Task recommendations for self-assigning spontaneous volunteers

Abstract

Spontaneous volunteers are a critical and often overlooked resource during disaster response. Frameworks that do consider spontaneous volunteers often rely on the use of a central controller to assign them to tasks, which is not always feasible due to privacy and legal liabilities. We investigate a new paradigm in which volunteers self-assign to tasks from an ordered list of recommendations. We present the problem of generating task recommendations as the Online Task Recommendation problem. Subsequently, we present the Maximal Ordered Multiple Matching problem, an integer optimization approach to construct task recommendations for volunteers by matching tasks and volunteers based on the expected volunteer utility from completing each task and the societal value of completing each task. We prove the presented matching model is polynomially solvable. We present a method to calculate improved objective coefficients by iteratively solving the proposed integer program and computing the likelihood that volunteers would select each recommended task. We demonstrate the benefit of the proposed methods with both real-world and synthetic case studies with spatially distributed tasks using Monte Carlo simulations. We find that recommendations generated by the proposed models guide the completion of the most critical tasks by reducing task over-service redundant volunteer responses often associated with spontaneous volunteers while recommending tasks that are preferred by the volunteers.