MUSINGS FROM BOSTON
As Daniel mentioned last week, we spent the first part of Labor Day weekend in Boston for the American Political Science Association annual meeting. We were different, obviously, than most people at the conference in that we were not to present new academic research, but provide a “practitioner” perspective. It also was an opportunity to learn more about several fields of study within political science and make new connections with scholars studying Congress. (Thank you to the folks who made the time, including Charlie Hunt and SoRelle Wykcoff Gaynor, who have a great weekly podcast on Congress called Highway to Hill.)
One thing stood out among the academics we met with whose work we admire: most had direct experience working in congressional offices, either as staff or long-term research fellows in graduate school. At a panel discussion on the study of congressional history, Arizona State University Steven Smith (the co-author of a recent authoritative history of Senate leadership) lamented that this type of experience is becoming less common in political science and noted the value of direct working experience for scholars. He himself had worked for Tip O’Neill.
We couldn’t agree more. We hope that scholars continue to have these opportunities and that graduate programs remain welcome to admitting staff leaving Congress to pursue academic careers because these types of experiences lead to better research questions and better perspectives in interpreting findings in the data. Knowing what’s important inside Congress to the people working there sharpens theoretical and methodological approaches to several subfields.
Congressional offices, for their part, need to remain open to hosting fellows, participating in research projects, and accepting interview requests from scholars because the more is learned about congressional practices and processes, the more effective proposals to change what isn’t working can be. Many of the American Governance Institute’s suggestions for chamber and caucus rules reform, appropriations requests, and modernization ideas we have developed through political scientist scholarship we’ve read and conversations we’ve had.
Academic fields of study are built around and governed by specific standards of evidence and methodological rigor. As a historian, the answers I could produce to the questions I was pursuing depended enormously on the archival and primary source material available. For political science, robust datasets may provide some of the answers. We know research questions are constrained by the legislative branch data available, as evidenced by the prevalence of the use of well-trod datasets like floor roll call votes and legislative effectiveness scores – even after offices have figured out how to game that system. They’re also constrained by political scientist awareness of publicly available information that could be used as data.
That’s why we also spent some effort in Boston encouraging scholars to reach out on data accessibility issues within the legislative branch so we can bring them up in forums like the Congressional Data Task Force meetings. It’s also why we join projects like the Legislative Branch Data Map, a resource researchers not only can utilize as a single-source roster of what’s available, but suggest additions. And although we see AGI’s role as being a bridge between different communities with a shared interest in federal governance, it would be great if academic researchers engaged directly with CDTF during its quarterly public forums, especially as legislative branch offices make strides in machine readability for the AI age. We’re always happy to do so for people, too – just reach out.
CONGRESSIONAL DATA
Speaking of the Congressional Data Task Force, its third meeting of 2026 took place via videoconference last Thursday. In previous years the September meeting took place at the Congressional Hackathon, but the House decided not to host that event this year and is moving to an every-other-year schedule. (In our view, this is a regrettable, if understandable, decision.) Here’s a summary of the more conventional meeting, which will be available on the Congressional Data Coalition website in a more complete form once the task force posts the event video.
Our partner Luis Kimaid from Bússola Tech joined all the way from Italy to invite the audience (and all of you) to watch proceedings from the sixth LegisTech Forum on September 22 through the 25th in Washington. The forum, which Bússola is organizing along with AGI, the Organization of American States, and the House of Representatives, will bring together parliamentarians, officers, and technologists from about 30 parliaments around the world to discuss how technology is transforming and challenging legislatures, with emphasis on how parliaments themselves are designing their own digital innovations. In-person participation is limited to specific parliamentary delegations.
The forum is an opportunity for participants to share their work and perspectives and discuss with peers common interests and concerns on topics like artificial intelligence, knowledge management, digital legislative processes, data interoperability, and public access to information among others. Representatives from the House, Senate, and Library of Congress will attend.
