The concept of “standard of care” is well known in professions such as medicine and law. It represents the benchmark for accepted practices in the job based on how “a reasonable professional in that position or industry would have [acted] under the circumstances based on then-existing knowledge.” By definition, the standard of care is intended to evolve based on constantly advancing levels of knowledge and capabilities.
This same concept can help guide how we in financial services think about compliance with core regulatory requirements, such as in fair lending. More specifically, what is the standard of care when it comes to avoiding disparate impact in lending to protected classes, and how should this standard evolve based on our knowledge and the tools we have at our disposal to mitigate such risks?
Let’s all first agree that the underwriting and extension of credit to all people, regardless of race, gender, or ethnicity, through an equitable and non-discriminatory process is the ultimate objective of a fair financial system. The challenge is to find ways to do it better and to ensure that our regulations don’t unintentionally lock in the status quo or unnecessarily delay progress.
This is the same challenge we face in other fields – especially those with technical underpinnings. When it comes to medical science, for example, without testing, experimentation, and measurement of outcomes it would be next to impossible to advance the state of the field and thereby evolve the standard of care. The unprecedentedly swift development and approval of several COVID-19 vaccines showed what can be done when regulatory systems are agile enough to facilitate critical innovations to benefit human lives.
When viewed from the other side of the same coin, the costs of not permitting innovation is stark: for example, imagine a scenario where COVID-19 vaccine development was blocked due to unnecessary regulatory barriers. Social costs would similarly be incurred if we do nothing to make lending more inclusive and fair, especially if it can be done without introducing risk to the financial system. For example, it is estimated that banks could generate up to $380 billion in annual revenue by closing the small business credit gap and bringing unbanked and underbanked adults into the formal financial system. And wider access to credit could boost global GDP by $3.7 trillion, and engender $4.2 trillion in new deposits and $2.1 trillion in additional loans, according to a report from McKinsey.
With this framing, it is worth considering how we can take steps to advance the standard of care in fair lending compliance.
To start, we have to recognize that systems today in the consumer lending space are highly technical: they are software and machine-driven and capable of processing and analyzing rich data sets. For this reason, it is critical for all involved stakeholders, including providers, innovators, and regulators, to have the requisite technological literacy to understand and assess new tools or approaches – both benefits and new risks.
Lending via artificial intelligence (AI) and machine learning (ML) now provide us the tools to advance the state of fair lending. It is time to improve what financial service providers view as reasonable.
Next, regulators should leverage modern regulatory tools, including sandboxes, labs, safe harbors, and no-action relief, in order to permit robust testing of new models that can help advance the state of the field. Zest works with consumer lenders of all sizes to deploy more accurate and fairer models using lots more data and better ML math. It’s new to the industry, and we understand and agree that the “proof is in the pudding” in terms of demonstrating real-world results. There’s a big need to collect and share more data industry-wide about the impact of AI and machine learning on financial inclusion. Regulators should embrace the appetite of innovators to test and demonstrate results by creating clear and formal paths to pursuing such ends.
Finally, regulators and industry must keep their eye on the prize – in this case a more fair and equitable financial system – in order to speed the pace of progress. Blind adherence to legacy approaches and processes should not be allowed to create a maze that unnecessarily prevents the field from advancing, and a sense of urgency is perhaps the best antidote. As we have learned over the past year with the development of life-saving vaccines in record time, we are capable of safely innovating and helping lives in short order.
With respect to advancing fairness in financial services, we should pull from the above learnings and pursue the same objective of helping lives. If tested and verified, advanced math-based machine learning tools applied to consumer lending and fairness hold substantial promise in advancing the state of the field and with it the standard of care.
Flo is general counsel and chief administrative officer of Zest AI.