Methodology

Theme Identification

1. All comments were gathered into a spreadsheet as they were originally received and corrected for typos.
2. An AI model was used to turn sentence fragments into complete sentences without adding additional words or meaning. Comments were marked in red where it had to guess. For example, a note reading "local hire" became "There should be local hiring."
3. A person then read each updated comment individually to confirm accuracy.
4. The same AI model was used to group the comments into five themes given the project RFP, the project guidelines, and community context. Because AI models can carry bias against people of color, documents from Healthy Black Families' research were provided to provide the model with the correct context.
5. All comments were then run through a second AI model to confirm the categorization into the five themes.

Data Collection Events

Input was collected from community members in multiple formats. In person events allowed for writing on sticky notes and index cards. Online events recorded input through chat. Some comments were given verbally to staff members, who recorded what they heard.