SBIR Phase I: Development of prediction-driven credit scoring and ruling platform for behavioral lending

这项小企业创新研究（SBIR）第一阶段项目的广泛影响将使目前服务不足的人口获得信贷。信誉在很大程度上取决于申请人的信用和财务历史的质量。然而，关于“信用隐形”（十分之一的美国成年人）和“薄档案”申请人（6200万美国人）的有限信息阻碍了对信用评分的准确评估。拟议中的基于人工智能（AI）的个性驱动借贷平台将为贷款人提供可靠的信用预测，而无需广泛的财务历史。此外，拟议中的系统将包括一个透明模块，以防止偏见，扩大信贷公平，并从代表性不足的群体中识别新的银行客户。该小企业创新研究（SBIR）一期项目旨在开发第一个人格驱动的信用评分和裁决平台。传统的评分模型基于财务历史数据进行评估，而该算法将根据申请人的个性产生可靠的信用评分。不需要金融交易历史（如煤气、水、电、电视、电话、宽带服务或租金支付记录）来得出关于申请人信誉的结论。这个第一阶段项目的研究工作将集中在1)评估替代数据点，如心理测试问卷、电信数据、人口统计、公司图和公开可用数据，2)开发半监督机器学习方法来预测个人信用可靠性，以及3)开发无监督聚类方法来识别与高信用可靠性相关的客户集群。此外，防止机器学习方法的基础训练数据中的偏差将是一个优先事项。该项目的主要目标是开发预测引擎作为拟建的个性驱动借贷平台的核心元素，并开发原型。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 of this Small Business Innovation Research (SBIR) Phase I project will enable credit to currently under-served populations. Creditworthiness is largely determined by the quality of the applicant’s credit and financial history. However, limited information regarding "credit-invisible" (one out of ten American adults) and "thin-file" applicants (62 million Americans) prevents accurate evaluation of a credit score. The proposed artificial intelligence (AI) based personality-driven lending platform will provide lenders a reliable creditworthiness prediction without an extensive financial history. Moreover, the proposed system will include a transparency module, preventing biases and expanding credit equity and identifying new banking customers from under-represented groups. This Small Business Innovation Research (SBIR) Phase I project aims to develop the first personality-driven credit scoring and ruling platform. While traditional scoring models base their assessment on financial history data, the proposed algorithm will produce a reliable credit score based on an applicant’s personality. No history of financial transactions (such as gas, water, electricity, TV, phone, broadband services, or rent payment records) is needed to reach a conclusion about an applicant’s creditworthiness. Research efforts of this Phase I project will focus on 1) the evaluation of alternative data points such as psychometric testing questionnaires, telecommunications data, demographics, firmographics, and publicly available data, 2) the development of a semi-supervised machine learning approach to predict credit reliability for individuals, and 3) the development of an unsupervised clustering approach to identify customer clusters associated with high credit reliability. Moreover, the prevention of bias in the underlying training data of machine learning approaches will be a priority. The primary objective of this project is to develop the predictive engine as the core element of the proposed personality-driven lending platform and develop a prototype.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.