Algorithmic decision-making is the use of data and computational models, from simple scorecards to machine learning (ML), to make or support decisions that were once taken by people. In finance it is used to approve or decline loans, set credit limits and prices, flag suspicious transactions, verify identity and target marketing.
In digital lending and financial inclusion, algorithms built on big data such as mobile phone records, payment histories and app usage allow providers to assess people with no credit report, at low cost and in minutes. The same features raise concerns. Models can reproduce or amplify historical bias against women, minorities or poor areas; complex models can be hard to explain to a declined applicant; and the data used may be collected without meaningful consent. Regulators and standard-setters therefore stress fairness testing, explainability, human review of significant decisions, model governance and data protection. A credit scoring model is the most common example in lending.
Ref 1300000380