Responsible AI for responsible lenders

NestEgg将借款人与负责任的贷款人联系起来。NestEgg平台是人们检查自己是否有资格加入以及是否有资格从英国各地负责任的贷款人那里获得贷款的一个地方。一旦有人找到了合适的贷款，NestEgg支持他们的贷款人做出公平的信贷决定。通过使用人工智能将信用参考和公开银行信息结合起来，这一过程将得到改善。这使贷方能够在双倍快的时间内做出一致和更公平的决定。但我们可以做得更多。目前，信贷决策的数据源分布在太多不同的地方。交叉引用数据源的手动检查耗时太长，而且容易出现人为错误。此外，英国金融市场行为监管局（Financial Conduct Authority）发现，信用信息不足会导致贷款人做出不反映个人财务状况的决定。因此，消费者更有可能获得他们无法偿还的信贷，或者被拒绝支付他们负担得起的信贷。该项目使用人工智能根据这些不同数据集之间的相互作用为负责任的非营利性贷方做出决策。在5个月的时间里，20多家中小企业贷款机构将通过交叉参考数据集，利用人工智能自动化信贷决策，最初的重点是由于生活成本危机而进行的负担能力评估。该项目将评估：a)被人类推翻的贷款决策的比例；b)人工智能是否公平地自动化了负担能力评估；d)人工智能如何帮助贷款人满足监管要求；d)基于这些结果共享信息在多大程度上改善了被拒绝申请人的信用状况。这个最初的小规模实验使用了一小部分数据。人类根本不可能考虑数据可以用来支持自动决策的每一种方式。因此，我们也将考虑引入机器学习的方法，以便我们能够更好地应对市场和经济的变化，同时仔细平衡自动化的风险，以确保决策算法如何工作的持续透明度。我们的贷款客户服务于不同的人群。我们必须确保人工智能不会削弱他们支持弱势群体的能力。

NestEgg connects borrowers to responsible lenders. The NestEgg platform is one place for people to check whether they are eligible to join as well as qualify for loans from responsible lenders across the UK.Once someone has found a suitable loan, NestEgg supports their lender to make a fair credit decision. This process will improve by using AI to combine credit reference and open banking information. This enables lenders to make consistent and fairer decisions in double quick time.But more can be done. Currently the data sources for credit decisioning are in too many different places. A manual review to cross reference the data sources takes too long and is subject to human error.Furthermore, the Financial Conduct Authority found that insufficient credit information leads lenders to make decisions that do not reflect an individual's financial circumstances. As a result, consumers are more likely to have access to credit they can't pay back or denied credit they could afford.This project uses Artificial Intelligence to make decisions based on the interplay between these distinct data sets for responsible, non-profit lenders. Over a five month period 20+ SME lenders will work together using AI to automate credit decisions by cross referencing data sets, with an initial focus on affordability assessment because of the cost of living crisis.The project will evaluate:a) the proportion of lending decisions overridden by humansb) whether AI fairly automates affordability assessmentc) how AI can help lenders meet regulatory requirementsd) the extent to which sharing information based on these results improves the credit profiles of declined applicants.This initial small scale experiment uses a fraction of the data. It simply isn't possible for a human to consider every way in which data could be used to support automated decisioning. We'll therefore also consider ways to introduce machine learning so we're better able to respond to changes in the market and economy whilst carefully balancing the risks of automation to ensure continued transparency in how decisioning algorithms work. Our lender clients serve a diverse population. We must ensure that AI does not undermine their ability to support those from disadvantaged backgrounds.