Abstract
Embedded knowledge generated during fieldwork in community development programs rarely enters organizational learning systems, despite widespread use of messaging platforms. These channels often fail to surface and scaffold tacit insights. The presence of supervisory authorities further inhibits vulnerable reflection. We introduce Sruthi, an LLM-powered peer-mode reflective agent that transforms fieldwork “sayability” using a non-evaluative relational voice grounded in two years of field transcripts. Sruthi scaffolds reflection through context-sensitive prompts that invite elaboration without disrupting conversational flow. Evidence from our five-day naturalistic deployment in rural India (n=73) on Telegram illustrates how Sruthi can act as a sociotechnical catalyst, coinciding with a shift in fieldwork conversations from rote logging to reflective, interpretive sensemaking. While the comparative design does not support causal claims, we reveal three mechanisms contributing to this shift: legitimizing emotional discourse, enabling retrospective reprocessing, and surfacing interpretive knowledge. We characterize boundary conditions that shape the limits of disclosure in fieldwork. We conclude with recommendations for peer-mode agents as sociotechnical infrastructure for reflective practice in fieldwork multi-party communication.
Funding Information
  • ICSSR (13/WLD-2025-1183/SCD)

This publication has 40 references indexed in Scilit: