Human-AI Narrative Synthesis to Foster Shared Understanding in Civic Decision-Making

Community engagement processes, like school district rezoning, generate massive volumes of feedback that overwhelm traditional synthesis methods, creating barriers to shared understanding between civic leaders and community members alike. To address this, we developed StoryBuilder, a human-AI collaborative pipeline that transforms community input into accessible first-person narratives, deployed through a mobile-friendly interface. Our mixed-methods evaluation found that narratives helped community members relate across diverse perspectives, and that experience-grounded narratives generated greater respect and trust than opinion-heavy ones. We contribute a human-AI narrative synthesis system and insights on its acceptance and effectiveness in a real-world civic context.

Accepted to CHI 2026!

Voice to Vision: Enabling Shared Understanding in Civic Decision-Making through Participatory Data Infrastructure

Trust and transparency in civic decision-making, like neighborhood planning, are eroding as community members frequently report sending feedback "into a void" without knowing how their input influences outcomes. To address this, we introduce Voice to Vision, a sociotechnical system that bridges community voices and planning outputs through interfaces for both community members and planners. Through iterative design and field evaluation, we find that community members want to see themselves reflected in the process and understand the rigor behind decisions, while planners value tools that make sense of diverse inputs. Our work contributes empirical findings and a complete system for how digital platforms can promote transparency and legitimacy in civic decision-making.

Accepted to CSCW 2026!

The Impacts of Transparency and Personalization on Feelings of Agency and Connection in Democratic Decision Making

Community engagement processes often shape policies that affect people's daily lives, yet they frequently struggle to build transparency, understanding, and agency. This study examines how varying levels and types of transparency, including personalization, in technology-enabled civic decision-making affect perceptions of agency, vertical and horizontal transparency, and community connection. Through an experiment where participants advocated for a local policy position and received a decision under varying transparency conditions, we find that increased transparency improved perceptions of agency and transparency, while personalization had limited effects. Qualitative reflections highlighted horizontal transparency as particularly valuable for opening perspectives. We discuss design implications for civic technologies.

Accepted to CHI 2026!

BoundarEase: Fostering Constructive Community Engagement to Inform More Equitable Student Assignment Policies

Public school districts shape access to quality education through student assignment policies like school attendance boundaries, yet community engagement processes for changing such policies are often opaque, cumbersome, and polarizing in ways that can perpetuate disparities. In collaboration with a large US public school district, we designed and evaluated BoundarEase, a web platform that allows community members to explore and offer feedback on potential boundaries based on their preferences. Through formative and evaluative studies, we find that BoundarEase prompts reflection on how policies impact families beyond one's own and increases transparency around policy proposals. Our work offers insights into community engagement challenges for student assignment policies and how sociotechnical systems can help mitigate polarization in local policymaking.

Accepted to CSCW 2025!

Coalesce: An Accessible Mixed-Initiative System for Designing Community-Centric Questionnaires

Incorporating community input into civic decision-making is crucial for inclusive governance, yet public officials often struggle to formulate effective questions due to constraints like time, resources, and limited questionnaire design experience. To address this, we present Coalesce, a mixed-initiative system that leverages large language models (LLMs) to help civic leaders craft tailored questions for surveys, interviews, and conversation guides. Through an iterative human-centered design process and a real-world user study, we found that Coalesce improved participants' confidence in questionnaire design and supported diverse workflows, while raising important questions about human agency and over-reliance on AI. These findings highlight the potential for intelligent interfaces to reshape how civic leaders engage with their communities.

Accepted to IUI 2025!

Sage: Design Implications of a Task-Oriented Hybrid Chatbot for Strategic Storytelling

People often turn to general-purpose chatbots for guidance in high-stakes self-advocacy contexts, yet these settings amplify risks around misalignment, over-coaching, and loss of narrative authority. To address this, we present Sage, a hybrid conversational interface that combines structured question flows and user-facing notepads with LLM-based language generation to support strategic storytelling — helping users reflect on lived experiences while preparing actionable messages for audiences in positions of power. Through expert interviews, we find the system promising for organizing overwhelming situations, prompting perspective-taking, and generating usable drafts, while surfacing design tensions around authenticity, guardrails, and memory management. We contribute design implications for hybrid conversational systems that support sensitive communication work without substituting for the user's voice.

Accepted to DIS 2026!

SenseMate: An Accessible and Beginner-Friendly Human-AI Platform for Qualitative Data Analysis

Community organizations face challenges harnessing qualitative data analysis to understand diverse constituent perspectives, with qualitative coding being particularly time-consuming and difficult to perform consistently across expert and beginner sensemakers. To address this, we present SenseMate, a human-AI system that leverages rationale extraction models to semi-automate qualitative coding, producing theme recommendations and human-interpretable explanations. Through an iterative, human-centered design process and an online experiment, we find that model recommendations and explanations improve intercoder reliability and coding accuracy while remaining accessible to beginners. SenseMate contributes a privacy-preserving alternative to LLM-based approaches and design insights for future qualitative data analysis systems.

Accepted to IUI 2024!

Future CMS Schools

In collaboration with a team from CCC and Charlotte-Mecklenburg Schools (CMS), we led a novel community engagement process around two magnet schools opening in fall 2023. We designed online platforms and helped facilitate 48 small-group conversations and interviews, engaging over 400 parents, students, and community members. The process collected 961 stories and suggestions on hopes and concerns for the new schools, informing the development of 8 magnet programs.

Facilimate

I applied user-centered design methods to create a digital support tool for people who facilitate small-group conversations. Using Facilimate, facilitators at any skill level can more easily manage time and follow conversation guides.

Shared Mobility Visual Analytics Tool

I created a visual analytics tool that can help city planners manage their shared mobility services. The tool is a React web application that analyzes scooter-share event data, estimates spatial demand, and generates an interactive data visualization page where users can map out usage and demand.

How to Not Get Rich: An Empirical Study of Donations in Open Source

Open source is ubiquitous and forms the digital infrastructure of our society, yet sustaining open source has become increasingly difficult due to growing demands and developer burnout. Donations are gaining in popularity as a potential method of sustaining open source. This research project is the first large-scale study to investigate the prevalence and impact of donations on open source.

The Relationship between Public Transit and Bikeshare Ridership

In this data science project, I explored the causal relationship between bikeshare and public transit networks in Boston, Philadelphia, and Washington DC using doubly robust estimators. I applied the economic concept of complements and substitutes to analyze how bikesharing could be used to support or supplant the first-mile/last-mile problem.

TASBE Flow Analytics

TASBE is a user-friendly and open-source environment that visually represents and analyzes flow cytometry data. In addition to feature development, I worked closely with biologists to design a customizable Excel interface that enables biologists without programming skills to set up analysis workflows in TASBE.

Interactive Robotics Research

I collaborated with three other students to analyze robot vision data using python OpenCV to program a robotic arm to autonomously play the card game SET (project 1) and to replicate user-built cube structures (project 2).

Computational Microbiology Research

I worked on several computational microbiology research projects with the goal of generating unified theories of microbial community function that impact all aspects of our lives, ranging from the environment to human health. I applied my programming and data science knowledge to analyze metagenomics data to understand nutrient cycling, apply network analysis to examine the effects of perturbation on microbial communities, and write software that mines genome databases.