10  Agent behavior and decision making

To analyze interviews and focus group data, we used the following process:

  1. Open Coding System: Transcripts were coded inductively, guided by the research questions (Glaser, 2016; Saldaña, 2015). The research team developed thematic memos for each producer group (beans, beef/dairy, leafy greens) through collaborative “theming” (Saldaña, 2015), returning to original transcripts throughout to preserve participants’ language. This produced four thematic areas: (1) personal values, (2) relationships and reliability, (3) market conditions, and (4) logistics and infrastructure.
  2. Secondary Coding: Transcripts were recoded using MAXQDA to identify key internal and external drivers within each theme. A structured codebook, which defined core themes as well as internal and external influences, was developed based on thematic memos generated during the analysis.
  3. Mapping to COM-B Framework: Following prior applications of inductive coding to adapted COM-B frameworks (Ambrose-Oji et al., 2022; Evans et al., 2025; Farrell et al., 2023; Mastenbroek, 2025), our study first inductively identified themes and internal/external factors, then deductively mapping those findings onto the adapted framework’s three domains to explore how potential NYC procurement policies might influence current behaviors.

10.1 How base agents respond to requests with offers

10.2 Scoring incoming requests for product in the model