▶ 0:00:42Without objection, the chair may declare a recess at any time. Also, without objection, the hearing record will remain open for five legislative days so that members may submit any materials they wish to be included therein. The title of today's hearing is modernizing public access to legislative data and information. Last december, the subcommittee held a hearing on the future of constituent engagement.
▶ 0:01:06And the big takeaway was that we need to keep pace with the way people communicate outside of congress, or we risk losing the ability to connect with the people we need to hear from the most. I think that we can all agree that sending form letters and emails to constituents who are using the latest ai tools and technologies doesn't always cut it. We need to meet our constituents where they're at and communicate like they communicate.
▶ 0:01:34The good news is that the subcommittee is actively working with house digital service to make that happen. Following our hearing in december, the subcommittee secured funding from the modernization initiatives account for a major new project that reimagines constituent engagement as dynamic rather than static and stuck in time.
▶ 0:01:57The project will pave the way for members to easily access a wide range of innovative and secure technologies for communicating with their constituents across multiple platforms. And as the way people communicate evolves, so will the technologies and tools available to members. I am really excited to see this project get off the ground. At the same time, I know we need to do a lot more to better connect with the people we serve in the people's house.
▶ 0:02:27It's one thing for us to reach out to our constituents, but another thing for them to reach out to the house or congress and find information and answers to their questions. We really need to think about the user journey and what that looks like. They will probably begin with a search engine query that directs them to house.gov or congress.gov. But once they get to those sites, is it easy for them to find what they're looking for?
▶ 0:02:55Is information presented in a way that makes sense, regardless of how much they know or don't about congress? And are the search tools and the sites powered by ai so that the queries don't have to contain specific words or phrases. I'm afraid the answer to all of these questions is no.
▶ 0:03:14Before the hearing today, I spent some time on both sites to get a sense of what constituents would experience, and my overall impression was that they'd likely be overwhelmed with the amount of information on both the home landing pages, and frustrated by the search tool capabilities. This is not the experience that we want them to have. You only get one chance to make a good first impression, and so these entryway sites matter a lot.
▶ 0:03:46They should be the digital voice of representative democracy. Not a long list of hyperlinks. The challenge of figuring out where to find information extends to members and staff as well. There are over 600 sites within the house.gov domain, about 70 of which belong to house officers and house support entities.
▶ 0:04:10Massive amounts of legislative data and information are spread across these sites, making it difficult to know where to find what you're looking for. And because different sites sometimes have different versions of the same data. Figuring out which data to use can also be confusing. If you want the most up to date information available, you have to continually monitor and refresh your screen. Because most of these sites don't have automated real time alert systems.
▶ 0:04:38This means that staff who are tasked with tracking the floor, calendar or bill status updates are tied to their screens, which strikes me as the opposite of efficient. These challenges explain why house staff have taken matters into their own hands, and built ai tools that automate these tedious tasks, and alert staff of any updates or changes in real time.
▶ 0:05:03And as we learned at our recent full committee hearing on crs and ai use, house staff have also built ai tools that summarize bills providing nonpartisan bill analysis. Find related bills and much more. The demand is clearly there and staff are not going to wait for the house, the library or crs to build the tools they need to do their jobs better.
▶ 0:05:28We want to support and encourage these efforts because they help the house stay current. In a world where ongoing innovation and change are the norm. We also want to ensure that the house and the library are capable of adapting and innovating in ways that meet current and future user needs.
▶ 0:05:47The data and information that various house websites and congress.gov maintain are essential to legislative branch functions, but the website's risk becoming obsolete as more and more users seek platforms and tools that can perform searches, complex tasks and analyze quicker and better. There's a lot of opportunity here to reimagine how we make the house and the work that it does make more sense to the american people.
▶ 0:06:16I'm looking forward to learning more about current and potential efforts to enhance the public user experience, and exploring some innovative ideas for making the digital voice of congress more meaningful and accessible. At this time, I would like to now recognize the ranking member, miss torres, for the purpose of providing an opening statement. Speaker 2: this from you?
▶ 0:06:43Speaker 3: well, thank you, chairwoman, and good afternoon, everyone. Today's hearing, modernizing public access to legislative data and information builds directly on this subcommittee's long standing work to strengthen transparency, usability and public trust in congress. According to the pew research center, just 17% of americans say they trust the federal government to do what is right.
▶ 0:07:14Most of the time. That crisis of confidence is rooted not only in partizanship, but in the difficulty the public faces in simply understanding what congress does and how we do it. My constituents in the inland empire work very long hours to put food on the table for their families.
▶ 0:07:37In trying to meet sports and, you know, that long agenda of being a parent, they want to know what congress is doing for them to alleviate their stress. They want to be able to use 21st century technology to get reliable information easily. For many americans, their relationship with congress begins online.
▶ 0:08:00It is simply not acceptable to offer dial up type service when the rest of the world is on 5g. Plus. Whether constituents are trying to identify their representative, ask for help for matters related to federal agency track legislation, read a committee report or understand how an amendment changes a bill.
▶ 0:08:26We must meet them where they are and quickly get them the information that they need. For example, regular people are not talking about h.r. Numbers. They or the technical names of bills or acronyms. The big ugly bill technical name is an act to provide for reconciliation pursuant to title two of h.r. Dot com dot rez .14.
▶ 0:08:56Try finding that. Try. Find out how congress voted on it and you will click through a page after page after page. And this is a piece of legislation that significantly impacts millions of people. Good or bad, they need to be able to find information about this legislation.
▶ 0:09:23If you need to know exactly what to search for, you're likely give up before finding any information. There are 625 publicly available websites within the house.gov umbrella alone. Having so many different places to get information creates unique challenges.
▶ 0:09:44Many americans simply do not know where the right information lives online, and often end up in a frustrating rabbit hole of clicks. That is unfortunate. People shouldn't feel like they are getting lost in a click spiral to get simple public information about their government.
▶ 0:10:07At the same time, the library of congress has documented through extensive research on congress.gov that the public often finds legislative information overwhelming, cluttered, difficult to search, and hard to navigate. 73% of surveyed users reported they could not find what they were looking for and simply gave up.
▶ 0:10:31This subcommittee exists to make congress more functional, more accessible, and more accountable to the public. That mission must include modernizing the digital presence of congress and the way the public access legislative information across the work conducted by our witnesses. A clear theme emerges.
