At the crossroads of political and racial discord, set against a backdrop of unprecedented data proliferation and technological advancement - there is certainly no shortage in the calls for self-reflection as based upon our current state of affairs.
Of the many difficult questions that lay ahead - one specific question continues to crop up in my news feed relating to all things AI: How can we prevent what is arguably the greatest side-effect of mass AI adoption - the prevalence of bias?
Undoubtedly, there needs to be some level of reckoning by both policymakers and the financial services industry alike. For policymakers – there needs to be clear guidance on "how much to prioritize accuracy at a cost of bias against protected classes”. For the financial services industry, or the architects of their AI – there needs to be a decision made on the value of faster and cheaper predictability when measured against the cost of non-compliance and negative social perceptions.
The linked article (part of a broader series of writings related to "AI Governance") puts forth a framework for evaluating the priorities and trade-offs that will need to be taken into account as we push forth on relinquishing human centered decision-making in favor of these algorithmic proxies.
The article highlights (as exemplified by the consumer credit landscape) some potential tradeoffs we might be willing to make between accuracy and bias in the application of AI for the financial services industry. Here in the U.S., the foundations for our regulatory framework to protect against discrimination in consumer lending (i.e., the Equal Credit Opportunity Act of 1974, the Truth in Lending Act of 1968 and the Fair Housing Act of 1968, etc.) were created at a time when high-quality data, standardized sources of information and the pool of available credit were all comparatively lacking. Fast forward to the present, and we see the opposite is true - where the quality and availability of standardized data is in abundance and the appetite for issuing new credit is seemingly infinite.
A recommended read for financial services and regulatory compliance professionals who have been professionally and personally grappling with similar considerations.
A broader conversation regarding how much bias we are willing to tolerate for the sake of improvement over the status quo would benefit all parties. That requires the creation of more political space for sides to engage in a difficult and honest conversation. The current political moment in time is ill-suited for that conversation, but I suspect that AI advancements will not be willing to wait until America is more ready to confront these problems.