Matching bill text
Living on the other side of the HouseNet firewall, the live demonstrations of the latest version of the House Clerk’s Comparative Print Suite and of the Congressional Research Service’s Text Analysis Project were exciting moments in the public forum. These tools are enormously valuable in providing insight into legislation for those who have access. Together, these tools can track how legislative language travels across bills – how particular language is incorporated into various bills and how it is reintroduced in multiple Congresses – and provide deep analysis into how amendments would change bills and bills would change the law. Knowledge is power, and timely knowledge empowers legislators to know what they’re voting on and be able to advance their own proposals. These tools are major success stories for legislative branch technological development. The limits on their availability also provide a window into the policy choices made regarding who they empower and the politics they must navigate.
Both of these tools were developed for particular legislative branch users and purposes. The Comparative Print Suite is a collaboration between the House Office of Legislative Counsel and the Clerk’s Office to assist staff in the bill drafting process, pushed forward by Rep. Posey, who wanted to understand how the text of legislation changed between drafts. TAP helps CRS produce bill summaries by giving analysts context and allowing them to find the correct versions of introduced bills that have been sucked up into large legislative packages. Certainly, limited legislative branch resources should go first to solving legislative branch problems, but these platforms would be of enormous help to the public trying to understand what Congress is doing through legislative data like bill text. Even broader availability within the legislative branch would be useful.
It also would help surface some of the limitations of these tools, such as the TAP’s underidentification of related bills, a result that is downstream of CRS’s political decisions more than limitations of the tool itself. That issue is an occasional but recurrent problem that runs across the services Congress.gov provides to members and the public: technically correct information that at times has the consequence of inadvertently misleading the users. For illustrations, notable examples include pages for the top ten viewed bills, the appropriations table, and (of course) the related bills feature. These problems do not outweigh the tremendous value of making these features available, but do point to the occasional misalignment between the purposes of the Congress.gov website and its implementation.
The Congress.gov team at the Library of Congress has included some related bill tracking functionality to the website, which is a welcome development. Pages for specific bills feature a “Related Bills” tab, which lists records including a link to related bills, their sponsor, title, specific relationship, and latest action. Kimberly Ferguson from the Library also demonstrated on the videoconference additions to members’ profile pages on Congress.gov under the “Member Activity” section that identify legislation members have reintroduced from a previous Congress. This addition addresses a recommendation of the House Select Committee on the Modernization of Congress, which called for creating ways that allow members to take credit for legislation that they authored, but for various reasons were enacted in a different legislative vehicle. We suspect that the underinclusiveness of the identification of related legislation may become a thorny problem.
How the Library is thinking about relationships, matching legislative text, and other services to members and staff in the next version of Congress.gov is something worth exploring during the upcoming public forum for the website on September 24 from 1:00 to 3:00 PM. We will be at the hybrid meeting and are looking forward to the opportunity to engage with the Congress.gov team.
Lobbyist disclosure: new meeting, same answer: The House and Senate have successfully created a unified search capacity across both chambers for the lobbyist disclosure website (LDA.gov) under development that allows public tracking of information submitted in lobbyist disclosure reports.
Four years into the project, however, we’re still waiting for unique identifiers for individual lobbyists so they can be tracked with precision. The Senate has been the lead on this project, and whatever decisions need to be made to move this feature along remain in the misty future.
Maker space – CDTF edition:
Dave Zvenyach returned to the CDTF to provide an update on AGI’s projects to improve the accessibility of appropriations data that is locked within committee reports and provide context on how appropriations report language changes over time.
The first project, on tracking appropriations spending over time, is challenging because the appropriations tables are challenging for a computer to ingest not only for their accounting-related formatting with numbers, but nested text that indicates to human readers specific programs under larger budget lines that make it difficult for computers to draw relational connections to structure data. We know this because we’ve had to build legislative branch appropriations datasets by hand.
The second project, on tracking changes in appropriations report language over time, makes it possible to see how report language has changed over the years.
Both projects make use of advanced AI capabilities,, which Zvenyach has been experimenting with thanks to a grant AGI has received. OCR can take data out of the source PDFs, then Gemini Pro and Nemotron parse the data to give it usable structure. It’s working pretty well, with 46 out of 52 appropriations bill reports reaching a strong matching rate between what the agents generated and what’s on the PDF.