▶ 0:10:55We cannot meaningful meaningfully increase transparency or public trust until we make legislative information easier to find, understand and use. That requires more than cosmetic improvements. It demands modernized data pipelines and machine readable formats.
▶ 0:11:17Consistent information architecture across congress websites, robust research, search and discovery tools, and opportunities for public dialog, but not chat bots and accessibility. Simple english and multi-language support that meets the 21st century expectations. This is not just a technology conversation.
▶ 0:11:41This is a conversation about making sure democracy works for all americans. It is about strengthening the public ability to follow their government and participate meaningful in the democratic process. Our goal today is to understand what investments, what governance changes, and what modernization strategies will allow congress to meet the moment and meet the public where they are.
▶ 0:12:07I want to once again thank our witnesses for their work, and I look forward to a productive discussion. And I yield back to the chair. Speaker 1: thank you, ranking member torres. And I'd now like to introduce you all to our panel of witnesses today. Our first witness is kirsten gullickson.
▶ 0:12:30Miss gullickson serves as the coordinator of the congressional data task force, where she works with legislative branch organizations to improve accessibility, coordination and long term stewardship of congressional data and information. She is also the director of analysis and quality assurance in the office of the clerk, where she has served for more than 25 years. Our next witness is john rutledge, the deputy chief information officer for the library of congress. Mr.
▶ 0:12:59Rutledge serves as a senior advisor to the cao and the librarian of congress on all technology matters. He directs and manages the day to day operations of the library's office of the chief information officer, and works to ensure that the library's it services are aligned with its strategic mission. Mr. rutledge joined the library in 2015 and has been a driving force for it. Modernization and centralization. And our final witness is charlotte lee.
▶ 0:13:26Miss lee is an award winning customer experience and human centered design practitioner. She's assisted senior government executives from over 15 federal agencies in designing, developing and implementing their plans for digital transformation.
▶ 0:13:42Miss lee also serves as an industry advisor to the american council for technology industry advisory, a nonprofit educational organization focused on improving government through the innovative application of technology, and a lecturer at the university of virginia's customer service leadership institute. Thank you so much to the panel of witnesses for being here today.
▶ 0:14:09Please remember, because I evidently did not to press the button in the microphone in front of you, um, to, um, make sure that the light is on. Um, when you begin to speak, the timer in front of you will turn green. After four minutes, the light will turn yellow. And when the red light comes on, your five minutes have expired. And we ask that you kindly wrap up. And at this time, I would be pleased to now recognize Mrs. gullickson for five minutes for an opening statement. Speaker 4: okay. Thank you chairwoman.
▶ 0:14:38Thank you, ranking member. Um, it's an opportunity to testify today. As you said, I have served the house, am I? I'm on. Right. I'm on. Okay, great. I serve the house for more than three decades. When I began in 1993, legislative information showed up each morning. It was tied in twine, and it was in big stacks and printed pages. Today, legislative information is available, as we all know instantly from virtually anywhere.
▶ 0:15:06Technology has changed dramatically, as you both said, but our responsibility has not. Congressional information must remain trustworthy, understandable, and available to those who need it. Congress is not only a lawmaking institution, it is one of the nation's most important information institutions. And today, how we organize it, connect it, and deliver that information matters just as much as the information it's itself.
▶ 0:15:32Public access, as you both said, is ultimately about trust, and people should not need expert level knowledge of our procedures or our structure to understand congressional information. As I prepared today, a colleague mentioned wayfinding. Anyone who has tried to navigate the capitol complex, even those of us who have been here for a while. It's difficult. Our digital environment, as you both alluded to, is the same every day.
▶ 0:16:00People try to find their way across our ecosystem of websites to information they need, and often they don't even start on a congressional website at all. Instead, they begin with search engines, apps, ai tools, and the like. That means congress must design two complementary experiences intuitive journeys for people and intuitive journeys for people in trusted, structured data for the digital systems they increasingly rely upon.
▶ 0:16:28Both should lead people back to our authoritative congressional information. The quality, structure and governance of our data matters more than ever. That is why modernization efforts, such as the adoption of structured legislative data standards like united states legislative markup, are so important. While these are largely invisible to us and most of our users, they are foundational. Better standards produce better data, better data produces better systems, better systems produce better public access.
▶ 0:16:57Trustworthy outputs require trustworthy inputs. As congress considers the future of public access, we should recognize that we are no longer simply publishing static websites. We are publishing information into a digital ecosystem. The early vision behind gpo access and thomas evolved into today's govinfo and congress.gov repositories, demonstrating the value of sustained investment, stewardship and collaboration across the legislative branch.
▶ 0:17:29We do not need to predict every future technology, and we shouldn't spend years on the perfect plan. But we do need to continue to invest in well structured data in a clear understanding of user needs and in the institutional capacity to adapt quickly. Like you said, my written statement outlines three recommendations.
▶ 0:17:46First, www.house.gov should evolve into a clearer public front door, helping people find the authoritative information without needing to understand the internal organizations of the house or even of the legislative branch. Second, the house should create a more consistent experience across all of our websites, using familiar navigation and design patterns that help people find what they need when they need it.
▶ 0:18:14Third, congress should adopt a data first strategy grounded in user research and designed around how people and the systems that they rely upon discover, access, verify, and use our information. These recommendations are not intended to be implemented through a single large modernization effort.
▶ 0:18:33They can begin with user research and discovery, continue through forward looking information architecture and user experience design and progress through a phased, iterative roadmap that continuously delivers improvement and value over time. Technology will continue to change. Congress's responsibility to the public will not.
▶ 0:18:54Our responsibility is to ensure that wherever someone begins their journey, whether that's on house.gov, congress.gov, one of your websites, an ai tool or technologies yet to come, they can find the information they need when they need it, and it's trustworthy and authoritative. Thank you for the opportunity, and I look forward to our discussion. Speaker 1: thank you, miss gullickson, for that insightful information. And now I recognize Mr.
▶ 0:19:22Rutledge for five minutes for an opening statement. Speaker 5: uh, thank you, chairwoman, vice ranking member torres, members of the subcommittee, thank you for welcoming the perspective of the library of congress at today's hearing. On behalf of everyone in the libraries office of the chief information officer, thank you for also recognizing and investing in library technology so we can better serve the united states congress and the american people.
