The Problem of Delegated Authority in Technical Domains
Last month, the Environmental Protection Agency issued guidance on how artificial intelligence systems should be evaluated for environmental compliance monitoring. Congress never explicitly told EPA to regulate AI. The Clean Air Act, written in 1970, contains no mention of machine learning algorithms or automated decision-making systems. Yet here we are, with environmental lawyers parsing federal guidance documents that reference “algorithmic bias in emissions modeling” and “transparency requirements for automated compliance systems.”
This scenario shows a basic tension in American governance that goes way beyond environmental policy. When Congress hands broad authority to regulatory agencies, it creates a gap between what lawmakers intended and what administrators actually do. That gap gets wider every time technology advances. The question isn’t whether agencies should fill this gap, but how we set up oversight to make sure they fill it competently.
Think about how complicated this gets. EPA’s air quality experts understand particulate matter and ozone formation, but they might not know much about algorithmic validation techniques. Computer scientists understand machine learning bias, but they might not grasp ambient air quality standards. Meanwhile, Congressional appropriators control agency budgets without necessarily understanding either domain deeply. We end up with a governance structure where technical expertise, democratic accountability, and resource allocation all operate on different timescales with different knowledge bases.
The Institutional Design Challenge
Traditional oversight mechanisms assume policy problems map neatly onto existing agency expertise and Congressional committee structures. The House Energy and Commerce Committee oversees EPA, but its members typically come from backgrounds in law, business, or traditional politics rather than environmental science or data analytics. This creates what political scientists call an “information asymmetry” problem. But the asymmetry runs in multiple directions at once.
Agency capture theory suggests that regulated industries will develop disproportionate influence over their regulators through superior technical knowledge and sustained engagement. But in rapidly evolving technical domains, even industry actors may lack comprehensive understanding. When the Federal Communications Commission regulates 5G networks, telecommunications companies understand the engineering better than FCC staff. But cybersecurity implications may be best understood by researchers at the National Institute of Standards and Technology, while economic effects fall under Federal Trade Commission expertise.
This fragmentation creates opportunities for what I’d call “expertise arbitrage.” Skilled policy entrepreneurs can use superior technical knowledge to shape regulatory outcomes in their favor, not through traditional lobbying, but by positioning themselves as essential sources of specialized information. Regulatory capture becomes more subtle and harder to detect. It operates through the provision of technical expertise rather than through obvious political pressure.
Congressional Oversight Tools and Their Limitations
Congress has several formal mechanisms for overseeing agency implementation: appropriations riders, confirmation hearings, oversight hearings, and the Congressional Review Act. Each works differently when applied to technically complex regulatory decisions. Appropriations riders work well for broad policy direction but poorly for technical specification. You can defund an EPA climate modeling program, but you can’t effectively micromanage the statistical methods used in climate model validation through budget language.
Oversight hearings reveal these limitations clearly. When Senator Joe Manchin questioned Federal Reserve officials about climate stress testing methodologies in banking supervision, the exchange highlighted fundamental mismatches between Congressional questioning techniques and technical regulatory substance. Senators are good at probing political motivations and policy priorities. They struggle to evaluate the technical adequacy of Value-at-Risk models or scenario analysis frameworks.
The Congressional Review Act presents a different problem. It allows Congress to overturn specific regulations with majority votes in both chambers, but it’s a blunt instrument. CRA resolutions can’t modify regulations, only eliminate them entirely. This creates perverse incentives for agencies to write broader, more flexible rules that are harder to challenge legislatively, potentially reducing regulatory precision in technically demanding areas.
The Expertise Validation Problem
Modern regulatory agencies increasingly rely on external technical advisory committees, peer review processes, and inter-agency consultation to validate their technical decisions. The Food and Drug Administration’s Vaccine Advisory Committee shows this approach in action: external experts review clinical trial data and provide public recommendations on vaccine approvals. But advisory committee structures work differently across agencies and policy domains, creating inconsistent standards for expertise validation.
Some agencies, like FDA, have statutory requirements for external advisory committees with specific membership criteria. Others, like the Department of Homeland Security, rely more heavily on classified inter-agency processes that provide less public visibility into technical decision-making. The Nuclear Regulatory Commission operates under different expertise validation requirements than the Federal Aviation Administration, despite both agencies making highly technical decisions with significant public safety implications.
This inconsistency matters because expertise validation mechanisms shape not just the quality of regulatory decisions, but also their democratic legitimacy. When EPA’s Science Advisory Board reviews the technical basis for air quality standards, that process provides both substantive validation and procedural transparency. When similar decisions happen through less formal inter-agency consultation, the technical quality may be equivalent but the democratic accountability is weaker.
Structural Reforms Worth Considering
Several structural modifications could improve oversight without sacrificing technical competence. Congressional committee staff could be professionalized differently, with longer tenure and deeper subject matter specialization, similar to how Congressional Budget Office staff operate. This would reduce the information asymmetry between agencies and oversight committees without requiring members of Congress themselves to become technical experts.
Inter-agency technical review processes could be standardized and made more transparent. When EPA consults with Department of Energy on energy efficiency standards, or when Federal Trade Commission coordinates with Securities and Exchange Commission on cryptocurrency regulation, these interactions currently happen through informal channels with limited public documentation. Making these processes more formal would improve both technical decision-making and democratic oversight.
Most importantly, we could experiment with sunset requirements for highly technical regulations, paired with mandatory technical review processes. Instead of regulations lasting indefinitely until explicitly changed, complex technical rules could include automatic expiration dates that trigger comprehensive technical and policy review. This would force agencies to regularly re-justify their technical approaches while providing natural oversight opportunities for Congress.
The deeper question these reforms address is whether democratic governance can effectively manage increasingly technical policy domains without either sacrificing democratic accountability or accepting technical incompetence. The answer probably depends less on finding perfect institutional solutions than on designing oversight systems that stay adaptive as both technology and governance challenges keep changing.