Many years of research have yielded computer modeling techniques that can predict the behavior of complex systems, such as traffic speeds in regional transportation systems, with high accuracy. However, the prediction accuracy suffers significantly when non-recurring events, such as traffic accidents, occur in these systems. Yet the impacts of such disruptions are precisely the events that vehicle operators need to be aware of when planning their trips. Techniques for autonomously detecting these events, such as automated incident detection from traffic flow data and computer vision, are active fields of research but currently offer significantly less accurate data than actual human observations. Therefore, introducing novel ways to identify and quantify disruptions using human input can improve modeling accuracy when speeds are disrupted, while raising new topics for research to address this large, unmet need. Blending human-relayed incident detection mined from social networks with existing traffic modeling techniques provides a promising new direction for improving accuracy in traffic speed prediction.