When Machines Will Take Over? Algorithms for Human-Machine Collaborative Decision Making in Healthcare

Artificial intelligence (AI) has increasingly become a popular alternative for performing tasks that are typically performed by humans. Mammography imaging is one context in which the role of AI is growing. Some experts claim that, with recent advancements in image processing algorithms and the increasing availability of data, AI will replace radiologists. Others argue that the rise of AI will change how diagnostic tasks are allocated, eventually paving the way for human-machine collaborative decision-making. In this research, we solve a hospital's AI acquisition problem for mammography imaging and redesign its operations for human-computer collaborative decision-making. To that end, we propose an optimization model for the hospital that minimizes costs related to mammography screening and determines whether and when a complete automation (AI alone) strategy or a delegation (collaboration between humans and machines) strategy is preferable to an expert-alone strategy. We find that the disease incidence relative to the ratio of follow-up against liability costs is an important determinant of whether the delegation strategy is preferable to the automation strategy. In addition, reductions in algorithmic cost could either result in the delegation (sharing of work between humans and machines) or full automation depending on the performance of the algorithm. Our work has implications beyond radiology imaging for the design of work in the AI era and in the human-machine collaboration context.