No. We need greater tech literacy in the agencies we already have.
Tim O’Brien is an Affiliate Associate Professor at the University of Washington’s Daniel J. Evans School of Public Policy & Governance.
The idea of creating a digital regulatory agency in the United States, a so-called “FDA for algorithms” is not new. In 2016, researcher Andrew Tutt published a paper with that very title, proposing the creation of a new federal agency to classify algorithms into regulatory categories, prevent the introduction of harmful algorithms into the market, and exercise broad authority to impose disclosure requirements and usage restrictions. Since then, we’ve seen more proposals over the years from the public sector, the private sector, non-profit researchers, and industry watchdogs.
While the principal aims in such proposals remain imperative, the addition of a new administrative agency is not the answer, primarily because adding complexity to existing bureaucracy will likely do more harm than good. This is especially true in a world where there really is no more “tech sector”: every industry is now a tech-centric industry. And it will do little to address the real issue: the lack of tech literacy in government, at all levels. If the shift from analog to digital cuts across all businesses and the public sector, then digital literacy has to permeate all efforts to create law and regulate industries in the age of digital technology. The public will be better served if this capacity is built within existing regulatory agencies.
What’s At Stake: NTSB Offers a Case Study
The National Transportation Safety Board (NTSB) was established in 1967 and mandated by Congress to investigate accidents and determine likely causes without assigning fault or blame. In March of 2018, the NTSB opened an investigation into a new type of accident: a crash involving an Uber automated test vehicle that struck and fatally injured a 49-year-old pedestrian in Tempe, Arizona.
The investigation was multi-faceted, involving study of the road, signage, and markings, mechanical condition of the vehicle, human decision-making, risk mitigation procedures for testing, and even the safety culture at Uber Advanced Technologies Group. But it also involved something relatively new – an examination of the algorithmically-powered automated driving system (ADS). The final report is comprehensive, and leaves no doubt that the NTSB either possessed or acquired the expertise necessary to uncover a serious object classification flaw in the ADS, namely the inability of the system to correctly perceive a pedestrian outside of a marked crosswalk.
While questions remain about why Arizona prosecutors assigned criminal liability to only the safety driver and not the company, the NTSB’s investigation succeeded in fulfilling its statutory mission, which included making recommendations to its sister agency, the National Highway Traffic Safety Administration (NHTSA). The outcome was consistent with the way in which the NTSB routinely engages existing agency structure and process to make public recommendations to the NHTSA, most recently citing Tesla’s ‘Full Self-Driving’ beta to call for stricter standards on automated vehicle technology.
Critics might cite the slowness of progress as proof of interagency ineffectiveness. The federal government is massive, bureaucratic, complex, and slow. But do we think the addition of a new agency that places new demands on interagency collaboration will make the situation better or worse? Would a new digital agency have done a better, faster, or more cost-effective job? If this new agency existed in 2018, what would its role have been in the Arizona investigation? Would it have led or supported the inquiry into the ADS system?
Every Industry is in Tech, Tech is in Every Industry
These counterfactuals are helpful to consider when querying how a new agency would work when companies traditionally considered “tech” are on the move via vertical investments and acquisitions, while companies in non-tech sectors are transforming into digital businesses, effectively becoming “tech companies” themselves. When tech companies look for new market opportunities via M&A in non-tech verticals, we still call them tech companies, but they’re really not.
Consider Amazon, which finds itself in retail, logistics, healthcare, banking, filmmaking, home security, and gaming sectors, and most recently made a bid to enter the vacuum cleaner market. There are regulatory agencies today for each of these market sectors, independent of each sector’s varying dependence on digital technology. Alternatively, considering a single vertical (financial services, for example) pulls in Google Pay, Apple Pay, and Samsung Pay among a slew of device-centric digital payment platforms in the market. These firms are technology companies at their core, but are also subject to financial services regulatory regimes enforced by existing agencies, an intersection that American Banker describes as the “biggest puzzle for regulation to solve”. Alliance for Innovative Regulation CEO Jo Ann Barefoot points out that the solution is “not structural change in the sense of creating a new ‘Department of Financial Technology Oversight’ … but rather new modes of working by the regulators.” She is quite right. And new modes of working should and will include incorporation of tech talent into today’s structure to adapt to a changing regulatory landscape.
