Bias can be built into algorithms
One might think that doing away with human bias in decisions would end discrimination. Instead it turns out that many algorithms have built in biases. Democratic presidential hopeful Senator Elizabeth Warren (D-MA) notes that the use of algorithms can potentially lead to lower levels of discrimination but algorithms can actually have biases hardwired into them and so should be checked for such biases before being used.
The case of African-American and Latino borrowers
Financial technology companies charge African-American and Latino homeowner an overall six to nine basis points higher interest rates than white and Asian borrowers. Researchers estimate that this adds an additional $250 million to $500 million a year in additional mortgage interest payments for US black and Latino borrowers.
Senator Warren’s letter
“
Warren, a Democrat from Massachusetts, is a member of the US Senate Committee on Banking, Housing and Urban Affairs. She sent a letter to Chairman Jerome Powell at the Board of Governors of the Federal Reserve System, Joseph Otting, Comptroller at the Office of the Comptroller of the Currency, Chair Jelena McWilliams at the Federal Deposit Insurance Corporation, and Kathy Kraninger, Director at the Consumer Financial Protection Bureau. The letter seeks information about what role banking agencies can play in making sure that lending algorithms serve US borrowers on a fair non-discriminatory fair basis.
The Warren letter, cosigned by Senator Doug Jones, an Alabama Democrat, reads: “Recent research highlights this tension, demonstrating both the opportunity of algorithmic underwriting’s potential to reduce discrimination, while also emphasizing the technologies’ current shortcomings… Researchers found that “algorithmic lending” appeared to reduce the likelihood that the loan applications of borrowers of color are rejected-thus reducing discrimination in lenders’ “accept/reject” decision-making process. However, the same study also found that “FinTech lenders [using algorithms] charge Latinx/African American borrowers 6-9 basis points higher interest rates” than comparable borrowers – which is almost identical to the interest rate premium that minority borrowers face with traditional underwriting. In other words, the algorithms used by FinTech lenders are as discriminatory as loan officers…. Senator Warren has requested responses to her letter by June 24, 2019.”
Warren is the latest Democratic presidential hopeful to demand answers about how algorithms negatively impact the lives of people of color. Last fall, a group of U.S. senators, including Warren and Senator Kamala Harris (D-CA) and Cory Booker (D-NJ), sent letters to federal agencies with questions about algorithmic bias and facial recognition software. t=_blank] recent article notes: “She is the latest Democratic presidential candidate to demand answers about how algorithms negatively impact the lives of people of color. Last fall, a group of U.S. senators, including Warren and Senator Kamala Harris (D-CA) and Cory Booker (D-NJ), sent letters to federal agencies with questions about algorithmic bias and facial recognition software.”
