Our San Francisco Winner, Community Voice is an application that enhances canvassing campaigns with real-time constituent sentiment analysis. 

The Team

Community Voice was developed by Ari, Ned, Mira, Rob, and Austin in partnership with Fair Count, a nonpartisan nonprofit organization founded by Stacey Abrams in 2019 through deep organizing, Fair Count builds civic power in communities historically undercounted in the census, underrepresented at the polls, and torn apart by redistricting.

Background 

“A major part of what we do at Fair Count is the collection of qualitative data through our organizing, whether it is door knocking, one-on-ones, communications with faith institutions and other partners,” said Dr. Jeannine Abrams McLean, FairCount’s Executive Director, and a computational biologist. “We have this vast database of qualitative data. Can we use what we know in these very specific communities based on geography, and can we overlay the data that would come from canvassing so that we can get a more holistic picture of the community? How can we take that database that we already have and really use it to inform our work.” 

Fair Count began exploring these questions with generative AI as a member of Cooperative Impact Lab’s first Generative AI Cohort (winter of 2023/2024). Through their pilot project, they collected 120 voice memos from canvassers in Mississippi, produced transcripts, and ran a sentiment analysis using Claude. The process generated insights and data with the potential to inform organizing strategy – from county-level breakdowns of where people live, a basic sentiment analysis of each county, and accurately estimated turnout.

The Challenge

Fair Count considered the pilot a success and decided to join the hackathon as a challenge partner, hoping to identify a framework and specifications for a tool that could make sentiment analysis in their canvassing campaigns less labor-intensive, more timely, and actionable.

Fair Count’s Challenge Statement for the San Francisco Hackathon:  

How might we leverage AI to analyze constituent sentiments in “real-time” to better understand and address motivations and challenges during canvassing efforts?

The Result

The result and the winning project was Community Voice by Fair Count. “The fact that the hackathon participants were able to create a functioning tool in that amount of time was really mind-blowing to me,” said Dr. Jeannine Abrams McLean.

Community Voice has three capabilities:  

1. Prime the canvasser with specific topics relevant to the community in preparation for a door-knocking conversation (i.e., potholes, ballot boxes, voter apathy).

2. A voice recorder for canvassers to summarize their conversations in 30-90-second memos right after they took place.

3. The ability to generate a report with a sentiment analysis using Claude focused on issues of importance to the community member, the person’s plan to vote, and information shared by the canvasser.

Community Voice effectively contained the three capabilities into a single application. In the pilot program, organizers collected voice memos from canvassers, transcribed them individually, and subsequently ran an analysis of the transcripts through Claude. Community Voice has a user-friendly interface, runs a real-time thematic analysis, and generates insights to inform organizing strategy immediately.

A Look Ahead

Fair Count’s goal is to further develop Community Voice in 2025, prioritizing the tool's sustainability and best practices around privacy and security for data collection.

The New York Winner, Truth Tea: Brewing Accurate Narratives identifies disinformation, analyzes political narratives, and generates shareable counter-messaging in language.

The Team

It was developed by Rosario, Sarah, Taren, Marshall, Anupam, Liam, and Tudor in partnership withAAPI Victory Alliance, an organization founded in 2017 by leading progressive voices in the AAPI community. AAPI Victory Alliance builds AAPI political power and advocates for progressive policies.

The Challenge 

Because the AAPI community is diverse and non-monolithic, with over 100 languages spoken, AAPI Victory Alliance is invested in reaching the estimated 34% of AAPI voters with limited English proficiency. The organization joined the hackathon to work on the challenge of widespread disinformation by building trust with rapid in-language messaging to under-engaged voters. 

AAPI Victory Alliance’s Challenge Statement for the NYC Hackathon:

Develop an AI-powered solution to identify, counter, and mass amplification of accurate information in response to mis/disinformation in language and across various media channels and messaging platforms

The Result

The result and the winning project was Truth Tea: Brewing Accurate Narratives.

