AI in Housing Regulations: Smart Insights from College Prodigy
The Department of Growth and Employment (DOGE) recently made headlines by recruiting 20-year-old college student Jack Kellogg to transform complex housing regulations using artificial intelligence. This unusual partnership pairs government expertise with Gen Z technical skills to tackle one of America’s most pressing issues: outdated and confusing housing codes that contribute to the nationwide housing shortage. Kellogg, a computer science major at University of California, Berkeley, is helping create an AI tool that translates dense legal jargon into clear, actionable guidelines.
How a College Student Landed a Government AI Project
Jack Kellogg wasn’t searching for a government position when this opportunity emerged. The sophomore was simply working on machine learning projects when DOGE officials discovered his work through academic channels. His programming expertise, particularly with large language models (LLMs), caught the department’s attention at the perfect moment.
DOGE had identified a significant problem: housing regulations across America often span hundreds of pages of complex legal language. Local officials struggle to interpret these documents, leading to implementation delays, confusion, and ultimately, fewer homes being built. The department recognized AI could help simplify these texts but lacked the in-house technical skill to develop such tools.
“When they approached me, I was skeptical at first,” Kellogg explained in a recent interview. “Government work wasn’t on my radar. But the challenge of using AI to solve a real housing problem that affects millions of Americans? That was too interesting to pass up.”
The Housing Regulation Problem
The current housing crisis stems partly from regulatory complexity. City planners, builders, and local officials face a mountain of paperwork when trying to approve new housing developments. These regulations include:
- Zoning ordinances that restrict where housing can be built
- Building codes that dictate construction standards
- Environmental requirements that mandate specific assessments
- Historical preservation guidelines that limit modifications
- Permit processes that can stretch for years
Many of these regulations were written decades ago in legal language that requires specialized knowledge to interpret. When local governments lack this expertise, housing projects stall or get abandoned entirely.
According to a Brookings Institution study, excessive or poorly written housing regulations contribute significantly to housing shortages and increased costs across the country.
AI as the Translation Tool
Kellogg’s project aims to create an AI system that performs several key functions:
- Convert complex legal language into plain English that non-specialists can understand
- Identify contradictions between different regulatory documents
- Highlight outdated rules that may no longer serve their original purpose
- Suggest simplifications based on successful models from other jurisdictions
The system isn’t designed to rewrite regulations automatically. Instead, it serves as an intelligent assistant for human policymakers. The AI flags problematic sections and provides clear explanations of existing rules, allowing officials to make informed decisions about potential changes.
“We’re not replacing human judgment,” Kellogg emphasized. “The AI is just making the information more accessible so humans can make better decisions faster.”
Technical Challenges in Regulatory AI
Building an AI system for legal text analysis presents unique technical hurdles. Regulatory documents use specialized vocabulary, contain cross-references to other documents, and often include provisions that only make sense in specific contexts.
Kellogg’s approach combines several AI techniques:
- Fine-tuning large language models on housing regulation datasets
- Creating domain-specific embeddings that capture the relationships between legal concepts
- Developing extraction algorithms that identify key requirements from dense text
- Building visualization tools that represent regulatory relationships graphically
The system requires careful validation to ensure accuracy. “Legal text is unforgiving,” Kellogg noted. “If the AI misinterprets a provision, it could lead to serious problems down the line. We’re implementing multiple layers of verification.”
Real-World Example
One test case involved a mid-sized city in California with a 437-page zoning code. Local builders complained that getting approval for basic multi-family housing took an average of 14 months, largely because different departments interpreted the code differently.
When Kellogg’s prototype analyzed the document, it found 23 instances where the code contradicted itself and 78 references to processes or departments that no longer existed. The AI generated a simplified guide that reduced the essential information to 40 pages of clear instructions.
“The planning department staff couldn’t believe it,” Kellogg shared with a laugh. “One official said, ‘I’ve been interpreting section 7.3.2 wrong for three years!’ They’re now using the AI-generated guide as a training document for new hires.”
