Integrating Algorithmic Decision Making into Small Business Credit Initiatives: a path to Enhanced Efficiency and Inclusive Economic Growth

Abstract

Purpose: This paper addresses the challenges faced by small businesses in accessing credit through Small Business Credit Initiatives (SBCI) in the United States. Despite the success of SBCI in creating jobs and fostering economic growth, there are limitations in the evaluation process. Methodology: The research design integrates advanced algorithmic decision-making, machine learning, and LLMs into existing credit evaluation process. Primary data is collected from various sources, including financial and business history, market sentiments, external factors, and utilization of sampling techniques if required. Document review, surveys and digital platforms are used for collecting data for LLMs to extract insightful information from complex sources. This comprehensive approach, combining with traditional and innovative methods, aims to establish a robust foundation for developing and evaluating a fair, efficient, and adaptive credit evaluation system for small business credit initiatives. Findings: The proposed framework integrates external market factors and use of LLMs for document review on top of primary data sources currently in adaption. Data processing could be amended by extracting features by using advanced natural language processing to enhance feature space by collecting valuable information which is expected to enhance predictive power, adjustment of thresholds and decision making along with a feedback loop. Unique Contribution to Theory, Policy, and Practice: Unique framework to accelerate small business credit initiatives by developing a new process of selecting and evaluating machine learning model centered on addressing associated risks, adapting to changes in government policy, improving current procedures, and incorporating feedback from stakeholders and applicants. This is done in an organized manner, with a focus on monitoring and maintaining algorithmic decision models.