Automated Valuation Models: A Beacon for Responsible AI in Housing
Automated Valuation Models (AVMs) are increasingly becoming a cornerstone in the housing sector, offering quick and cost-effective property valuations. However, technology is not without its pitfalls, especially when it comes to perpetuating long-standing discriminatory practices like redlining. This blog post delves into the multi-faceted roles AVMs can play in shaping a more just housing and lending landscapes, guided by frameworks such as the OSTP’s AI Bill of Rights, NIST’s AI Risk Management Framework, and NFHA’s PPM framework for algorithmic auditing. It is inspired by a recent publication authored by Alex Engler and Sylvia Brown.
The Ghost of Discriminatory Policies
AVMs are only as good as the data they are trained on and the subjective decisions behind their architectural designs. Historical housing policies like redlining, which were discriminatory by design, can haunt even the most well-intended machine learning models. AVMs can be used as a case study for how to design safeguards that ensure AI solutions do not ride on the roads of discriminatory legacies to constitute systemic and structural barriers to well-resourced communities and equitable lending.
Regulatory Tools for Responsible AI
Housing and lending regulators have a plethora of tools at their disposal to ensure that the benefits of AI in housing far outweigh its challenges. The OSTP’s AI Bill of Rights, NIST’s AI Risk Management Framework, and NFHA’s PPM framework provide comprehensive guidelines for algorithmic auditing. They can be used to build, study and test AVMs for architectural blueprints and foundational blocks that can be used to construct effective and enforceable regulations that make AI a force for good.
The Role of Research Institutions and Nonprofits
Research institutions and nonprofits have a unique role to play in shaping the ethical landscape of AI in housing and lending. They can perform surgical operations on AVMs to make a compelling case for policy solutions that guarantee consumer protection from algorithmic harms. Their independent status allows them to offer unbiased insights that can be invaluable in shaping AI regulations.
Declaring War on Algorithmic Apartheids
The emergence of algorithmic apartheids in housing and lending is a pressing concern that requires collective action. Exploratory surgery on AVMs can provide a roadmap needed to combat algorithmic injustices and a blueprint for reconstructive surgery to heal society from negative legacies rife in housing and lending data. It’s not just about creating ethical algorithms; it’s about waging a war against systemic biases and structural barriers that have found a new breeding ground in AI and AVMs can be testbeds for learning how to design effective ammunitions for the war against algorithmic injustices.
Governance Solutions that Use AI as a Wheel for Distributive Justice
A recent Brookings report on AVMs offers insights into governance solutions that can ensure the benefits of AI reach those who need them most. From transparent algorithms to community-driven data collection, these solutions offer a roadmap for making AVMs a tool for social justice and economic empowerment and the roadmap can be adapted for a national strategy that delivers AI dividends to society and stops perpetuation of and creation of AI biases.
Conclusion
Automated Valuation Models are not just tools for property assessment; they are mirrors reflecting the complexities of our housing and lending ecosystems. AVMs can serve as a guide that beckons us to address historical injustices, wield the right governance frameworks, and declare our commitment to a future where algorithmic technologies uplift society rather than divide it. AVMs can provide insights for designing a moral compass that guides us through uncharted territories as we journey towards algorithmic justice.
