ICYMI: Empowering Legislative Staff with AI on Capitol Hill

POPVOX Foundation and R Street Institute recently hosted an introductory discussion on the integration of Artificial Intelligence (AI) into the workflows of Congressional legislative staff. The event included an overview of current institutional guidance and featured insights from prior Hill staff and policy experts. Here’s what you need to know if you missed this enlightening discussion.

AI as a Congressional Capacity Builder

Aubrey Wilson, Director of Government Innovation at POPVOX Foundation, set the stage that AI technology may not be as new as it seems, with its long-standing presence as text prediction on phones to Netflix recommendations. However, generative AI (GenAI) is new, and this emerging technology is capable of generating new text from a diverse and overwhelmingly large collection of sources. OpenAI’s ChatGPT is one of the most well known examples of GenAI, and as the only currently authorized private large language model (LLM) approved for Congressional use. It provides comms teams with a valuable new tool that can boost creativity, cut down on information processing time, expedite memo drafting, and streamline additional day-to-day functions.

Institutional Guidelines and Best Practices

Aubrey discussed the House's internal policies on AI use. ChatGPT4 Plus is the only authorized LLM for use in the House at this time and it should be used in accordance with the House’s official guidance. If a legislative staffer works for the Senate, that chamber has yet to issue any official AI guidance, so Aubrey encourages Senate staff to seek guidance from senior staff and leadership. One additional option available and recommended for all staff on the Hill is to start the conversation with your team about an internal office-specific AI use policy to ensure that approved use cases and best practices are clearly understood.

Practical Applications of AI in Communications: Enhancing Creativity and Efficiency

For audience members who had not yet explored ChatGPT, Aubrey performed a quick demonstration of the tool and submitted a few sample prompts, showcasing how AI can save staff time and build capacity.

One of the amazing things about ChatGPT is that you can really ask it anything…A very basic question to put into it to get the creativity and brain flowing is something like “I'm an LA on the Hill. How can I use this tool? What are some examples?” I think one of the hardest parts about starting to experiment with a new tool like this is just thinking outside of your normal workflows to learn how it can be useful and helpful… Just ask it and then follow up on one of the examples it provides and just try it out.

Taylor Swift, Deputy Director of Governance and Innovation at POPVOX Foundation, shared his enthusiasm for AI in aiding legislative staff with tasks like synthesizing complex information, drafting hearing questions, and generating bill acronyms, significantly saving time and enhancing creativity.

The thing that really, really excites me about this technology is, as someone who worked on the Hill, time is such an important currency within the institution, especially if you're a Legislative Assistant or an aide. You're spending so much time information gathering, researching, understanding different opinions and perspectives, and this tool - this OpenAI ChatGPT4 tool - there's such a return on your time investment.

[ChatGPT] will be really helpful to use in building your capacity and saving time on common tasks like synthesizing really complex information or bills like a clause in the US code or a contract. It's going to be able to help explain those things easily and, of course I mentioned legislative history, doing some Horizon scanning on tons of documents that are already publicly available, and then providing talking points and insights for your Member so that they could know more of the narratives compared to the US code and legislative lingo that everybody uses on the Hill.

If your [Member] isn't an appropriator ...it's going to be super important to have background research on every single [appropriations] subcommittee bill that potentially could affect your district. But a lot of the appropriation subcommittee bills are not readily available as data, and it's really hard to synthesize all of that information and then analyze it quickly. I tell that story a lot better using ChatGPT. I can go back and find those bills and go through everything and provide summaries of what I need.

Shoshana Weissmann, Digital Media and Policy Fellow at R Street Institute, shared how she uses AI to assist with background research, brainstorming, and drafting.

I was trying to remember certain bills to cite because I knew there was a bill that banned all algorithms … so I wanted to cite some examples and I asked [ChatGPT] for examples and it reminded me of the ones that I had forgotten and linked the bill text. In the last session our speaker made a really good point too that it's helpful if you ask it for citations so then you can go to the original link.

I'm sure you all are writing so much all the time…You all know this feeling when you're at the screen like reading it for the 10th time. Instead of doing that you can also have ChatGPT tell you [where edits are needed]. I've had really really great luck with it editing for me and even just giving me suggestions for better writing. Also memos are a good idea! You can have it turn stuff into memos, you’ll still want to re-write it in your own words after, but the way it distills information sometimes is really helpful if you've been really, really deep in something and need help.

 The Importance of Caution, Verification, and Knowing AI’s Limitations

All presenters agreed on the necessity of double-checking AI-generated content for accuracy and appropriateness. As an evolving technology, GenAI tools need to be used with transparency and caution, and although they offer the ability to streamline tasks for legislative staff, they should never be trusted as an official source, final product, or faultless assistant. As Aubrey put it:

One of the really great ways to think about generative AI, especially with ChatGPT4, is to think about it as a really qualified intern or a really junior research assistant. It's ‘somebody’ who you know you can ask to do a task for you, but you always want to double check the work… For a draft, you still need to read what it has produced, think through what it means, and you double check everything. Then you take inspiration from it to write something final.

The model doesn't necessarily know everything and it can definitely misinterpret things very easily. It's always good to use it for brainstorming, to use it for a summary of things, but it's important to ask it “what specific part of the document did you summarize to find this point that I'm going to use?” …It's a shortcut but not a substitution for the work that you should be doing anyway to really know the policy.

Encouraging AI Literacy and Experimentation

Panelists encouraged legislative staff to experiment with AI in their professional and personal lives to understand its capabilities and explore ways it can make their day-to-day to-do list a little less daunting. As Shoshana recommended:

AI really, really thrives with repetitive tasks and I know that there's a lot of repetitive tasks in Congress, understandably. But this can really help with that to stop the burnout. So when you feel like you're burning out or you feel like you're doing something again and again, those are really good moments to think through “should I check out what I can do through AI?” 

The AI 101 for Legislative Staff webinar aimed to start the conversation about how GenAI can be used by legislative correspondents, legislative assistants, and legislative directors to aid them during busy session weeks and boost their capacity over time. It provided insights into best practices, official guidance, privacy considerations, and the importance of ensuring that AI-users know both the opportunities and limitations of the tools. As AI continues to evolve, Congressional staff should take time to explore how these tools can be leveraged to enhance efficiency and creativity in their roles. In addition to this webinar, POPVOX Foundation and R Street Institute hosted a second presentation focused on AI use by communication staff.

Stay tuned to POPVOX Foundation’s resources for ongoing discussions in the field of AI adoption on Capitol Hill. 

Previous
Previous

The Senate Issues Guidelines for Responsible Internal AI Usage

Next
Next

ICYMI: Embracing AI in Congressional Communications