▶ 0:19:44The library's two most popular websites, loc.gov and congress.gov, see nearly half a billion page views each year, not counting ai crawlers and bots for the library. User experience or ux is all about those half a billion pair of human eyes locked in on the way. We present historical collections and legislative data across these websites. Good user research helps us understand the people behind these eyes, what they need, what they expect, and how they feel about what they see.
▶ 0:20:13This is how the library approaches ux design. We learn why users interacted with our data. These users are members of congress, quickly checking a bill status in their phone, on their phone, in the cloakroom. They're also high school teachers, like my brother, who uses congress.gov with his students to learn more about the constitution and the legislative process. We have a responsibility to ensure those users and everyone in between get what they need from our websites.
▶ 0:20:40Ai has changed the technology landscape faster than any of us have expected. On one hand, our systems are being stressed in new ways. In a single 24 hour window. Last month, our human verification check on congress.gov blocked approximately 98% of incoming requests as non-human traffic. However, ai is also giving the american people access to more information than ever before.
▶ 0:21:06But authenticity and objectivity matter, and people know they can depend on congress.gov to meet those standards. I'd like to outline lessons learned, lessons we have learned from the past, and how we are engaging in thorough and thoughtful user research in the present, so we can further democratize access to information and the legislative process in the future. In the mid 1990s, congress tasked the library with getting legislative data online.
▶ 0:21:30We achieved this mission through two websites, thomas for the public and the legislative information system, or lis for the congress. Ultimately, two sites with different designs and approaches led to a digital disconnect and confusion between the members of congress and their constituents.
▶ 0:21:48With congressional support, we built a new system, congress.gov, which launched to the public in september of 2012 with an updated design, modern infrastructure, mobile friendly access and new search capabilities. But we didn't stop there. New features, including committee schedules and videos, publicly available crs reports, bulk data download and an api are now drawing more users to the site than ever before.
▶ 0:22:15Today, congress.gov is more than the sum of thomas and lis. It's the viewfinder that brings into focus multiple modern data flows from across the legislative branch, alongside library resources from crs and the law library of congress. With this. With this expansion, we're complete. Now is the time to turn our focus to making congress.gov as intuitive and user centered as possible.
▶ 0:22:39The library follows established, user centered design principles and a repeatable five phase process for conducting user research. First, we discover this involves researching and gathering everything we know from users, stakeholders, and data. Next, we define where we need users to inform requirements and constraints. Then we ideate. This leads us to the prototype stage where we get to visualize what we what the product will look like, and then finally, we validate with.
▶ 0:23:09User testing will often return to the idea stage to improve and advance our ideas. Based on what we heard in testing over the last year, we completed the discover phase by conducting one on one interviews sessions with public participants. This research showed us once again that congress.gov is a valued and trusted resource. We are now in the final stages of the defined phase of our ux process. Our priorities, as derived from the discover phase, are updating the design and enhancing congress.gov search.
▶ 0:23:40We plan to have a prototype of the updated congress.gov interface to validate with congressional users, no later than the end of this calendar year. Additionally, two items within the library's fiscal 2027 budget request the ai enterprise platform and the web application delivery request will directly support long term goals for congress.gov.
▶ 0:23:59The ai enterprise platform will enable experimentation with and implementation of ai based tools and enhancements across the library, including congress.gov. One idea we are actively exploring is how the popular ask a librarian feature on loc.gov, coupled with ai tools, could further enhance the congress.gov search experience.
▶ 0:24:21In closing, as we move through the user design process, we welcome the continued partnership with this subcommittee. Speaker 1: thank you so much, Mr. rutledge. And finally, I recognize miss lee for five minutes for an opening statement. Chairwoman bice, ranking. Speaker 6: member torres, and members of the subcommittee, thank you for the opportunity to appear today.
▶ 0:24:47I'm grateful for the privilege of representing my brilliant peers who have shaped these ideas and the future generation who will who are digitally native and will never walk these halls. I grew up a few miles from here, translating government, its forms, offices and systems for my father, a baptist preacher who spoke no english, and for the community center for people experiencing homelessness in the shadows of this dome.
▶ 0:25:13It is where I first asked where power comes from and why some have it while others do not. A lifetime of sitting with that question. My belief holds that in america, power comes from engaged people in this country. Any informed person with an idea and the support of like minded peers can influence change for their community and country.
▶ 0:25:36So it falls to those who serve the public to keep access to information equal and evolving with technology, so that the social contract stays strong. So I offer this subcommittee one key takeaway. The challenge is how the house meets people where they are by demonstrating representation, by showing, speaking and listening to the public. Democracy requires dialog, and dialog strengthens trust.
▶ 0:26:05In 2018, I was invited to help make it easier to understand how the legislative drafting process. Easier to understand. My expertise is in designing systems that are intuitive and solve the needs of users. There I learned that a bill is never just a bill.
▶ 0:26:21It is a complex negotiation of intent and outcome that work by amending layers of existing law for an individual to track these changes across thousands of pages of code is nearly impossible without specialized help. That is the difference between availability and access.
▶ 0:26:43The tools that interpret data into useful information often sit behind paywalls, so staffers, lobbying firms and everyday constituents hold unequal levels of access. The same public data. The rise of ai changes the playing field. Ai generated answers now reach a billion people a month. Search. How does this bill I heard about on a podcast affect my benefits?
▶ 0:27:08An ai generates answers from the the blogs, the press and trade association websites. It would be far better if the answers came from the authoritative sources from inside the house. Even inside the house where information is available, we haven't built the intuitive layers that help members and staff flatten the learning curve to become productive immediately. Because nothing has been designed around the natural churn of the house every two years.
▶ 0:27:36Both examples illustrate where technology designed around human centered measures can be used to close the gap in representation. Fortunately, the house has demonstrated that it can build technology around the needs of members and staff with a comparative print suite. I envision the next phase of the digital people's house feeling like a visit to the library of congress, or really any library.
▶ 0:28:02Engaging with congress online should feel the way a librarian remembers the book the children loved, and also helps the investigative reporter personalize enough to understand why I am here and to send me off knowing what comes next. We have the tools and expertise today to design that secure, personalized experience on top of ongoing digitization of congressional data.
