Breaking new ground in generative AI: Zillow’s Fair Housing Classifier

Zillow’s latest technology promotes fair housing practices within AI-powered search and chatbots

Posted in

Even though generative AI has become a trending topic lately, you’ve most likely become familiar with Zillow’s original AI model, the Zestimate, since its release in 2006. For almost two decades, the Zestimate has helped millions of people understand the potential market value of real estate properties. 

Since launching the Zestimate, Zillow has continued to incorporate AI capabilities into its products and services to help homeowners and home buyers make the most informed decisions throughout their real estate journeys. Zillow’s latest AI innovation, the Fair Housing Classifier, is designed to establish guardrails in real estate conversations that are powered by large language model (LLM) technology. 

What’s an LLM?

In the simplest of terms, an LLM is a form of generative AI that can be thought of as a storage space for vast amounts of conversational language data. Many chatbots are powered by LLMs, and when you ask a question, the chatbot goes to its LLM to look through its stored language data and provide you with the correct answer. A chatbot is only as good as the data contained in its LLM.

Out of the box, the LLMs behind many AI solutions do not consider fair housing requirements in conversations with users, and when deployed, can perpetuate bias. To address this, Zillow has built the Fair Housing Classifier, which prioritizes fairness as we advance our commitment to innovation with generative AI technology.

The Fair Housing Classifier focuses on mitigating the risk of illegal “steering,” which is the practice of influencing a buyer’s choice of communities based upon legally protected characteristics under federal law. The classifier detects language that goes against federal, state and local antidiscrimination laws. When the classifier spots cases of noncompliance, it’s up to the system developers to figure out how the LLM will respond.

As a part of our commitment to responsible innovation, Zillow is releasing this classifier under an open-source license. This will empower others in the real estate, technology and civil rights sectors to join in strengthening fair housing practices to foster a fair and equitable real estate and technology landscape.

Explore more:

  • For more about how we’re working to promote responsible AI in real estate, explore our AI principles
  • Organizations looking to adopt the Fair Housing Classifier can access the code and comprehensive framework on GitHub
  • To learn more about how the LLM Fair Housing Classifier works, its innovators and the fair housing laws it protects, check out the Zillow tech blog post “Navigating Fair Housing Guardrails in LLMs.”