Enabling Insights by Long-Term Evaluation of Social Impact Indicators of Engineered Products for Global Development using In-Situ Sensors and Deep Learning

Abstract

Remotely measuring social impact indicators of products in developing countries can enable researchers and practitioners to make informed decisions relative to the design of products, improvement of products, or social interventions that can help improve the lives of individuals. Collecting data for determining social impact indicators for long-term periods through manual methods can be cost prohibitive and preclude collection of data that could provide valuable insights. Using in-situ sensors remotely deployed and paired with deep learning can enable practitioners to collect long-term data that provides insights that can be as beneficial as data collected through manual observation but with the cost and continuity made possible by sensor devices. Postulates related to successfully developing and deploying this approach have been identified and their usefulness demonstrated through an example application related to a water hand pump in Uganda in which sensor data was collected over a five month span. Following these postulates can help researchers and practitioners avoid potential issues that could be encountered without them.