Doctoral Dissertation Research in DRMS: Building a comprehensive understanding of enterprise risks and their interdependencies for improved risk-intelligence

企业无处不在，与人类生活密不可分。蓬勃发展的公司为一个公平的社会提供了必要的基本活动，其好处从现成的商品和服务到支持经济繁荣的区域就业。同时，COVID - 19等高风险事件令人沮丧地提醒人们，企业持续受到威胁，建立企业抵御能力至关重要。然而，企业弹性的基石之一——风险情报——或对各种风险因素及其复杂的相互依存关系的充分认识，仍然严重不发达。该项目通过开发企业风险的复杂网络视图来解决这一知识缺口。利用大数据和人工智能工具，该项目确定并开发了一个与企业无关的风险清单，该清单比任何此类公共可用资源都要全面得多。此外，专家的意见允许将看似不相关的风险因素转化为相互关联的复杂风险网络，从而能够寻找可能复合并导致更严重不利影响的风险链。这项工作建立了一个知识资源库，有助于解释级联风险的机制，并预测风险事件对企业的不同影响。因此，这项工作符合国家利益，并与NSF促进科学进步的使命保持一致，从而促进国家的繁荣和福利。这项工作以企业风险管理的复杂系统观点为基础，旨在建立一个全面的、基于数据的动态风险网络的观点，以系统地提高企业的风险意识，并为真正的风险智能组织的发展做出贡献。这一观点是通过以下方式实现的：a)开发全面的企业不可知论风险因素清单，b)生成映射风险因素相互关系的风险网络，以及c)探索这些风险网络的复杂动态。这项工作需要一种混合方法，利用大型公司风险管理数据集上的信息提取（IE）和用于复杂风险网络开发和分析的模糊认知映射（FCM）。公共信息（美国证券交易委员会文件）与标准普尔500指数企业的私人风险数据（分析师报告）相结合，提供了对真实风险因素的强大覆盖。语料库使用IE原理进行分析，其中包括词性标注、依赖解析、n - gram提取和主题建模。来自行业和学术界的被调查和/或访谈的专家告知在FCM过程中导致复杂风险网络发展的风险相互作用（依赖关系、方向和影响程度）的定性度量。由此产生的风险网络进行定量分析，以揭示风险的中心性、风险之间的距离以及风险网络中的子群体结构等见解，这些见解可以为关键风险的顺序提供信息。该研究通过提高学术界对影响企业的风险广度的认识，为企业风险管理领域做出了贡献。此外，通过FCM，这项工作将专家的理解转化为可量化的网络，将重点放在风险的相互依赖性上。此外，通过网络分析，这项工作阐明了在传统观点中可能被忽视的关键风险和传播机制。总的来说，这项工作为影响企业的风险因素、它们的（非直觉的）相互作用和相关动态提供了一个广泛的、基于数据的观点，从而推进了ERM研究的复杂性观点，并提高了企业预测由看似不相关的事件引起的级联效应的能力。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

Businesses are ubiquitous and inseparably merged with human lives. Thriving corporations provide the fundamental activity necessary for an equitable society with benefits ranging from the ready availability of goods and services to regional employment in support of economic prosperity. Simultaneously, high-risk events such as COVID‐19 are bleak reminders that enterprises are continuously threatened, and that building enterprise resilience is crucial. Yet, one of the cornerstones of enterprise resilience - risk intelligence – or the adequate awareness of the various risk factors and their complex interdependence, remains grossly underdeveloped. This project addresses this gap in knowledge by developing complex network views of enterprise risks. Employing big data and tools of artificial intelligence the project identifies and develops an enterprise-agnostic risk inventory that is considerably more comprehensive than any such publicly available resource. In addition, expert input allows for converting seemingly disconnected risk factors into interconnected complex risk networks, which enable the search for risk chains that may compound and lead to more significant adverse effects. This work builds a knowledge resource base useful to explain mechanisms of cascading risks and to predict the varying impacts of risk events on enterprises. Thus, the work serves national interest and is in alignment with NSF’s mission to promote the progress of science, and via that, advance national prosperity and welfare. This work is grounded in the complexity systems view of enterprise risk management and seeks to build a comprehensive, data‐informed view of the dynamic risk network influencing enterprises to systematically enhance risk awareness and contribute toward the evolution of truly risk‐intelligent organizations. This perspective is achieved through: a) the development of a comprehensive enterprise agnostic risk factor inventory, b) the generation of risk networks that map risk factor interrelations, and c) the exploration of the complex dynamics of these risk networks. The work entails a mixed-methods approach utilizing information extraction (IE) on a large, curated dataset of company risks, and Fuzzy Cognitive Mapping (FCM) for complex risk-network development and analyses. Public information (SEC filings) are augmented with private risk data (analyst reports) for enterprises in the S&P 500, providing robust coverage of true risk factors. The corpus is analyzed using IE principles, which include part‐of‐speech tagging, dependency-parsing, n‐gram extraction, and topic modeling. Surveyed and/or interviewed experts from industry and academia inform qualitative measures of risk interaction (dependencies, direction, and degree of influence) during the FCM process leading to complex risk network development. The resulting risk networks are analyzed quantitatively to reveal insights such as centrality of risks, the distances between risks, and sub‐group structures within the risk networks that could inform an order of critical risks. The research contributes to the field of Enterprise Risk Management (ERM) by increasing scholarly awareness on the breadth of risks affecting enterprises. Further, via FCM, the work converts expert understanding into a quantifiable network, bringing focus on risk interdependencies. In addition, via network analysis, the effort illuminates critical risks and propagation mechanisms that may be overlooked in traditional views. Overall, this effort provides an expansive, data‐informed view of risk factors affecting enterprises, their (non-intuitive) interactions, and related dynamics thereby advancing the complexity-view of ERM research as well as sharpening an enterprise’s ability to predict cascading effects caused by seemingly unrelated events.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.