▶ 0:28:25To do this, the north star should be to build an engaging place that shows context, speaks the truth, and is designed to listen, show, speak and listen to show design for their user and their journey. Stop treating the digital house as a collection of static pages organized by each office, and map it to the actual needs of the people it serves an educator, a journalist, and a constituent are on entirely different missions.
▶ 0:28:51The digital infrastructure should guide each along their path with clarity and context at every step. To speak. Be the primary, trusted source of truth. Wherever dialog is happening, the house is authoritative information. The law, records and statuses should be structured and accessible enough to be cited accurately wherever people already discuss policy.
▶ 0:29:14In an era when ai assistance is sure to turn into ai agents, silence is not neutral. Invest in modern, well documented apis and clear, neutral explanation layers so that trusted data surfaces truth reliably in the public square. To listen, commit to feedback loops. No institution can serve a public.
▶ 0:29:38It has difficulty listening to build structured, bounded channels to acknowledge, process, and respond to the signals coming from the people and users. Meaning light touch, honest communication that manages expectations that build trust for a healthy democracy to run on. I leave you with the question I most want us to answer together. How might congress use technology to assert its place as the authoritative source of truth and representation?
▶ 0:30:06The answer ensures that a teacher in portland, oregon, a veteran in portland, maine, and people from every district have access to the information that encourage engagement and that the house continues to lead, not follow in using information and technology to strengthen the institution and the processes over which it presides. Thank you. I look forward to your questions and the work ahead. Speaker 1: thank you, miss lee, for those opening statements.
▶ 0:30:34And thank you to all of our witnesses again for being here. We'll now move to questions for the witnesses, beginning with myself and followed by ranking member torres. And if our additional committee members can be present as well. Without objection, the five minute rule is waived. As you can see in this particular subcommittee, I think it is much more effective to have dialog. Um, and that is why we are in a roundtable format and setting rather than at the dais.
▶ 0:30:59Um, but being that being said, while the timer will be turned off, I do ask that you please keep your responses brief. Any member wishing to be recognized should signal the request to the chair. And I now recognize myself for the purpose of questioning the witnesses.
▶ 0:31:14Um, to miss gullickson, uh, today and at our recent hearing on ai, how crs is using ai, we heard about staff who were using ai to turn great ideas into effective tools that help them do their jobs better.
▶ 0:31:30In your testimony, you noted that the challenge is not a lack of ideas, but figuring out effective pathways for those ideas to be evaluated and turned into actual tools and systems that can be supported over time. And I think for me, one of the things I think about is doing all of this in a timely way is also incredibly important, uh, given the rate of change. So what is the biggest barrier to creating these pathways? And does it come down to capacity and funding?
▶ 0:32:01Speaker 4: yes. Speaker 1: yes. Speaker 4: that's my short answer. Yes, we have a lot of ideas, and I do think we need to create more ways to collect those ideas and then decide who's going to evaluate them. My idea is to create a repeatable innovation cycle.
▶ 0:32:16So one that we evaluate those ideas, we test them quickly, we route them to the right tech team, not only the right tech team here in the house, but if we have a tech team across the ledge branch, either at the library or at gpo, maybe we need a partner with the senate. So route them to the right tech team, invest in what works and scale. Like you said quickly, I think the biggest barrier from where I'm standing is who's going to who's going to do it, who's collecting those ideas and how we're going to do it.
▶ 0:32:48As you know, the cao has put up that great vendor site, so that's great. But how do we get the we have so many staff who have great ideas. How do we get them to do it here? Um, I don't want to I'm not advocating for more bureaucracy, like you said. Thank you. And then I do want to say that you're absolutely right. Capacity, velocity and funding also matter. Um, one of the challenges we have is the congressional calendar, the needs that you all need as members.
▶ 0:33:17And your requests don't always match with the fiscal calendar. So, um, the funding and know your funding have been really flexible in some of the projects I've been, um, involved in. So I continue to, to find where that, that right balance is perfect. Speaker 1: this is a very bipartisan committee and as such, we are just going to go back and forth with questions.
▶ 0:33:45So I recognize ranking member torres for a follow up question or another question. Speaker 3: yes. Thank you. Our constituents are mutually frustrated as we are. Um, miss gullickson and Mr. um, rutledge, when when my constituents call my office. Um, it's after they are completely frustrated. Either non-responsive agency and they're trying to find information.
▶ 0:34:14Um, they don't have a bill number. Um, if they're looking for information on a bill, they, they're simply heard and can't remember, you know, everything that they heard that they have an idea of what it is. Um, it simply, what they want to know is how do I find information in your website? Um, what is congress actually doing to improve my quality of life?
▶ 0:34:41Uh, right now, the honest answer is that it takes real digging. You almost need a phd in how to find information on, you know, congress, you know, dot gov. Um, so we need to be more transparent and accountable to them.
▶ 0:35:00Are there any opportunities that you see to develop features on congress.gov that would track members voting record, uh, amendment or appropriation tracking as easily as someone could check a flight status. I mean, nowadays I could text my flight number to myself and get updates on that status of that flight.
▶ 0:35:30How is it that they can do that? And we can't do that? Speaker 5: uh, I mean, from, from the library's perspective on congress.gov, we can put the data up as quick as we can get it. Um, and right now we get data updates every 15 minutes. Um, we can certainly increase the frequency. We're just reliant on, on our data partners to provide that information, uh, to us. So if the data is available, we can get it up.
▶ 0:35:56Speaker 3: so explain to me how that works when oftentimes we're taking, let's say tonight we have 830 votes, but sometimes we have after midnight votes. Does that mean that your folks are keeping up with. Oh my goodness. Speaker 4: absolutely, ma'am. So the data flow goes from the clerk's office. Um, our floor staff. So the floor staff you interact with on the floor, they are doing real time data entry. Electronic voting system is involved. That data is going into machine readable formats.
▶ 0:36:28What sometimes what we call an api and it sends it over to congress.gov, like john said, every 15 minutes. And they're being able to, um, then make it available. I think one of the challenges is then how is that user experience? We have the data, then how do we find it? Speaker 3: who translates it? Speaker 4: we don't need to translate it. The machine does it. Yeah. We have a great team of engineers that make it happen.
