Human-machine-learning integration and task allocation in citizen science

Abstract

AbstractThe field of citizen science involves the participation of citizens across different stages of a scientific project; within this field there is currently a rapid expansion of the integration of humans and AI computational technologies based on machine learning and/or neural networking-based paradigms. The distribution of tasks between citizens (“the crowd”), experts, and this type of technologies has received relatively little attention. To illustrate the current state of task allocation in citizen science projects that integrate humans and computational technologies, an integrative literature review of 50 peer-reviewed papers was conducted. A framework was used for characterizing citizen science projects based on two main dimensions: (a) the nature of the task outsourced to the crowd, and (b) the skills required by the crowd to perform a task. The framework was extended to include tasks performed by experts and AI computational technologies as well. Most of the tasks citizens do in the reported projects are well-structured, involve little interdependence, and require skills prevalent among the general population. The work of experts is typically structured and at a higher-level of interdependence than that of citizens, requiring expertize in specific fields. Unsurprisingly, AI computational technologies are capable of performing mostly well-structured tasks at a high-level of interdependence. It is argued that the distribution of tasks that results from the combination of computation and citizen science may disincentivize certain volunteer groups. Assigning tasks in a meaningful way to citizen scientists alongside experts and AI computational technologies is an unavoidable design challenge.