Criticisms and Concerns
Not everyone supports the use of AI in regulatory reform. Critics have raised several valid concerns:
Algorithmic Bias
Some housing advocates worry that AI might perpetuate existing biases in housing policies. If trained on historically discriminatory regulations, the system could normalize problematic approaches.
Kellogg acknowledges this risk: “We’re very careful about training data and have implemented specific checks for language that could have discriminatory effects. The system flags potentially biased provisions for human review.”
Technical Authority
Others question whether a 20-year-old student, regardless of technical skill, has sufficient understanding of housing policy to lead such a consequential project.
DOGE officials defend the approach, noting that Kellogg works with a team of experienced housing policy experts. “Jack brings technical expertise, not policy judgment,” explained DOGE Director Maria Vasquez. “The decisions still rest with housing professionals and elected officials.”
Privacy and Security
Some local governments express concern about sharing regulatory documents with AI systems, particularly if they contain sensitive information about city planning or infrastructure.
The project addresses these concerns through local deployment options that don’t require sending data to external servers. Municipalities can run the analysis tools on their own systems if preferred.
Early Results and Future Plans
Despite being only six months into development, the AI regulation tool is showing promising results. In addition to the California test case, pilot programs in three other states have helped identify outdated or contradictory housing regulations.
DOGE plans to expand the program in three phases:
- Phase 1 (Current): Analysis and clarification of existing regulations
- Phase 2 (2024): Comparison tools that highlight regulatory differences between similar jurisdictions
- Phase 3 (2025): Recommendation engines that suggest evidence-based regulatory improvements
The department hopes to make the tools available to local governments nationwide by next year, potentially accelerating housing development across America.
Gen Z in Government: A New Model?
Kellogg’s involvement represents a new approach to government innovation. Rather than relying solely on career civil servants or expensive contractors, agencies are increasingly open to partnering with young technical talent on specific projects.
“Students bring fresh perspectives and current technical skills,” noted Dr. James Harrington of the Digital Government Institute. “They’re not embedded in traditional ways of doing things, which can be valuable when you’re trying to solve stubborn problems.”
For his part, Kellogg plans to complete his degree while continuing to consult on the project. “I’m learning as much as I’m contributing,” he said. “This experience connects classroom concepts to real-world impact in a way I never expected.”
The Bigger Picture: AI and Regulatory Reform
The housing regulation project represents just one application of AI in government modernization. Similar approaches could help clarify tax codes, environmental regulations, or healthcare policies – all areas where complexity creates barriers to implementation.
The key insight is that AI isn’t replacing regulation but making it more effective. Clear rules that everyone can understand are more likely to achieve their intended purpose than complex provisions that only specialists can interpret.
As research from the Urban Institute suggests, AI tools can significantly improve administrative effectiveness when deployed thoughtfully and with appropriate human oversight.
Lessons for Other Sectors
The DOGE-Kellogg collaboration offers valuable lessons for other organizations facing similar challenges:
- Don’t overlook young talent for specialized technical needs
- Consider project-based partnerships rather than traditional hiring for innovative initiatives
- Focus AI applications on making existing human processes more effective rather than replacing them
- Address ethical concerns proactively throughout development
- Test with real-world use cases early and often
These principles apply whether you’re reforming housing regulations or simplifying other complex systems that create unnecessary barriers.
The Path Forward
As the housing regulation project continues to develop, DOGE is already considering how similar approaches might apply to other regulatory domains. The department has created a dedicated AI task force to identify additional opportunities.
Meanwhile, Kellogg hopes his experience inspires other students to consider how their technical skills might serve the public interest. “There are so many government problems that need technical solutions,” he said. “You don’t have to wait until you graduate to make a difference.”
The unlikely partnership between a federal department and a college sophomore reminds us that innovation often comes from unexpected places. As housing regulations become clearer through this AI initiative, the real winners will be Americans seeking affordable housing in communities across the country.
Have thoughts on using AI to simplify regulations? We’d love to hear your perspective in the comments below.