However, blindly relying on AI would be a serious mistake. It routinely makes serious mistakes. We have been benchmarking different AI tools to evaluate their analytical capabilities for legislative analysis as well as using them to assist with creating deterministic tools to perform the work. Dave called for others interested in this work to help us engage in quality assurance on the data derived from these efforts.
Daniel reiterated to the forum that this is an interim step in making past appropriations data machine-readable that we think should be available from official sources in the House and Senate. We’ll have a more in-depth update on these projects later. The Library of Congress helpfully noted that our work to make current and historical data available is complementary to their efforts to make certain sets of legislative data publicly available on a going-forward basis. We agree and would welcome the opportunity to share ideas, techniques, approaches, and work to close gaps in data availability.
Will Adler from the Bipartisan Policy Center demoed an open source legislative tracker that his team at BPC’s elections project built to help organizations track state and federal legislation. The goal of the tool is providing a collective way for legislative affairs teams to collaboratively track various bills, set team positions, and comment on progress without having to purchase expensive software licenses from vendors. From our perspective, this is a great way for organizations or organizational coalitions to track and share information on legislation that is of interest.
Called FloorVote, it pulls in LegiScan account data on introduced legislation across 50 states and Congress and uses AI to score bills for relevance to issue areas and other criteria that users can identify. Users can then set customizable filters to assist in legislative affairs workflows. The system also can set calendar notices for upcoming hearings on relevant topics and legislation.
BPC conceived of the tool to assist partner election law organizations in keeping track of legislative activity for no additional cost beyond LegiScan and Cloudflare accounts. This makes the tool affordable to just about everyone. Adler noted that user feedback and feature requests can be made on the FloorVote GitHub repository. The platform has been public for only a few weeks, and users can try a demo.
From CAO
The Office of the CAO announced a new staff leave tool and AI programming for staff. The House Digital Service has started a pilot with 10 member offices to test a new staff leave and telework request system called LeaveTrack. HDS product manager Ananda Bhatia said the platform will allow staff to request time off and telework and see their time off accruals and office policies. It also will sync with Outlook so member offices can view who’s out of the office or working remotely conveniently.
The goal is the tool will make it much easier for staff to manage these back-office tasks, replacing bespoke spreadsheets or text documents.
CAO’s Stephen Dwyer highlighted a hands-on AI testing event through the House AI Center happening today and two open positions for the new office (more on that below). The AI Center also has launched its AI Innovators pipeline project that the Committee on House Administration announced recently, which will give tech-savvy staffers access to secure data sandboxes in which they can build and test new AI-powered tools for their offices. The project is limited right now, but is a major development for in-house (or in-House) modernization capacity.
More on the Data Map: Mark Caudill from the Government Publishing Office explained the functionality of the Legislative Branch Data Map, which is living on GPO’s innovation hub (along with CDTF records). It’s set up for different types of users, with links to the webpages that host different data sources and access to the raw DCAT metadata for those who want to repurpose the data. GPO also is accepting feedback and contributions through the GitHub for the project, so this is very much an open, growing project. This project is a fun success story, starting off as an AGI project at the Congressional Data Coalition website, then becoming a joint HDS-CDC project collaboratively built at the last Congressional Hackathon, and now fully managed by the Congressional Data Task Force with buy-in from internal and external stakeholders.
Kirsten Gullickson from the Clerk’s Office added that her team worked this summer on developing subject matter terms in a controlled vocabulary to add to the data map at a later date, perhaps at the end of the year. It should aid navigating the resources available.
Congressional MPCs
Caudill also shared that the documentation for the GPO’s new MPC server is up on the office’s GitHub, which folks can use to point agents to the source of authoritative legislative branch data rather than bouncing around the web or making things up.
Forum participants asked several questions about other MPC opportunities, including whether the Library is developing one and how the House is thinking about providing structured data to AI vendors to use to train agents directly. These ideas clearly are being explored in the House from the answers provided, but it’s a question to ask the Library again during the upcoming forum. While providing data via MCP is useful, the emphasis should remain on making more congressional data publicly available, whether in bulk, via API, or through an MCP server.
