Is the Era of Fairer AI Finally Here?

The future of innovation is global. We discuss it here.

Image of an AI robot holding the scale of justice

Co-authored with my partner in crime at The Pnyx podcast Vikas Raj, and originally published on our Forbes column here.

The intersection of artificial intelligence (AI) and financial services is a hot topic. The potential for AI to revolutionize the financial services industry is immense, and it is being reflected in an endless parade of fundraise announcements, but all this AI buzz also raises important questions about the fairness, ethics, and the regulation of using AI in the provision of core financial services.

In a recent episode of "The Pnyx," we interviewed Kareem Saleh, CEO and Founder of FairPlay AI, where he delved into these issues, sharing insights on how his company is leveraging AI to create fairer lending practices. Kareem’s unique perspective, shaped by his background in both law and finance, offers a compelling look at the challenges and opportunities in this rapidly evolving field.

Here are the three main takeaways from the conversation:

1. Innovative Uses of AI Can Drive Fairer Lending

Kareem makes the case for leveraging AI for fairer lending. He shares his “ eureka moment” when he discovered that by applying adversarial debiasing techniques to a major mortgage originator's credit model, approval rates increased for Black consumers by about 10% without additional risk. By reducing reliance on conventional credit scores and optimizing the influence of other predictive variables, the mortgage originator was able to grant many billions of dollars in additional credit. This breakthrough not only facilitated 50,000 more Black families in securing homes but also demonstrated that fairer lending practices can align with financial institutions' profitability goals. Others have certainly taken notice. Last month, AI-startup Stratyfy partnered with cash flow and underwriting platform Prism Data “to help lenders make more informed decisions in cases where traditional credit data is insufficient.” Clearly, there are real opportunities for AI to democratize credit and expand lenders’ books.

2. We Must Address Disparities in Credit Underwriting

Disparities in decisioning systems often stem from “limitations in data and mathematics” rather than “people of bad faith building those models,” according to Kareem. To combat this, companies need to fine-tune the weights of various predictive variables to enhance fairness without compromising accuracy. For example, while traditional credit scores can dominate underwriting decisions, they may inadvertently perpetuate historical biases. By rebalancing the significance of these scores and integrating alternative data points, more equitable outcomes can be created for underserved communities, ensuring that deserving individuals aren't unfairly denied credit opportunities. As an industry, the solution is to be proactive; we can’t let the limited data and mathematics have their way. Investors, in particular, need to diligence AI models, as Accion details more here, prior to pouring capital fuel on the fire.

3. Balancing Innovation and Regulation is Vital

Regulation in AI is a double-edged sword and represents “one of the toughest questions” according to Kareem; over-regulation can stifle innovation and entrench the dominance of big tech companies. To ensure successful progress in AI, society must adopt smart, adaptive regulations that prevent misuse without hindering competition. Drawing parallels with the nuclear power industry's stringent regulations, Kareem argues for a balanced approach that fosters innovation while implementing safeguards to prevent ethical breaches and harmful practices. The CFPB is already taking steps in regulating AI, such as with home appraisals earlier this summer. Whether making use of other industries' frameworks or drafting something entirely new, it is paramount that regulators balance the risks alongside the benefits of AI technologies.

Conclusion

AI boasts substantial transformative potential in financial services and beyond, while also demanding careful consideration of ethical and regulatory challenges. FairPlay AI's innovative approach and vision offer an example of how we might ensure that AI-driven decisions are fair and equitable, benefiting both individuals and society as a whole. If companies implement AI correctly and ethically, it may just lead to the “strong competitive advantage” that Ganesh Rengaswamy (Managing Partner at Quona Capital) shared in our previous reflections on 2023.

The responsibility of ensuring AI is a beneficial force in society does not fall on fintech startups alone, rather all of us. Where do you see AI making significant improvements in financial services? What risks do you worry about? Let us know in the comments.