Prospects and Risks of Algorithmic Lenders Using Big Data (IV)

The risks of Algorithmic Lender 2.0

Algorithmic Lender 2.0 can reduce the occurrence of human bias, while there is also the risk of human bias infiltrating the algorithmic decision process. In addition to this case, learning algorithms may still have discriminatory: First, the algorithm may acquire biased training data; Second, the algorithm may be biased in the initial programming; Third, the complexity and self-learning of the algorithm can cause to the bias that human don't understand.

There is no simple method to avoid the problem that Algorithmic Lender 2.0 may be discriminatory in its use of big data. Without training data, algorithms often cannot learn to make decisions. But in an American society where discrimination is still widespread, Algorithmic Lender 2.0 will perpetuate bias without aggressive intervention.

The important goal that client financial regulators should pursue is to prevent Algorithmic Lender 2.0 from being disadvantageous to a certain group of people and to mitigate the harm of algorithmic lending. The most effective solution may be that the CFPB could embed "regulators" in algorithms.