Algorithmic Rural Road Planning in India: Constrained Capacities and Choices in Public Sector

Abstract

The use of AI and other algorithms in resource-constrained public sector settings of the developing world face unique technical and social challenges. The organizational and institutional realities of the public sector such as legacy data/IT systems, embedded work culture, bureaucratic norms and resource-constraints both explicitly and implicitly shape deliberation, design and deployment of public sector data science projects. Through the case of algorithmic rural road planning in a large-scale government program in India, our work demonstrates how algorithms can be positively utilized within the context of constrained capacities and choices. As practitioners deeply involved in the entire project life-cycle, our action-research provides an intimate and reflective account of how production of even seemingly "simple"algorithmic projects pose non-trivial complexities and challenges in the public sector. We situate the conversation around the humans-in-the-loop in our setting and show how public sector characteristics impact participatory design, choice of interfaces, data inequities and algorithm design. Further, we show how the preparation and production of technology by constrained capacities can be counter-productive and detrimental even before the technology is put to use. This further expands the scope of the debate concerning the use of public sector algorithms. Understanding the nuances, practices and constraints in production of data science in the public sector will not only allow more responsible production of data sciences, but also help formulate realistic strategies to mitigate the risks involved in the use of algorithms in high-stakes public policy situations.