Artificial Intelligence and the Collaboration Paradox: Evidence from Innovation Networks In China

Using patent data from 284 prefecture-level and above cities in China over the period 2010–2022, this paper constructs inter-city innovation collaboration networks and systematically examine how artificial intelligence development affects such collaboration. The findings indicate that artificial intelligence development significantly reduces inter-city innovation collaboration. Mechanism analysis reveals two primary channels: a cost channel, whereby artificial intelligence-driven increases in urban specialization and centralization raise coordination costs and information asymmetries; and a substitution channel, through which artificial intelligence expands the efficiency and reach of knowledge spillovers, allowing informal knowledge exchange to crowd out formal collaborative arrangements. The findings reveal a collaboration paradox—artificial intelligence enhances individual innovation capacity while undermining inter-regional networks—with important implications for regional innovation policy. This work advances understanding of artificial intelligence’s economic effects by shifting the focus from firm-level outcomes to the meso-level structure of innovation networks and develops an integrated analytical framework linking artificial intelligence development to inter-city collaboration through cost and substitution mechanisms.