I-Corps: Knowledge Graph Embeddings-based Explainable Artificial Intelligence for Enterprise Performance Management

I-Corps项目更广泛的影响/商业潜力是为投资者、客户、供应商、员工和社区开发一个企业绩效管理（EPM）平台。该技术旨在将知识范围从以财务为中心的绩效扩大到涉及所有利益相关者的经济、社会、心理和身体健康的跨学科框架。此外，这项技术可能会使人工智能（AI）大众化，让那些可能不具备复杂分析技能的普通组织经理也能使用。目前的AI模型缺乏交互性和直觉性的故事叙述。将数据的层次聚类与因果知识图相匹配，所提出的技术将以模仿总经理直觉思维的方式准备用户数据。这项技术解决了市场上的一个商业空白——缺乏规定能力，即告诉最终用户他们应该做什么。数据可能从组织中的不同来源收集，因此它们是碎片化的，并且丢失了因果关系。因果知识图的外部来源通过呈现和解释EPM数据中隐藏的因果关系来填补空白。该项目旨在帮助管理人员制定干预措施，以提高所有利益相关者的福祉。这个I-Corps项目的基础是开发一个基于知识图嵌入的平台，用于企业绩效管理（EPM）数据的统计和机器学习模型。该技术旨在利用自然语言处理模型，将组织科学中的大量科学研究转化为因果知识图，该知识图将嵌入到可视化分析平台中，以构建和解释企业管理数据。目标是通过可视化和直观地解释隐藏的因果路径来帮助EPM用户，从而改进组织管理。提出的技术结合了组织科学和计算机科学的研究成果，涉及两项创新：关于组织绩效因果关系的科学知识图和基于知识图嵌入的新可视化技术，以实现可解释的人工智能（XAI）。层次聚类是对数据中变量的描述进行解释和组织的方法。因果假设是基于知识图中已知的因果关系自动开发的，然后在统计和机器学习模型中进行实证检验。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。

The broader impact/commercial potential of this I-Corps project is the development of an enterprise performance management (EPM) platform for investors, customers, suppliers, employees, and the community. The technology aims to broaden the scope of knowledge from financial-centric performance to an interdisciplinary framework of economic, social, psychological, and physical well-being concerning all stakeholders. In addition, the technology may democratize artificial intelligence (AI) to ordinary organizational managers who may not possess sophisticated analytics skills. The current AI models lack interactive and intuitive storytelling. Matching the hierarchical clustering of data with a causal knowledge graph, the proposed technology will prepare user data in a way that mimics a general manager’s intuitive thinking. The technology addresses a commercial gap in the market - that of a lack of prescriptive capability, that is, telling end-users what they should do. Data may be collected from different sources in an organization, so they are fragmented and the causal links are lost. The external source of a causal knowledge graph fills the gap by presenting and interpreting the hidden causal links in EPM data. The project seeks to help executives to prescribe interventions to enhance the well-being of all stakeholders.This I-Corps project is based on the development of a knowledge graph embeddings-based platform for statistical and machine learning models of enterprise performance management (EPM) data. The technology is designed to engage natural language processing models to convert a massive volume of scientific research in organizational science into a causal knowledge graph, which will be embedded into a visual analytics platform to structure and interpret enterprise management data. The goal is to help EPM users by explaining the hidden causal pathways visually and intuitively to enable improvements in organizational management. The proposed technology combines research outcomes across organizational and computer sciences and involves two innovations: a scientific knowledge graph on causes-and-effects related to organizational performance and a new knowledge graph embeddings-based visualization technique to enable explainable AI (XAI). Hierarchical clustering is used to explicate the descriptions of variables in data and organize these descriptions. Causal hypotheses are automatically developed based on the known causal links in the knowledge graph and then empirically tested in statistical and machine learning models.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.