Removal of Learning Opportunities Through Task Automation: Why a Shortcut Can Complete the Task and Skip the Lesson

This conceptual paper introduces Removal of Learning Opportunities Through Task Automation, a causal mechanism in which a shortcut produces a result while removing encounters that could have developed a capacity. Completing a task and learning from it can therefore come apart. The account assesses learning through the ability to reproduce or evaluate results independently. Removing steps does not itself establish a learning loss: some effort teaches nothing, and tools can preserve useful practice. A claim of reduced learning requires evidence connecting the removed encounters to independent performance. The discussion connects the mechanism to Bainbridge’s account of the ironies of automation and examines Bastani and colleagues’ field experiment with generative AI in high school mathematics. Both tools improved practice performance. The GPT Base group performed worse on the subsequent unaided exam than the control group; GPT Tutor showed no statistically significant exam difference from that group. The researchers interpret reliance on supplied answers as displacing students’ own practice. Part of the Shared Capacity and Learning series within On Human Modeling, the paper distinguishes supported results from evidence of developed capacity. Its framework can inform human–AI research on assisted performance, preserved learning opportunities, and what people can do after assistance is removed.