RescueLens: LLM-Powered Triage and Action on Volunteer Feedback for Food Rescue

TL;DR

RescueLens tackles the challenge of converting large-scale volunteer feedback into timely, targeted actions for food rescue operations. Using an LLM-driven pipeline, it automatically categorizes feedback into seven categories and supports interventions by ranking donors/recipients and generating updated directions, achieving high recall and competitive precision on a real-world dataset from 412 Food Rescue. Deployed in practice, RescueLens reduces organizer workload by focusing on a small subset of donors who generate the majority of issues, while providing evidence-based direction rewrites to improve volunteer experiences. The work highlights practical integration considerations, governance, and generalizability to other nonprofits, and shares code and prompts to enable replication in similar contexts.

Abstract

Food rescue organizations simultaneously tackle food insecurity and waste by working with volunteers to redistribute food from donors who have excess to recipients who need it. Volunteer feedback allows food rescue organizations to identify issues early and ensure volunteer satisfaction. However, food rescue organizations monitor feedback manually, which can be cumbersome and labor-intensive, making it difficult to prioritize which issues are most important. In this work, we investigate how large language models (LLMs) assist food rescue organizers in understanding and taking action based on volunteer experiences. We work with 412 Food Rescue, a large food rescue organization based in Pittsburgh, Pennsylvania, to design RescueLens, an LLM-powered tool that automatically categorizes volunteer feedback, suggests donors and recipients to follow up with, and updates volunteer directions based on feedback. We evaluate the performance of RescueLens on an annotated dataset, and show that it can recover 96% of volunteer issues at 71% precision. Moreover, by ranking donors and recipients according to their rates of volunteer issues, RescueLens allows organizers to focus on 0.5% of donors responsible for more than 30% of volunteer issues. RescueLens is now deployed at 412 Food Rescue and through semi-structured interviews with organizers, we find that RescueLens streamlines the feedback process so organizers better allocate their time.

Figures (8)

• Figure 1: We introduce RescueLens, an LLM-powered tool that automatically analyzes volunteer feedback in food rescue. Our tool first categorizes volunteer issues into different categories, such as Recipient Problem and Update Contact. RescueLens then uses these predictions to discover trends in volunteer feedback, identify which donors and recipients require interventions, and suggest updates to the directions based on feedback.
• Figure 2: After each rescue trip, 412 Food Rescue collects a rating out of four along with text-based feedback.
• Figure 3: We conduct an ablation study to understand the importance of different aspects of RescueLens. We find that RescueLens performs well because of a combination of few-shot learning and task-specific guidelines.
• Figure 4: We assess the performance of our direction rewrite module according to three criteria: helpfulness, novelty, and clarity. Across these criteria, we find that RescueLens performs well, averaging over a 4.7/5 across all three categories.
• Figure 5: We compute scores for donors and recipients using the formula from Section \ref{['sec:donor_recipient']}, then plot the distribution of scores. We show that only a few donors and recipients require intervention, and by focusing on these few, we can reduce organizer efforts.