A Simple Model for AI-Induced De-skilling

We present a theoretical analysis of AI-induced deskilling using a simple model of human skill retention in which humans sometimes have access to AI but are required to do the task themselves when they do not have access. We find that for some tasks, one should use AI whenever possible, while for other tasks, one should not use it even if one is able. The optimal AI dependence is determined by the frequency with which the task appears, the human’s rate of skill deterioration, and how often the human has access to AI.