Credit scoring is an increasingly central and contested domain of data and AI governance, frequently framed as a neutral and objective method of assessing risk across diverse economic and political contexts. Based on a nine-month ethnography of credit scoring practices in Nairobi, Kenya, we examined the sociotechnical and institutional work of data science in digital lending. While established regional telcos and banks are leveraging proprietary data to develop digital loan products, algorithmic credit scoring is being transformed by new actors, techniques, and shifting regulations. Our findings show how practitioners construct alternative data using technical and legal workarounds, formulate risk through multiple interpretations, and negotiate model performance via technical and political means. We argue that algorithmic credit scoring is accomplished through the ongoing work of alignment that stabilizes risk under conditions of persistent uncertainty, taking epistemic, modeling, and contextual forms. Extending work on alignment in HCI, we show how it operates as a two-way translation, where models are made "safe for worlds" while those worlds are reshaped to be "safe for models."

翻译：信用评分日益成为数据与人工智能治理中备受关注且争议不断的领域，常被视作一种跨不同经济和政治背景评估风险的中立客观方法。基于在肯利亚内罗毕长达九个月的信用评分实践民族志研究，我们考察了数字借贷中数据科学的社会技术与制度运作。尽管成熟的区域电信运营商和银行正利用专有数据开发数字贷款产品，但算法信用评分正被新兴参与者、技术手段及不断变化的法规所重塑。研究发现揭示了从业者如何通过技术与法律变通构建替代数据，通过多重解读来构建风险概念，并借助技术与政治手段协商模型性能。我们认为，算法信用评分是通过持续的对齐工作实现的——这种工作在持续不确定条件下以认知、建模和情境化三种形式稳定风险。通过拓展人机交互领域的对齐研究，我们揭示了这种对齐如何作为双向翻译运作：既使模型"适配世界"，又重塑世界以"适配模型"。