Agenda setting for LA wildfires recovery: Exploring community feedback
The agenda setting phase of the Los Angeles wildfires recovery engagement is complete. During this phase, community members shared thousands of comments about what they need most for recovery and rebuilding. You can download the comments to see what people said.
Breaking down the complexity
Residents responded with overwhelming volume and a wide range of views. We made data visualizations to simplify this complexity. They organize and show comments by topic and theme. This tool will help you explore what the community said and see patterns in the feedback.
We designed the analysis and visuals below to help you:
- See the full range of community perspectives on recovery priorities
- Understand how different viewpoints cluster around key issues
- Explore the complexity of challenges facing fire-affected communities
- Prepare for upcoming community discussions
Note: The info on this page does not offer final answers or solutions. Instead, it shows us the questions and tradeoffs we need to work through together. During the deliberation phase, we'll find the best way forward. Community members will work through these complex issues in a thoughtful manner.
We will invite California residents to help us with this. Sign up to stay informed.
How to interpret these charts
Each chart shows the comments made about one topic.
- Each dot represents an individual comment.
- Dots are color-coded, with each color representing a group that came up in the comments.
- We used AI to identify themes. The comments that were identified as similar appear closer together on the chart.
Data visualizations
Download the full datasetHow we built these charts
Below is a step-by-step explanation of the methodology for these charts.
-
Process and restructure the dataWe did this for all the comments within a topic. We turned each comment into numerical data, or semantic embeddings. These embeddings show the meaning of the comments in a way that computers can understand and use. We did this using Snowflake Cortex Embed Text 1024.
-
Find conversation groupsWe used the BERTopic framework to group comments within each topic area by similarities. This helped us sort them into different categories based on common themes. We then used UMAP to reduce the dimensionality of the comments. This tool made complex data simpler to visualize. We used density-based clustering with HDBSCAN and then tuned the algorithm settings for each opportunity area.
-
Label the conversation groupsWe grouped the comments by topic, then used the 'Claude 3.5 Sonnet' model to create the initial label for each group. Our team of data analysts then reviewed and edited these labels for clarity. Each label is a short phrase that reflects the main idea or topic that the comments have in common.
-
Plot the commentsWe used a topic-supervised UMAP. We tuned it for each topic.This allowed us to turn all the numerical data into two simple coordinates that we could plot on a 2D chart. The comments are color-coded by conversation group. This makes it easy to see patterns and connections.
Download the data
The comments dataset does not contain personally identifying information. We have not edited or modified the comments in any way.
Download the full datasetData definitions
These are the columns in this dataset and their definitions.
Comment ID
A unique value to identify each comment.
Comment
The full text of a comment submitted by a participant. Comments have not been edited for clarity, content, grammar, or for any other reason.
Topic
This column lists the topic a commenter was responding to. There were 10 topics.
- Emergency communication
- Debris removal and environmental recovery
- Financial and legal assistance
- Emotional and mental health support
- Housing and rebuilding
- Economic recovery and small business support
- Infrastructure and utilities restoration
- Wildfire prevention prioritization and accountability
- Climate and community resilience
- Evacuation and emergency coordination
We also included comments that were not in response to a specific topic. We labeled those comments with “general comment.”
Conversation group
The conversation group a comment aligns to.