(Credit: Zivica Kerkez / Shutterstock.com )
AI systems in lending discriminate against women. Study shows: Women pay significantly more for car loans. But why does AI fall back on old prejudices?
Many banks use artificial intelligence (AI) in lending. Some use AI only to reduce the processing time of applications by capturing documents faster. Others also let it decide whether a loan is approved or rejected.
Advertisement
Use of AI puts women at a disadvantage when it comes to car loans
The idea behind this is that AI algorithms can quickly analyze large amounts of data and provide more accurate assessments of creditworthiness. But they can also lead to systematic discrimination against women, as researchers at the University of Bath in the UK have found.
The scientists examined credit decisions in Canadian car dealerships. It was found that the use of AI significantly increased the profits of lenders. It also did not increase prejudices against customer groups that are traditionally excluded in society. Women were a “glaring exception”.
Christopher Amaral from the School of Management at the University of Bath explained:
We found that lenders were able to increase their profits by applying machine learning techniques to thousands of auto loans. But this gain also comes with a significant downside: AI can cause auto loans to be disproportionately less favorable for women, exacerbating social injustice in the system.
Historical discrimination against women continues in AI systems
The study found that lenders can increase their annual profits by up to eight percent by using AI to optimize sales commissions. However, this comes at the expense of customers, especially women. The expected utility of loans fell by 20 percent overall, and by as much as 42 percent for women.
Women have historically been treated differently than men. Amaral points out that the bias against women in this sector is well documented. Salespeople will negotiate worse credit terms for women than for men.
Read also
Show moreShow less
Little research has been done into why this is the case. However, according to Amaral, one study suggests that salespeople assume that women are less able to correctly assess credit terms. This seems to be due to the fact that they know less about certain types of credit than men, but also because women are seen as less assertive.
AI and lending: solutions for fair decisions
AI systems trained with such data reproduce these credit decisions. But they can also provoke measures by regulators to prevent discrimination against women. Against this background, it makes sense for companies to forego AI altogether.
Unless the algorithms are reprogrammed as the researchers have done. This means they are no longer disadvantaged, but the profit can only be increased by four percent.
“Instead of restricting the use of AI, we should encourage companies to use it responsibly and ensure that it serves both their business goals and social justice,” Amaral concluded.
