Automated Redlining in 2020 - Is Tech Systemically Biased?

The Covid-19 health and economic crisis has expanded and accelerated our adoption of technology at a rate never seen before. We are finding ourselves in a place that might otherwise have taken years to get to and are just starting to become aware of the implications.  

The crisis is also showing us that while technology can move us forward in leaps and bounds, new automated systems  are dragging along the systemic biases of the past.

Technology In Hot Water

Remember the proverbial frog in boiling water? If put suddenly into a hot pot the frog will jump out, but if put in tepid water that is slowly brought to a boil, the frog will be cooked to death before becoming aware of the danger. 

Technology has changed our lives dramatically but incrementally over the past several decades. The heat has been turned up slowly making it difficult to see many of the looming dangers that we may confront as we rely more and more on technology. Suddenly the Covid-19 crisis has shoved us forward in time and tipped us over the edge. Whether it’s finance, health, social interactions or education, tech-driven applications have become the default rather than an alternative to the old ways of doing things. 

Technology’s Unfulfilled Promise 

We are now in the middle of a massive health and economic crisis requiring us to drastically alter our work and personal lives by utilizing technology-driven applications that we had little incentive to use before. Old habits and ways of doing things are dying fast. It has taken a crisis of epic proportions, but we are seeing that much of the convening, convention-ing and meeting we spent so much time and money on pre-Covid-19, may be unnecessary.

Technology has always held out the promise of lowering costs and increasing access and opportunity for marginalized communities and families in our economy, but instead, today’s technology is being used to reincarnate and amplify barriers and systemic biases from the past. The Covid-19 crisis has revealed that technology’s great promise to enable a more inclusive economy will not be kept without a broad effort to change direction.

Redlining 

Redlining was the decades long practice of excluding low-income people of color from obtaining federally backed mortgages by drawing red outlines on paper maps showing where lending was prohibited. This practice essentially legitimized housing discrimination and segregation and made it difficult and expensive for BIPOC (Black, Indigenous, People of Color) families to realize the same benefits from home ownership as white homeowners. 

The practice of redlining was finally outlawed by the Fair Housing Act of 1968. No longer could the federal government legally administer loan programs in a way that was discriminatory in intent or impact. Now, more than fifty years later, we find that redlining has been re-enabled and enhanced with technology. 

Others Need Not Apply

The CARES Act, which was signed into law at the end of March, created the Payroll Protection Program (PPP). The PPP is the largest and most advantageous single loan program ever offered to small businesses in the United States by the federal government. The $350 billion in loans made by the SBA under the program were not only offered at a reduced interest rate, they are also forgivable if the loan proceeds are used to pay employees and normal business expenses - this provision effectively turns the PPP loans into free grants for the businesses that received them.

Congress included language in the CARES Act requiring the SBA to prioritize female-owned, black-owned and rural businesses in the loan program but what happened was just the opposite. 

While the Trump Administration has refused to provide exact data, it appears that less than 2% of the loans went to the businesses intended to be prioritized and applications from white owned businesses were approved at a rate upwards of ten times greater than applications from BIPOC enterprises. It should be noted that the selection process did not include credit evaluation because the loans were all guaranteed by the federal government and repayment was not required if proceeds were applied to payroll and other expenses.

A report released on May 8th by the SBA Inspector General found that the SBA did not comply with Congress’ mandate.  As a result, the Inspector General wrote, “These borrowers, including rural, minority and women owned businesses, may not have received the loans as intended.” 

14 Years into 14 Days

PPP was launched on April 2nd.  By April 16th, close to $350 billion had been lent out by the SBA to one and a half million businesses around the country. In a contactless business environment  the SBA loaned the same amount of money out in 14 days as it had in the prior 14 years. 

To emphasize, this was all done in a contactless business environment: no in-person meetings, no lunches and no paper document signings. This simply would not have been possible without the use of a broad array of technology applications that PPP participants had to adopt because of the Covid-19 crisis and the “network effect” of everyone being forced to adopt the same or compatible technology simultaneously.  

The New Redlining 

While the tech-enabled efficiency of the PPP program is extraordinary, it was also a colossal failure in that it ended up discriminating against the very businesses that the program was supposed to prioritize and support. Why did this happen?

Let’s assume that most of the government officials, bankers and other business people involved in the PPP did not overtly intend to discriminate against businesses in BIPOC communities. Instead, the discriminatory outcome seems to have been hard-wired into a tech-driven system that amplified and accelerated bias in ways not seen in the past.

Rapid and easy access to loans under the program was largely made possible through pre-existing relationships with industry associations, law firms, accounting firms and banks. These organizations deployed tech guidance and processes to coordinate and facilitate applications on behalf of their best members, clients and customers. The fact is that white-owned businesses are far more likely than others to have these kinds of relationships due to historic discrimination and systemic bias in general.

Because female and BIPOC owned businesses did not have access to the business networks and relationships with banks and professional firms, they also did not have access to the tech-driven process and expert guidance that was required to compete for a PPP loan. While no one may have intended it, BIPOC businesses were effectively red-lined from access to the largest small business loan and grant program in the history of the country.

If there are any winners in the Covid-19 crisis it may be technology. The forced adoption of new tech-driven processes has accelerated what we would have seen anyway a few years down the road. There will be no return to “normal” after the crisis has passed. 

Automated Systemic Bias

Past injustices can build-in bias, discrimination and disparities in today’s automated systems. This first became evident with Machine Learning and other AI tools in which algorithms can demonstrate unwarranted discrimination and bias - even if not consciously intended by programmers. 

The problem is that if algorithms utilize data that reflects inputs and outcomes generated during times of systemic bias, racial discrimination and unequal opportunity, we are effectively hard-wiring those discriminatory outcomes into the AI-driven world of our future.

The Covid-19 crisis is showing us that the problem extends far beyond biased AI algorithms. As computers and software replace humans and become more essential to our interactions at work and home and increasingly automate decision making, those with limited access and understanding of technology will be left behind. 

The Problem Goes Far Beyond Government Programs

The problem extends far beyond government loans. Automated payment systems, the process of buying a home and getting a mortgage, learning about and applying for employment and shopping for goods are only some of the systems that will soon be almost entirely automated and online. Equal ability to access and benefit from automated systems is necessary for people to have a chance at financial well-being. 

While not a focus here, we must also consider the rapid growth of online health care which has grown at triple digit rates during the Covid-19 crisis.  Again, those lacking access to the technology to participate in remote care, will soon be at a distinct disadvantage to those that can.

What Can Be Done?

1. Unequal access to the technology needed to remain financially and physically  viable in life needs to be identified, publicized and studied.  Otherwise, the injustices and biases of the past are going to be brought forward and magnified in the future.

2. We must learn from the PPP debacle. Automated private sector and government processes need to be examined to make sure that BIPOC communities are not further marginalized by automation. There must be equal access to programs and resources when automated processes favor those able to utilize technology to participate. Accommodations must be made if needed for equal access

3. Government and nonprofit organizations that provide services to those needing help must somehow come up to speed with the private sector with regard to utilizing technology - they are light years behind. They must also help their clients and users get access to the technology they need so they have an equal opportunity to participate in our economic and health care systems in the future.