Wrapping up
The Congressional Modernization Staff Association made its first official appearance at a task force meeting, although its presence is often felt because its founder and long-time president, Ananda Bhatia, has presented at many meetings. With Bhatia installed at the House Digital Service, she has passed the MSA leadership torch to Paul Hurton, a staffer with the Senate Homeland Security and Government Affairs Committee. In addition to introducing himself and the MSA, he noted the organization’s new Substack (more on that below). Membership is open to all congressional staff, including in support offices and agencies.
The updated version of Senate.gov is almost ready for launch, Secretary of the Senate webmaster Arin Shapiro reported. The goal of the update was to improve accessibility, including new “mega menus” to help users drill down into different sections of the site faster. The new site also will have an alert when the Senate is in session at the top and a link for the live video feed as well as a larger footer. He provided an excellent visual overview of the forthcoming changes.
CALENDAR
Three bills of note heading to the House floor this week before the chamber goes poof again until after the midterms:
H.R. 9360, the Advisory Committee on the Records of Congress Sunset Act, which closes a committee of House, Senate, and National Archives officers that has worked through management and preservation practices of congressional records at the Archives.
H.R. 9342, the GPO Modernization Act, which updates a variety of requirements for GPO operations for the digital age.
H.R. 10204, The Legislative Branch Agencies Clarification Act, which eliminates the presidential nomination process for appointment of the Librarian of Congress and Director of GPO, allowing them to be chosen exclusively by Congress through a bipartisan commission. It also removes the Librarian’s supervisory authority over the Register of Copyrights and the Copyright Office while maintaining the support it receives from the Library. The bill also establishes an Office of Inspector General at the Copyright Office and requires GPO to establish a human management system.
The Congressional Modernization Staff Association is now on Substack. Congrats to new President Paul Hurton and outgoing President/founder Ananda Bhatia, who started the whole thing about a decade ago. The MSA Substack has a good calendar section built out that should be filled in with more specifics for several upcoming events, so check back.
Of note, they highlight the House AI Lab, hosted by the CAO’s House AI Center TODAY from 2:00 PM to 4:00 PM in the Rayburn Gold Room (Rayburn 2168). The event is open to legislative branch staff and has a registration page, but walk-ins should be fine.
Per Steven Dwyer of the CAO during the aforementioned Congressional Data Task Force meeting, the workshop will be a hands-on event with beginner, intermediate, and advanced experience tiers.
Several DC-based civic tech groups and gov tech professionals have put together a Google Meet event “Lightning Demos: Budget & Approps Edition!” with four demos of related projects on October 2 from 12:00 to 1:15 PM. Register here.
Daniel will be speaking at the International Conference on Legislation and Law Reform (iLegis) taking place October 29-30 in Washington D.C. The conference brings an international perspective on legislative drafting, the legislative process, and current legislative issues. For more information, visit the iLegis registration page.
JOBS
The House CAO has posted two positions it is seeking to fill for its new House AI Center, which will be critical to congressional development of AI capabilities and overall modernization. These should be really fun jobs and a great opportunity for people to make their mark on public service in a whole new way.
One opening is for an AI software engineer to “build innovative software, prototype new ideas, and collaborate with product managers, designers, and engineers to tackle unique technical challenges inside the House of Representatives.”
The other is for the AI program coordinator, who will support the center’s initiatives and stay connected to House customers and stakeholders.
On the topic of employment, Omar Awan at Hillclimbers notes that the latest staffing data from House member offices show that the initial bump in staffing growth in 2024, the high-water mark for total non-intern staff since 2012, has not lasted. The House is down nearly 240 non-intern staff since that period, and has more than 1,000, or 13% fewer staff since 2010.
ODDS & ENDS
The Office of the Whistleblower Ombuds has announced its latest fact sheet, Federal Judicial Branch Whistleblowing, which identifies the principal avenues available to report judicial branch misconduct.
The Foundation for American Innovation’s Soren Dayton and James Wallner have a terrific essay for NYU Law’s Democracy Project on how the parties need to get over “faction stigma” in order for Congress to work better. “Factions are not the source of Congress’s dysfunction,” they explain. “The source is the centralized process Congress has engineered to suppress conflict between its members.” We talked about this very issue with them a few months ago.