各预测因素的预测效应也存在差异，因此 研究四对研究一至研究三的结果进行了元分析，验证及比较各预测因素的预测作用。元分析的结果表明，赌博非理性认知、风险寻求和社会规范对网络赌博有显 著正向预测作用，且因变量类型（赌博意图/赌博行为）和样本类型（一般人群/ 特殊人群起到调节效应。对于赌博意图，赌博非理性认知、社会规范具有正向预 测作用；对于赌博行为，风险寻求、社会规范、性别（男性）、负债具有正向预 测作用；在一般人群中，赌博非理性认知、社会规范、性别（男性）、负债具有 正向预测作用；在特殊人群中，性别（男性）、负债具有正向预测作用。 总之，本研究发现：1）在中国社会背景下，赌博非理性认知、社会规范可 以预测网络赌博意图，风险寻求和社会规范可以预测赌博行为；2）赌博非理性 认知中介了其他因素对赌博意图和行为的预测作用；3）认知、人格和社会三个 层面的多个赌博影响因素中，赌博非理性认知是帮助精准鉴别网络赌博人群的最 有效预测因素；4）采用机器学习方法可以有效预测和识别中国网络赌博人群， 且 AdaBoost 是最优算法。 本研究提出并验证了以赌博非理性认知为核心的多层面网络赌博预测模型， 丰富和完善了网络赌博心理机制的相关基础理论，并可为网络赌博问题治理提供 基于心理学理论的基础证据，为我国进行网络赌博人群精准识别提出实证建议， 为治理我国网络赌博高发问题提供心理学视角的可能操作方案。 |
| 语种 | 中文 |
| 源URL | [http://ir.psych.ac.cn/handle/311026/48154]  |
| 专题 | 心理研究所_社会与工程心理学研究室 |
| 推荐引用方式 GB/T 7714 | 肖怡. 中国网络赌博人群的精准识别： 赌博非理性认知为核心的多层面因素预测模型[D]. 中国科学院心理研究所. 中国科学院大学. 2024. |
入库方式： OAI收割
来源：心理研究所
浏览0
下载0
收藏0
其他版本
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。