When AI does the work, what skill remains?
Diener and colleagues model workers choosing when to accept an AI answer and when to solve a task themselves. In their model, relying on AI can cut practice for new workers and dull expert skills even when today's answers are good. The paper tests a theory, not a real training plan.
What the source reports
Diener and colleagues built a model of work across a series of tasks. A worker can take an AI answer or solve a task without it. The model tracks how skill changes over time. Under its assumptions, using the tool to finish each task quickly can leave new workers with low skill. Skilled workers can lose ability too, even while their AI-assisted answers remain accurate. The authors study ways to reward independent practice. This is a theory paper. It does not show that a particular workplace training plan works or how large the skill loss would be outside the model.
Original source
- Title
- Today's Work, Tomorrow's Workforce: Incentivizing Learning in AI-Assisted Work
- Author
- Moritz Diener, Daniel Hendriks, Philipp Spitzer, Gerhard Satzger, Joshua Holstein, Wei Sun
- Publication
- arXiv cs.HC
- Date
- Thursday, November 5, 2026