Here is how it works: A grassroots organizer finds a video in Hindi they want to analyze. They paste the video link in the Truth Tea input bar. Truth Tea gets the transcript, runs a gen-AI-powered analysis on its claims and political narrative, generates a summary of both, and identifies disinformation. Truth Tea then recommends counter-messaging in the form of sharable content. The prototype can be applied to different languages.

“We iterated on the challenge statement, and I was a little worried because I think we probably spent a good four hours just talking through the problems. And even when we thought we had an aha moment, we still did a gut check around the room to make sure it worked on the technical side, and that it was really solving the problem that we wanted to solve, and that it was practical,” said AAPI Victory Alliance Executive Director, Varun Nikore.

“Being from a technical background, I've never thought more on the political side, so the brainstorming session itself was great. Having a partner who potentially could use what we built was extremely useful as a sort of directional pull. Is this useful to you? No. OK, we probably shouldn't build it. I think that helped us focus down quite a bit. It was obviously a big topic that we were trying to tackle.” – Anupam Tiwari, Hackathon Participant.

“In the end, I feel that we ended up so much further than anything I had envisioned when I even came up with the challenge statement. I feel like not only do we have a solid roadmap for things that can be utilized in our community but also in other communities like the Latino community as well. This can be a template for other language-specific or community-specific initiatives.” - AAPI Victory Alliance Executive Director Varun Nikore. 

“I went to the AI Hackathon in hopes that I would get to find other organizing techies like me who wanted to tackle disinformation in multilingual spaces. I was pleasantly surprised to be able to be part of the team Varun (from AAPI Victory Alliance) had already planned. Our team built a prototype webpage with high coding programs like the Streamlit app that uses an API and with low code tools like a chatbot and specialized ChatGPT Meme-maker. Then, we had the team at SoSha integrate a nice social media strategy to take the wiki project a step further. It was amazing to see all the skillsets in our team working together to build this in 6 hours. These opportunities to collaborate in an open-source prototype matter because there's already a couple of projects, like one I am working on for Georgia organizing work, that will benefit from the knowledge this election cycle.” - Rosario Palacios, Hackathon Participant. 

A Look Ahead

AAPI Victory Alliance is working with members of its hackathon team to develop Truth Tea further.

Our Chicago Winner, Reddit Pulse Checker is a social listening tool for analyzing Reddit content to understand AAPI community conversations and sentiments.

The Team

The Reddit Pulse Checker was developed by Trevor, Nathan, Arthur, and Jalen in partnership with Asian American Futures. This organization activates young AAHHPIs and experiments with evidence-backed programming to shift mindsets at scale.

The Challenge

AAF came to the hackathon interested in exploring AI applications for social listening. “I wanted help summarizing and getting a pulse on the kinds of content and the conversation coming out of Asian America from top media and social media accounts,” said Kana Hammon, AAF’s Narrative Strategist. “A lot of what our organization is focused on is trying to understand these people who may not be as politically engaged or engage with social justice orgs and learn how we can reach them and engage with them to bring them along a journey towards supporting racial justice. Getting a pulse on what they're consuming and seeing and hearing what their version of Asian America looks like is a really, really important step for us to be able to empathize and connect with people who are not your typical nonprofit organization audience.”

The Result 

The result and winning project was the Reddit Pulse Checker. Here is how it works: The user identifies a subreddit of interest and runs the Reddit Pulse Checker Python script locally, which scrapes Reddit and prepares the data for LLM use. The script uses Clause 3.5 Sonnet to summarize the content by topic and analyze it according to six personas. Asian American Futures developed the six personas through extensive surveying and interviews to represent six common characters in their audience. Once finished, the script exports its analysis to the Miro App (running locally) for the staff to integrate into their communications workflow.

“One thing that my team was able to do was match content to guess which kind of persona it might speak to, and that was really exciting to see. I hadn't expected that they would do or try that level of analysis. Typically, we have to infer and guess who we think certain content is reaching. It was nice to see that it might be possible to systematize that more,” said Kana Hammon, AAF’s Narrative Strategist.  

A Look Ahead

Asian American Futures is interested in investing in AI-powered social listening tools to cut down on the time and work required of staff.