Algorithms and bias: What lenders need to know JD Supra

Much of the software now revolutionizing the financial services industry depends on algorithms that apply artificial intelligence (AI)--and increasingly, machine learning--to automate everything from simple, rote tasks to activities requiring sophisticated judgment. These algorithms and the analyses that undergird them have become progressively more sophisticated as the pool of potentially meaningful variables within the Big Data universe continues to proliferate. When properly implemented, algorithmic and AI systems increase processing speed, reduce mistakes due to human error and minimize labor costs, all while improving customer satisfaction rates. Creditscoring algorithms, for example, not only help financial institutions optimize default and prepayment rates, but also streamline the application process, allowing for leaner staffing and an enhanced customer experience. When effective, these algorithms enable lenders to tweak approval criteria quickly and continually, responding in real time to both market conditions and customer needs. Both lenders and borrowers stand to benefit. For decades, financial services companies have used different types of algorithms to trade securities, predict financial markets, identify prospective employees and assess potential customers. Although AIdriven algorithms seek to avoid the failures of rigid instructions-based models of the past--such as those linked to the 1987 "Black Monday" stock market crash or 2010's "Flash Crash"--these models continue to present potential financial, reputational and legal risks for financial services companies.