This problem is two-sided: tech companies are moving into verticals (e.g., Apple Pay), and vertical players are moving deeper into tech (e.g., banks’ increasing reliance on algorithmic decision systems). For existing agencies, the former is very much in focus. The latter is a much bigger problem if measured by the vastness and breadth of issues commanding the attention of existing agencies who lack the tech expertise to even frame the problems correctly (much less to initiate rulemakings to address them). This challenge is unlikely to be solved by adding a new agency into the regulatory mix. Each agency tasked with oversight of its respective sector(s) needs to ask Congress to fund acquisition and retention of technology expertise to fulfill its mandated mission.
Intermediate Steps Toward Progress
There are small signs of progress in this direction. Legislative proposals to address the risks and harms associated with algorithmic systems have come to the fore in recent years and provide a more reasonable alternative to creating a new peer agency. One such proposal is the Algorithmic Accountability Act, first proposed in 2019 and amended in 2022. The Act is centered on the FTC and includes a provision that “establishes a Bureau of Technology to enforce this Act and support the Commission in the technological aspects of its functions.” Greater tech literacy in the FTC is a good thing and a work already in progress, but creation of a Bureau within the FTC as a center of excellence for technical expertise feels like an intermediate step. This might be pragmatic in the near term, as solving the tech literacy deficit is not a step-function fix that magically transforms regulatory agencies into hubs of tech expertise.
Several agencies have taken the approach of engaging external tech advisory committees, under the provisions of the 1972 Federal Advisory Committee Act (FACA). The Federal Communications Commission (FCC) has a Technological Advisory Council. The Commodity Futures Trading Commission (CFTC) has a Technology Advisory Committee. The Department of Commerce (DOC) has both an Emerging Technology Technical Advisory Committee and a Visiting Committee on Advanced Technology. The list goes on. The notion that many (if not most) of these advisory councils are comprised in part by private-sector executives working for companies that are regulated by the agencies they advise raises the necessity of a separate discussion about public-private conflicts, lobbying, corporate power, and revolving doors.
But for the time being, external advisory bodies represent a shortened path between regulatory problem solving and understanding of fast-changing technical issues. A new bipartisan Senate bill, for example, aims to fill a regulatory void by placing oversight of cryptocurrencies including Bitcoin and Ether under the purview of the CFTC, creating an immediate need to add to the agency’s existing knowledge base of derivative securities with crypto expertise. This kind of technical proficiency, experience, and know-how must eventually become part & parcel to these agencies at their core. Administrative agencies cannot lose sight of the long game by simply relying on external experts or creating captive entities to house their tech talent. The end goal is an agency in which all personnel possess a baseline level of technology expertise as part of the agency’s culture.
It bears repeating that suggestions of new, tech-centric organizational entities within existing agencies are embedded in legislative proposals because these agencies require a mandate (and funding) from Congress to create them. If more of these internal centers of expertise take shape, the knowledge gap between sector experts in administrative agencies and lawmakers in Congress will widen. This is a problem, and the solution is exceedingly simple: bring back the Office of Technology Assessment (OTA), created in 1974 to provide lawmakers with objective analysis and assessment of new technologies and their impacts, but defunded in 1995 following the Republican takeover of the U.S. House of Representatives amidst a promise to voters to reduce government spending and curb congressional hostility to business. Calls to revive it are increasing, with consistent and compelling rationales. As the Center for American Progress’ Bianca Majumder wrote in 2019,
“[C]ongressional resources for science and technical support—including science and technology functions within the Government Accountability Office (GAO), the Congressional Research Service (CRS), and the National Academies of Science (NAS)—did not and do not have the internal structure, structural autonomy, and culture that made the OTA so successful in providing direct, on-demand science and technology support to lawmakers.”
The OTA is still a legally authorized organization, and reviving it would help bridge the inevitable divide between growing tech expertise in administrative agencies and the lack thereof among lawmakers in Congress.
All Regulatory Agencies Must Be Digital Regulatory Agencies
The goals proposed in Tutt’s 2016 article have proven very durable, as reflected in today’s federal and state legislative proposals that would require governments and businesses to provide algorithmic transparency and submit to accountability regimes for harms that occur. The growing awareness of algorithmic harms and continued work on legislative and regulatory remedies reflects the importance of increasing tech literacy in government. This can and should be done without a structural addition that would serve to burden existing agencies by requiring them to dedicate even more resources to cross-agency collaboration.
The administrative state is complex enough already. It’s time to prioritize agency modernization through talent acquisition and the fostering of tech-centric culture transformation inside agencies with vertical expertise, and give lawmakers in Congress the resources they need to maintain baseline knowledge parity with the agencies to whom they’ve delegated rulemaking.