▶ 0:36:54Speaker 3: and are there plans then to have the interactive calendar, um, that could be integrated within each individual member, um house that gov website. Speaker 4: yes. Um digital, the house digital services is working on house gal and we continue to work with them to where the clerk's office is feeding some of that, some of that data to them, but continuing to work with them, how we can service that on member websites and on congress.gov and on the clerk's website.
▶ 0:37:22Speaker 3: and lastly, just when do you expect that to happen? Speaker 4: can I get back to you on that? Because I don't know the time frame for implementation. Speaker 2: thank you, thank you. Speaker 1: um, I recognize myself. I think there's a couple of pieces of that that I may want to extrapolate on. And that is, um, it seems as though there are a lot of hands in the pie, if you will. Um, is there a way to streamline that so that the functionality is a little bit smoother and easier?
▶ 0:37:53Uh, is there either technology or ways that that can be improved upon that you could see? Speaker 4: I think there's a, there's a lot of purpose behind how our data flows, right? Because we're bicameral. So we need to make sure that the senate state is over on congress.gov to tell the whole story. Um, we are in the clerk's office. We are working on improving our data transfers to the library.
▶ 0:38:19And we have ongoing projects to continually update and modernize everything. Um, I do think that, um, we have good foundations to do this work and we just need to continue to invest. And I certainly, we have some great ideas to optimize our, what we're doing already. Um, so not anything immediate, but continuing to, to continue to the work that we're already doing and the foundational work. Speaker 2: thank you.
▶ 0:38:48Speaker 1: I know that, um, next major redesign of congress.gov begins. Um, it'll be important to not only get input, public input member and legislative staff input. Um, but one of the things I worry about is time. Um, so, so two part question. When was the last time that congress.gov or house.gov has been, has had a complete refresh? Is 2012 the last time that we've actually seen something sort of new and innovative?
▶ 0:39:17And then to follow up with that, if we're going to do a redesign of these sites, what's a timeline that you think is realistic to actually accomplish that? Speaker 5: so we've actually already started the redesign of congress.gov. We actually began it last year. We held, um, a couple of rounds of interviews, uh, one starting in november of 2025, um, with 15 congressional and public, um, users and again in march of 2026, again with 15 public and congressional users.
▶ 0:39:48Speaker 1: when you say public and congressional users, can you extrapolate who that is? Speaker 5: um, different demographics on the, on the public side? So educators, for example, researchers, for example, congressional staff, different types of congressional users. Um, we can certainly provide the demographics if you're interested. Speaker 1: I was just curious, like who you're actually reaching out to to get insight. Speaker 5: we typically get users on the public side through congress.gov public forum.
▶ 0:40:13Um, after today's hearing, my brother's very interested, by the way, and being one of those people, um, which I'm certainly want to help him connect with the team because I didn't know that before this hearing happened. Um, but through those two rounds of sessions, by the way, your opening statement hit the nail on the head in terms of their feedback, right? The search is, is complicated in terms of the results that come back. The, the sites, the pages are cluttered, right?
▶ 0:40:39It's difficult to find the data, especially on broad search terms. Speaker 1: I mean, as a member of congress, when I'm trying to search to see if I if I have actually cosigned a bill, it's incredibly complicated. Yeah. And that shouldn't be the case, right? Speaker 5: yeah. And if you don't understand the legislative process, that's also somewhat complicated because the site isn't geared towards more towards a member of congress or a staff member versus the public. It's not geared towards constituents.
▶ 0:41:03Um, and the site has evolved in terms of its user interface ever since it was launched in 2012, but it's just been adding more content, right? It's been very content rich. Uh, in fact, we've grown congress.gov from about a million items to 2 million items over that period of time. Speaker 1: but the interface hasn't changed. Speaker 5: the interface has just become more cluttered, if you will, not in terms of modernizing the look and feel.
▶ 0:41:31We started that effort last year and we're continuing it now. Uh, as I talked about in my opening statement, we're in the second phase of a five phase process. We're going to start the id phase here in just a couple of weeks. Um, and we plan to have, um, the validation phase done, which means more engagement feedback, um, actual, um, prototype where you can click and actually experience it and get feedback completed by the end of this year. Right?
▶ 0:41:57Which means we want to have a completely redesigned congress.gov up next year. So completely in production, whole new site experience. We want it to be intuitive, we want it to be engaging, and we want to offer a personalized experience next year. That's our target. And honestly, I'm hoping that by the congress.gov public forum in september, we have something to show then. Speaker 1: so my fingers are crossed crossed. Yeah.
▶ 0:42:22Well, I think that's really exciting news given that you're exactly right. You can see that there's been a lot of content added to the site, but it hasn't been structured in a way that's easy to actually navigate for your average user, including your member of congress. And so, um, I appreciate the update on that.
▶ 0:42:40And my only, I think concern is that technology is moving at such a rapid pace that if we take too long to actually get something stood up, that we may be falling behind already, um, because innovation will be lapping us. So it's nice to hear that we're trying to do this in a very expedient way. Uh, at this time, I'll recognize the ranking member for an additional question.
▶ 0:43:01Speaker 3: I think that that is a system right now where we are innovating moving forward, but it takes so long to adopt policies that by then, by the time we do it, it's already a whole new way of doing things. Um, members of congress are very creative in legislating and how they go about legislating, and some are very successful in that, whether it's through an amendment or they introduce a, a bill that gets folded in a much larger,
▶ 0:43:33Um, package, for example, moving through the appropriation process, a constituent following a bill that I introduced that gets added to, um, to that in that process, they lose total information about that bill and they will not be able to get any updates on that, on that bill. So that thread really ends when that bill is transferred.
▶ 0:44:01Um, is there a way to make it easier to go on congress.gov and follow a bill all the way from inception to even when it gets added into a must pass piece of legislation, how do you continue to follow it? Speaker 5: yeah, I think that's one of the primary goals in terms of updating the bill tracker. That is part of the feedback that the team received through those two rounds of of forums that we that we met with.
▶ 0:44:31Um, the bill tracking process is, is complicated. In fact, it seemed like it was confusing to some users, um, in terms of where a bill was in its particular process. So that is one of the key goals we want to achieve. Speaker 3: as well as if you vote yes in committee, but no on the floor because suddenly it changed in um the rules committee. Speaker 2: um.
