Automation or Augmentation? Task Characteristics, Human-Centered Goals, and the Deployment of AI

Abstract

The rise of artificial intelligence raises the question of when AI is adopted and whether it is deployed to replace versus augment workers. Studies of prior technologies such as robots and computers emphasize the role of task routinization and the cognitive vs. physical nature of tasks, yet theorizing and empirical evidence in the context of AI technologies are scarce. Moreover, the diffusion of AI is not automatic: Decisions on whether and how to use AI are made by organizations, and these organizations may pursue not only production efficiency but also goals that require the continued involvement of humans (e.g., providing jobs, human capital development, community engagement). We develop hypotheses regarding the role of task characteristics and human-centered goals and test them using data on more than 1,200 projects that involve participants in scientific research. Although task characteristics predict AI exposure in a general sense, they relate differently to automation and augmentation, emphasizing the need to distinguish different types of exposure in future work. Projects pursuing human-centered goals are less likely than others to use automation, while they are just as likely to use AI to augment. We contribute to the literature on AI in organizations by providing novel evidence from task-level data and by highlighting that the impact of AI on humans is, partly, a matter of human choice.

Keywords: Augmentation, automation, task routinization, human capital, crowd science