The AI Paradox: How Efficiency is Dismantling the Relational Infrastructure of Transformative Learning

Abstract

AI agents make students 60% more productive—but reduce their social and emotional communication by 23% (Ju & Aral, 2025). For networked learning, this is a crisis masquerading as progress. We're optimizing for efficiency while accidentally dismantling the relational networks where transformative learning happens. AI doesn't just augment learning networks—it reconfigures them, weakening peer-to-peer ties and replacing dialogue with transactions.

This roundtable challenges the productivity narrative and asks: Can we design AI agents that strengthen learning networks instead of substituting for them? What if AI created connection opportunities instead of answering questions? What if it surfaced emotion instead of eliminating "coordination effort"? What if we evaluated AI on network-strengthening rather than task completion?