Making Sense of Civic Voice Using Computer-Driven Analysis with the NYC Civic Engagement Commission
by Joy Ming
The Challenge
How can we make sure that when people engage with the government, their voice matters?
The New York City Civic Engagement Commission (CEC) is mandated to directly address questions of civic trust and democracy. One of its initiatives, the People’s Money, is the City’s first participatory budgeting process, launched in 2022. Through a series of public outreach campaigns, all New Yorkers were encouraged to share their voices and the 12,000+ residents who participated generated 4,000+ ideas for how $5 million of the city budget could be used to address local community needs. 110,000+ residents then voted on ideas that best represented their priorities and 46 ideas were ultimately selected to be funded. As a Siegel Family Endowment PiTech PhD Impact Fellow this summer, my goal was to analyze the rich feedback that New Yorkers shared with their government through the participatory budgeting process in order to help the CEC improve the next iterations of the People’s Money and to help other city agencies better understand people’s concerns.
The Concerns New Yorkers Voiced
I applied natural language processing techniques to analyze data consisting of 4,000 independent text entries, each representing a New Yorker’s idea (see the pipeline here). My primary analysis used unsupervised learning to highlight key themes and visualize their relationships. I conducted it using BERTopic, which implements topic modeling to identify 39 topics and provides hierarchical analysis to help merge these topics into 16 key themes. I was able to validate that these themes were fairly reflective of the concerns of NYC residents because they mostly matched and extended previous issues identified by an annual community board needs assessments. Using further qualitative analysis, the themes were summarized in the following 6 problem areas, ordered by the number of ideas submitted:
Activities for Youth and Seniors: The largest percent of submitted (27%) and selected (28%) ideas suggested different programs for youth (i.e., sports, arts/culture) and seniors, including intergenerational mentorship or technological help.
Community Health and Wellness: Another 28% of selected ideas addressed access to health and mental health education/services as well as to nutrition/food justice.
Worker Support and Opportunities: Many ideas promoted vocational education to address unemployment, especially to help immigrants overcome language/legal barriers, give teens internship opportunities and support working families.
Issues Related to Homelessness: Top of mind were also ideas related to maintaining cleanliness of neighborhood streets, addressing problems around housing and shelters, and preventing and rehabilitating drug addiction.
Community Safety: The second largest percentage of ideas selected to be funded (15%) were related to reducing crime through improving relationships with law enforcement and decreasing youth gun violence and gangs.
Neighborhood Infrastructure: Residents raised concerns about public spaces (gardens) and infrastructure (sidewalks, transportation) and promoted neighborhood projects to help fundraise for parks and schools.
Figure 1. Bar chart showing the number of ideas from each of the 16 themes at each stage of the process.
Figure 2. Visualization of how the 16 themes relate to the 6 problem areas.
Joy Ming
The Impact and Path Forward
I discussed these insights in interviews and focus groups with 11 CEC staff members and 5 other government officials from different agencies. Through these interviews, I found that there were two main ways that the CEC could improve the next iterations of the People’s Money. The computer-generated analysis could help:
Inform the City’s resource allocation strategy by identifying popular concerns, connecting themes for deeper conversations, and distinguishing people, problems, and solutions for more precision.
Make the participatory budgeting process more efficient by automatically categorizing ideas by keywords and identifying duplicates, highlighting the importance of using verbs and specificity for more concrete ideas and help people understand and build on each others’ inputs, and monitoring the different distribution and relationships among themes as well as see how the structures of ideas evolve over time.
More work will be done to make my insights available to NYC Open Data, other city agencies, community-based organizations, and participants in past and future participatory budgeting processes.
The first contribution of this work is that it helped uncover insights in the civic feedback from the NYC participatory budgeting process, highlighting the lived experiences of residents to policymakers. Additionally, this work proposed improvements to future processes for the CEC. These will inform ongoing discussions on how to encourage meaningful dialogue with communities and integrate computer-driven analysis. Finally, this work can serve as a launchpad to further explore the role of computer-driven analysis in governance and how these processes could be used even in settings with limited resources.