AI Assistance May Degrade Productivity, New Model Shows
A new model from arXiv (2605.11350) reveals that increased AI assistance can paradoxically reduce productivity. Researchers modeled human agents with varying skill levels exerting utility-maximizing effort with AI help. They found that when skill development or AI unreliability is endogenous, a productivity paradox emerges: more AI aid degrades output. Long-term, AI literacy gaps lead to skill polarization in steady state. The study highlights that AI's impact is not uniformly positive and depends on human adaptation.
Key facts
- arXiv paper 2605.11350 proposes a model of human-AI interaction
- Endogenous skill development or AI unreliability can induce a productivity paradox
- Increased AI assistance may degrade productivity
- Skill polarization emerges when accounting for heterogeneity in AI literacy
- AI literacy is the capability to identify and adapt to inaccurate AI outputs
- The model examines long-term distributional effects of AI on skill
- Human agents with varying skill levels exert utility-maximizing effort
- The study elucidates mechanisms by which AI affects productivity
Entities
Institutions
- arXiv