Standard behavioral frameworks often overlook how social network topology shapes the coordination of inflation expectations. To address this, I build a hybrid agent-based model that integrates “narrative-rooted” heuristics within a standard New Keynesian structural scaffold. In this framework, beliefs evolve through dual channels: performance-based selection (heuristic switching) and social diffusion (DeGroot learning). Simulations across canonical topologies reveal that seeding a target-based narrative in high-centrality nodes compresses forecast dispersion and accelerates convergence. However, a structural trade-off emerges: while performance evaluations dampen distorting narratives, strong social persuasion reduces volatility but simultaneously decouples expectations from economic fundamentals. This occurs because agents prioritize peer alignment over objective data, highlighting the double-edged nature of social networks: they stabilize expectations when credible narratives diffuse through hubs, but sever the link between beliefs and fundamentals when conformity overrides economic signals.