OPTIMAL COGNITIVE DIVISION OF LABOR IN HUMAN–AI SYSTEMS
The growing integration of artificial intelligence (AI) into intellectual work raises a fundamental question: which cognitive operations should be performed by humans, which should be delegated to AI, and under what conditions should responsibility shift between them? Existing approaches to human–AI collaboration frequently emphasize task performance, automation, augmentation, or comparative advantage. However, maximizing immediate productivity does not necessarily maximize decision quality, meaningful human oversight, or the preservation of human cognitive capabilities. This article proposes a conceptual framework termed Dual-Horizon Cognitive Allocation (DHCA). The framework treats cognitive division of labor as a dynamic, context-dependent allocation problem involving two interconnected objectives: immediate task effectiveness and long-term human cognitive competence. It integrates six considerations: comparative task performance, consequences of error, independent verifiability, metacognitive calibration, learning value, and accountability. The framework distinguishes human-led execution, AI-led execution, sequential collaboration, parallel evaluation, and deliberate non-delegation. It further proposes that optimal allocation depends not only on who can generate an output more efficiently but also on who can recognize errors, justify decisions, recover from failures, and maintain essential skills over repeated interactions. Five testable hypotheses and an experimental evaluation strategy are presented. The proposed model is theoretical and has not yet undergone empirical validation. Its intended contribution is to connect performance-oriented task allocation with psychological mechanisms of metacognitive regulation and longitudinal capability development.
Authors
- Roman Shaban (ORCID: https://orcid.org/0009-0009-5259-6102)
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-14
- DOI
- https://doi.org/10.5281/zenodo.24216641
- Primary Topic
- Human-Automation Interaction and Safety
- Type
- preprint