Abstract:
Debates about generative AI in software engineering often center on cognitive offloading and the possibility that developers will lose skills. This position paper argues that this focus is important but too narrow. Evidence of durable cognitive decline remains mixed, and offloading can involve task substitution, skill displacement, skill decay, or failures of skill formation. A more useful unit of analysis is the changing human–AI work system, including its metacognitive, social, motivational, and organizational conditions. Drawing on research on extended cognition, learning, teamwork, motivation, automation, and developer productivity, the paper examines how GenAI may reshape judgment, novice learning, shared understanding, communication, ownership, accountability, professional identity, status, and the distribution of productivity gains. It argues that the central question is not whether developers should avoid offloading, but which capabilities and forms of agency human–AI systems should preserve and cultivate, for whom, and under whose goals. Evaluations of AI-assisted work should therefore extend beyond task-level speed to learning, team knowledge, meaningful work, well-being, and the allocation of any time and effort that AI releases.
PPIG 2026 - 37th Annual Workshop
What if Offloading Is Not the Problem?