▶ 0:44:54Speaker 3: then your vote, the bill might show that you only voted, you know, yes or no and doesn't really give any feedback. Yeah. Speaker 6: actually. Can I add some context to that? So around the same time we were doing all the comparative print I was working with, um, was it demand progress? Um, to actually do bill tracker, we called it bill map. Um, and I think it was handed over to the library of congress. Um, and it was somewhat integrated, but I do have existing research that does it.
▶ 0:45:23And really it seems like a behemoth of a problem. But really, if you think about it, all you have to really decide is, is this person entering with information? Like, what is the person entering preexisting? Do I know my member? Do I know my issue? Right? Like there's different ways to categorize what someone might know or want when you first get there. And then talking about the temporal context of it is very, very hard.
▶ 0:45:47And that's the part I think, that we can collaborate more on, which is that there isn't an ability to see what is current and what is related to it. But I think that there has been, I want to say on that there is evidence of that work, and I would love to share that so that we can expedite the process.
▶ 0:46:07Speaker 4: and I can back that up, that we definitely have had meetings with the staff of the subcommittee and the library on, um, specifically recommendation 120, which is to show your work better. And I know that, um, we have a commitment to get some of those features out so we can talk about them at the congress.gov forum. So you can go to your member profile page and see where a bill might have been incorporated in a larger bill. And that's the bill that you sponsored.
▶ 0:46:36So we've been actively working on that, um, this year, which is, um, very much related to what charlotte has said and so excited about that work. The other thing that I'm going to squeeze in here, because we're having a discussion is one of the things we can take a look at is, um, semantic search. And how do we use ai to provide a semantic search box across all of our websites?
▶ 0:47:00Um, so that it doesn't, it understands not just what the words I type, but what I mean. So what charlotte's saying, and so I'd really love to follow up with the committee and john about to ask a librarian because it to for me, it should be expanded to include your websites and to gpo and some of our other large branch repositories in such a way that it then can bring back the citations so it so it can brought you back to maybe a press release or route you back
▶ 0:47:31To a committee report on, um, gpo or back to a resource just on congress.gov itself. Speaker 2: thank you. Speaker 1: I recognize myself and this is a perfect segue for the next question to charlotte, miss lee. Um, how can we leverage large language models and ai to personalize legislative information while at the same time ensuring accuracy and trust? We've had lots of conversations, um, in this subcommittee about the utilization of ai.
▶ 0:48:01And the phrase that jumps out at me is garbage in, garbage out. When you're looking at these large language models, if you are feeding it full of, um, maybe not necessarily misinformation, but not completely accurate information, then what you're getting out of it is not going to be accurate.
▶ 0:48:18So, you know, I always think about the fact that we live in a day and age where it is, um, we have more access to information, particularly legislative information, than we've ever had in any time in this country's history. But people are more distrusting of what we're seeing. And so how do we use these large language models and use ai, but also build accuracy and trust in the information that's being presented? Speaker 6: I love this question. Thank you. You're welcome.
▶ 0:48:48I think that one of the first things I learned when I was working with leche council was the definition of accuracy means two different things to everybody. Accuracy in the like exact word I wanted and accuracy in the outcome and intent that it has. Right? So already accuracy is is a, is a challenge to define. Actually when you say something is not accurate.
▶ 0:49:13Um, the other thing is that there's a stanford study talking about like leading commercial legal ai, like the really like top tier ones. Um, it's still hallucinated 17 to 33% of queries. Even when it was given the data, the clean data. So I know that it's easy to say it's garbage in, garbage out, but it's not really the nature of generation creates discrepancy.
▶ 0:49:39And so I don't know, um, if there's ever going to be a 100% accuracy, which is not possible. I don't think that we'll ever reach that. However, I think that with human centered design, you can design around the critical points of human judgment and enforce those guardrails very easily. So I can give you the example. When we were doing comparative print, um, not all of the changes could be routed, routed to a current law or a statute. Right.
▶ 0:50:12And so we did these user notifications where they had to manually go and check, like, these are the five things that did not compute. Like what are these things? Please resolve. Right? And just like that, if you're establishing something and saying, well, these things you can't like, you can't move on, you cannot move on this journey until you have confirmed a, b, c, d and we can absolutely, in user experience, establish those boundaries.
▶ 0:50:36So, um, yeah, points of human judgment, like what is the thing that they are going to be held accountable? So, um, any decisions, any facts, right? Like who like that is the big difference between large language model and human is the accountability, right? So one example of a point of judgment is what are the things that you are going to be accountable for? Is it this fact? Is it this line? And then putting that onus back onto the user is important.
▶ 0:51:06Speaker 1: do you think that we are in a, in a day and age where, um, you can sort of utilize an llm to be able to give you some sort of context about, let's say a bill, um, and, and then have, you're still going to have a review of that. How long do you think it's going to be before we actually don't have that any longer? Or do you think that's ever going to be the case? Speaker 6: well, it depends on the engineer using. So sometimes when I are we allowed to say names of like actual softwares? Sure. Okay.
▶ 0:51:35Like, so some of the fully generative things like chatgpt will just kind of aggregate and like spit something out. But some of the other things I love, like perplexity has a citation and a like, like tags on every single thing. So you can go check that citation. Right? So I think that that's a standard that they set. So I mean, it's all new for everybody. But if that is the standard that the house says, right?
▶ 0:52:01Like every line has to have a source or something that traces back to there's no reason technically we can't do that. Speaker 1: so it's sort of like writing your high school english paper. Speaker 6: yeah. But like. Speaker 1: you have to cite everything. Speaker 6: yeah. Speaker 1: fantastic. Um, thank you for that response, miss lee. And I'll yield to miss torres for an additional question.
▶ 0:52:20Speaker 3: so with the sheer, um, information, volume of data that, um, is now available on congress.gov, I'm curious, um, what that tells us about what the public is actually wanting to know. Um, what is the most viewed or most searched content on that website?
▶ 0:52:45And does that usage data shape how you prioritize new features? Speaker 5: yeah. I mean, the obvious clear most popular content are bills. They are looking to find particular bills. But what's interesting is they are not going to congress.gov and doing the search on congress.gov.
