From the start of my research, my mentor has reminded me that one of the biggest challenges would be obtaining reliable data. My original plan was to reach out to nonprofits and food rescue organizations directly to learn from their routing and distribution data. However, because communication with organizations can take time, I began exploring publicly available datasets from the NYC government. I was surprised to find that the city has many high-quality datasets related to food insecurity rate, which has helped me keep building my project while waiting for nonprofit responses.
One ethical issue I am grappling with is how to define “better” in an optimization model. At first, I thought the goal was simply to maximize efficiency: reduce travel time, minimize cost, and deliver as much food as possible. But as I started planning the algorithm, I realized that this philosophy has a serious flaw. If the model only rewards efficiency, it may simply prioritize nearby locations or easy-to-serve neighborhoods. That might look good mathematically, but it does not necessarily mean the food distribution system has become more just or more effective.
This has changed the way I think about my research. An algorithm is not neutral just because it is mathematical. The objective function reflects human values. If I only tell the model to save time, then the model will save time, even if that means leaving out communities with greater need. In food rescue and food insecurity work, the goal cannot only be operational efficiency. It also has to consider equity, urgency, perishability, and the needs of different neighborhoods.
In response, I am trying to build fairness into the structure of the model itself. Instead of only asking, “What is the shortest route?” I am also asking, “Who is being served?” “Which neighborhoods are repeatedly left out?” “How should we weigh distance against need?” and “How can an algorithm help organizations make decisions without hiding the ethical tradeoffs?” I am considering using NYC public data to create an equity index or priority score, so that the routing model accounts not only for distance and capacity, but also for community need.