Interpretive infrastructure for public participation
Tracing Matters of Concern is a research and development project focused on computationally assisted analysis and visualization of large volumes of open-ended text. Its goal is to help practitioners and publics make sense of complex and nuanced public opinion and stakeholder input on a wide range of issues.
The problem we're solving
Public agencies, foundations, non-profits, researchers, and civic organizations routinely ask communities what they think. They design surveys, run town halls, open public comment processes. The responses come back in the thousands — and then, more often than not, they get summarized by a small team using judgment calls that are difficult to audit and impossible to fully justify.
This is not a failure of intent. It is a failure of infrastructure. The tools available for analyzing open-ended text at scale either flatten complexity into numbers — sentiment scores, frequency counts, Likert scales — or require so much manual effort that only a fraction of responses ever get read.
Voxil was built to fill that gap. It processes open-ended responses at scale, finds the natural structure in what people say, and organizes findings into navigable hierarchies where every theme is traceable to its source. The goal is not to replace human judgment — it is to give human judgment something trustworthy to work with.
We do not collapse complexity. The diversity and nuance of public perspectives are preserved in the output — which means minority views and dissenting voices appear alongside majority ones, not averaged away.
Every analytical object — every category, summary, and finding — links back to the original responses that produced it. This is not just a technical feature. It is an accountability commitment.
Topics emerge from the data, labeled in ways that support interpretation without forcing premature conclusions. We describe what communities say; we do not classify them.
The system is designed to support human deliberation and sensemaking, not replace it. Outputs are starting points for conversation, not answers to be accepted at face value.
How we think about the work
Open-ended responses as structured meaning
Most analysis tools treat open-ended responses as the messy residue left over after the real data — the closed questions — has been collected. We treat them as the most valuable part: structured expressions of meaning that can be traced, compared, and mobilized without collapsing their complexity.
Rather than forcing responses into predefined categories, Voxil uses computationally assisted categorization — combining statistical and language models to surface recurring patterns while preserving access to the underlying texts that produced them. This makes it possible to work at scale without sacrificing interpretive accountability to source material.
Tracing matters of concern
The approach draws on science and technology studies — specifically the distinction between stabilizing premature facts and tracing live matters of concern. Public opinion on contested issues is rarely settled. It is conditional, ambivalent, and context-dependent. Voxil is designed to surface that complexity rather than resolve it prematurely.
This shapes every design decision: why topics emerge from data rather than being imposed in advance, why every analytical object remains traceable to original responses, and why the system is built to support deliberation rather than deliver verdicts.
From measurement to sensemaking
Conventional survey analysis is built around measurement — extracting a signal from noise. Voxil is built around sensemaking — helping people understand what is being said, why it matters, and what to do about it. The difference is not just technical. It reflects a different relationship between analysts and the communities whose voices are being analyzed.
Interpretive infrastructure
Voxil's outputs are designed to function as interpretive artifacts: materials that can be used to make sense of complex input and guide conversation, planning, and action. Analysis is a starting point, not an endpoint. The central question is always what different groups understand the findings to mean — and how that understanding can be translated into next steps.
Project Team
Gabriel Harp is the principal investigator and developer of the project. He leads research on computationally assisted sensemaking, its application to public engagement and policy analysis, integration with participatory design processes, and its impact on implementation-driven project and programs.
University of Michigan
Ann Arbor, MI
How people work with us
Most conversations start with a dataset someone has been sitting on and doesn't know how to use, or a community process they're about to run and want to think through more carefully. We're happy to talk at either stage.
Common starting points include:
- • Research collaborations and pilot projects with new datasets or communities
- • Integration with existing engagement platforms and processes
- • Session design for workshops, planning meetings, or stakeholder convenings
- • Methodology questions and research design consultation
- • Speaking, media, and academic inquiries
Want to learn more?
Explore our case studies to see how Voxil has been applied, or read our detailed overview to understand the workflow and data requirements.