▶ 0:53:0795% of the public actually comes to congress.gov from an other outside source, whether it's google or social media link or an email from a friend. Um, 95% of the public comes from somewhere else. Um, what's interesting, more than 70% of congress comes from an outside source though, too, right? Speaker 3: so can I just say it's easier for me to go on google and search? Yeah. Um, for the bill.
▶ 0:53:33And then it takes me to your website and to go to your website and search, because then it gives me a bunch of other stuff. Speaker 5: we recognize that and that's why we're going through a redesign. So, you know, that's, that's something that we, um, have been wanting to do for a while. We're excited to be going through this redesign process and, and working with, um, the house, the senate, and this committee to be part of that redesign, by the way. Speaker 2: so, so.
▶ 0:53:57Speaker 3: um, can you speak to what criteria should congress use to decide whether to adopt ai generated bill summaries, which I think is really critical. We have to get that right. Um, real time bill trackers, topic based notifications and personalized dashboards. If I'm only interested in this subject, can I sign up to get alerts just on this subject as an example? Speaker 2: yeah.
▶ 0:54:30Speaker 4: we one of the structural things that we need to invest in is that structured data. So we have projects already going on. We have a structured data format called united states legislative markup. Gpo is leading that project. We have a group of, uh, clerk employees, secretary of the senate employees, library of congress employees who participate weekly in those calls. We have a roadmap continuing to invest in that is going to fuel congress.gov.
▶ 0:54:56We also have two other projects, ma'am, the clerk of the house. We are my office is modernizing our legislative information management system. So that's the data that is feeding congress.gov and the senate of the secretary. I said that backwards. The senate secretary is also just starting modernizing their system.
▶ 0:55:17So those foundational building blocks to get that data over there so that we can make those decisions on what is the most important to surface in that user centered design that john is talking about is just continuing to, to invest in making sure we have the right staff and, um, being able to increase our velocity on doing the work because, um, as you know, we have competing priorities. So sometimes our work isn't as moving as fast as all of us want it to move. Speaker 2: thank you.
▶ 0:55:48Speaker 1: thank you for that. I recognize myself and I'm going to follow up on that question. You have a lot of different competing interests and entities that you're working with. Uh, you mentioned the senate. Is that does that hinder your ability to actually move faster because you are working with what is essentially two different, um, customers, if you will, and having to try to figure out how to incorporate what you're trying to do into in two separate systems. Speaker 4: no, not, not necessarily. No.
▶ 0:56:17The legislative process is very complex. The house rules are very complex. So I think sometimes just making sure that we are making sure the systems can respond the way they need to because of the complexity of the process, sometimes gets us slowed down. It's not, as I always say, we're not building a shopping website, right? If we were building a shopping website, we might be done. We're not. Speaker 1: people are shopping for data, right? Speaker 4: they're shopping for data that's already in.
▶ 0:56:45As you know, when we take a an amendment and it gets into nda or a bill into nda, then to tell that story is really difficult. Absolutely. We can do it, though. Speaker 1: uh, Mr. rutledge, I want to pivot to you. You mentioned that there are now 2 billion visits to congress.gov. Did I did I get that number right? Speaker 5: 2 billion visits?
▶ 0:57:10Um, much more than two, uh, five, half a half a billion visits to congress.gov just on the human side. Speaker 1: okay. And so my follow up question was actually going to be how many, um, sort of, uh, what machine traffic visits are happening on congress.gov currently?
▶ 0:57:30Speaker 5: yeah, that's, that's a great question because we were seeing in april, we got 16 million visits from a bot traffic in may, we got 20 million visits in one day in the middle of june, 16th of june, to be exact. We got 22,000,000 in 1 day. So it actually caused our infrastructure to come to a little bit of a grinding halt. And you might have experienced that. So we had to make a couple of emergency changes to address that on congress.gov.
▶ 0:58:01So, um, one of the things we had to do was fully separate out the public instance from the congressional instance because it caused the data ingest issue. Uh, the other thing is we put that human verification check on the search for, uh, the site, which is where we've discovered 98% of that traffic was non-human traffic. So we had, um, and it was not malicious in any way, shape or form, just vibe coders, recreational coders playing with ai tools.
▶ 0:58:27We think learning about congress.gov, it's a great site. Uh, just wanting to learn more about the data. Just hit the site at all at the same time. Speaker 1: so you've seen an exponential increase in machine traffic on congress.gov in the last couple of years. Speaker 5: absolutely. And it's still going up. Speaker 2: so perfect. I will yield to miss torres if she has additional questions. Speaker 3: um, just briefly. Speaker 2: um.
▶ 0:58:58Speaker 3: your testimony mentions a couple of items in your fiscal year 27 budget request as an appropriator. Um, the ai enterprise platform, which would enable experimentation with an implementation of ai based tools to systems across the library and the web application delivery and management, um, improvements to enable cloud based services. What's the found?
▶ 0:59:27The funding request for both of these. Speaker 5: I'll take the second one first to follow up to Mrs. wise's question. The um web application delivery is to move congress.gov and llc.gov to the cloud. Um, that exponential increase in ai traffic really impacted our infrastructure. Um, and we were not able to accommodate because it's physical hardware on an on premise data center. Speaker 1: it's locally hosted. Speaker 5: it's locally hosted. That's exactly right.
▶ 0:59:56Um, and we need to move it to the cloud so it's more dynamic. We can add resources more dynamically. But the other challenge that we're having on the infrastructure side for locally hosted is the cost of hardware is literally going up five times. So one server, um, a year, a year ago cost us 55, 000. We got a quote for 254 000 for the exact same server. And that's the problem that we're seeing. Uh, and we have to keep the hardware up to date. We have to refresh it.
▶ 1:00:25And it's a challenge. Speaker 3: why is that? Is that the lack of chips? Uh, tariffs. What, what is that? Speaker 5: well, honestly, it's ai. I mean, a lot of the vendors are building data centers. Speaker 3: to the demand. Speaker 5: is so high. I mean, go to ashburn and you'll see the data centers popping up like mushrooms. I used to live there. So I know, um, and they're, they're getting preferential treatment by the, by the vendors. So they're buying out all the inventory.
▶ 1:00:51And government agencies like the library are sort of having to wait and the cost is going up. Uh, but moving to the cloud is really a priority for us. And it's going to help us reduce our overall cost, but also help us from a performance perspective, right? Not only on congress.gov, which is important, but also.gov as well. Um, real quick aside, since we're not watching the clock, I learned just the other day that congress.gov is in the top 4% for government attacked websites.
▶ 1:01:18Llc.gov, interestingly, is in the top 6%, which I would not have expected. So we get a lot of attention from nation state actors and our particular. Speaker 3: library of congress. Speaker 5: yeah. So we are a popular agency apparently. Um, but then on the ai one, uh, so with the, the web application, we're asking for $2.4 million for that to move it to the cloud. Speaker 2: 2.4, 2.4. Speaker 5: 2.475 to be exact.
▶ 1:01:50Um, on the, on the ai enterprise platform, the library's total ask is 5.4, but ocio is ask is 3.3. And that's broken down really into two buckets. One is 1.6 for contract support, uh, for maintaining the platform itself. Uh, and then 1.7 is for the cloud infrastructure to handle all the compute resources that we need for the environment.
▶ 1:02:18And it's a, it's an investment up front, but it's got long term return on investment for us and gives us a more secure environment that we can control, especially with our data, the speech or debate data, the rights, restricted information, and building our own language models to support you in congress and other library needs. Speaker 2: okay. So thank you. Speaker 1: uh, and for my final question this afternoon. Um, this is to you, miss lee.
▶ 1:02:42Your testimony states that a website waits to be visited while a voice travels. Can you expand upon that and describe how that principle could be applied to house.gov, the house.gov website? In other words, what would giving the house.gov website a voice actually look like?
▶ 1:03:01Speaker 6: I think we've extended covered the notion of generative ai being the main search engine, but it looks like, um, the fact is that pew tracked 29,000 or 69,000 google searches, and then when the summary appeared, people went to a website 8% of the time versus 15% without one. So I think away from that, I think that we can turn content into conversation based on where it is happening.
▶ 1:03:30So actually, the simplest thing I can say is that when you post on reddit or in facebook, you can respond like right away, like you can activate things and invest in things that have that dialog.
▶ 1:03:46When they say this is happening, you can actually appear and be dynamic in social media platforms or discord, or like create apis or encourage the development of apis that constantly, like are embedded in the way that constituents really talk about things. Like, I can always appreciate the role of congress.gov and house.gov, like as a website, but I feel that there is a digital voice that is lacking.
▶ 1:04:17So I think the united kingdom just set up a communications office to to be that voice. I don't think we have one that are similar, that are responsible for, I guess, outbound communication. Right. Everything is very reactionary. Like we said, like somebody has to ask for it for us to give it to them just like on the website. Right? I have to ask for it. I can't discover anything. Right.
▶ 1:04:44So I think that what I say voice, I loosely mean that the information has to be encouraged and actively participating in dialog where it is happening. Um, I also think that, um, the reason that you might be getting more visits when you have, uh, ai answer your questions, um, is actually because it prefers structured websites.
▶ 1:05:10So government websites are well established enough that it can actually be the number one search result, but we haven't intentionally done it. And I think that, um, the change of the way that people are engaging online has never been regarded for a digital democracy like conversation, right? Like where are people?
▶ 1:05:37Um, and actually can remember, one of the strains that I learned at ledge council was this. And this is actually the root of the question. And perhaps the thesis itself is that, um, when I was working with ledge council, I was working with wade at the time, and he said that they're in an existential debate and, um, concern all the time negotiating the difference between theory and practice of something, meaning that bills are now being used. So bills were written to be permanent.
▶ 1:06:07Every member or staffer who submits a bill assumes it could be permanent. So ledge council takes that very seriously from what I have learned. Right. And so they have to be very accurate because it could become permanent. And when that happens and members. But in in that level of scrutiny, right? Like the organization is like, we're taking this very, very seriously.
▶ 1:06:30But really the intent of that bill was to have dialog because they needed to go on a news channel and say, I did this thing or be able to point to that, that thing. I think that what I'm saying is when we lose public forum, like there's not people like outside on a soapbox anymore, right? We've lost that actual public forum. Speaker 1: I think if you talk to my social media, there'd be a lot of people that are on their soapbox, but that may be a different conversation. Speaker 6: the digital soapbox, right? So this absolutely.
▶ 1:06:57I think that there and that is what I mean. I mean, that we have lost this ability to have any kind of conversation. And the house hasn't acknowledged at all. The communication portion of dialog. Speaker 1: would you agree, though, that part of this is being able to have a dialog in a respectful and thoughtful manner? Because I think people have asked me before, how could you get more bipartisanship? How could you get more dialog? And I said, my response to them is, turn the internet off.
▶ 1:07:26Um, and although that's not a reality, nor does that solve the issue, I think making sure that we recognize that we need dialog, we need to be able to talk through these issues and come to some sort of agreement that, you know, we may not we may not agree on the issue, we may disagree vehemently, but but we can move forward in a different direction.
▶ 1:07:46Um, and I think that's the hard part about, um, about the online conversation is in some cases it is, um, obfuscated because the person doesn't want to be publicly known. They use fake usernames or pictures. They're not, they're not really wanting to have a dialog. They're sort of wanting to have an argument rather than being able to sit down and, and hash out real issues.
▶ 1:08:09Speaker 6: I feel like we could actually sit over a glass of wine and really actually think about the notion of dialog in democracy at its length. Speaker 1: I think miss torres and I are a four year deal. Speaker 6: um, and, um, but actually dialog, one thing I will say is dialog isn't always a conversation. It is sometimes a confirmation or an acknowledgment. It doesn't have to be like back and forth, back and forth. Right?
▶ 1:08:38But right now, we're not actually even doing the bare minimum of acknowledging or place wayfinding. So, I mean, before we can get into a conversation or two way, as we say, let's think about the one to many that that we can fix now. Speaker 2: perfect. Speaker 1: thank you, miss lee, for that. And, um, with no additional questions, the members of the subcommittee may have additional questions for you. And we ask you to respond to those questions in writing.
▶ 1:09:08Uh, without objection, each member will have five legislative days to insert additional materials into the record or to revise and extend their remarks. I want to say thank you to all of our witnesses today for the robust dialog back and forth. Um, and your insight into this very important topic. We appreciate you being here this afternoon. If there's no further business, I thank the members for their participation. And without objection, this subcommittee stands adjourned.