Watts (2002) threshold model: global cascades from local interactions
Each node i adopts if the fraction of its neighbors already adopted exceeds threshold φᵢ (here uniform φ). Unlike simple SIR contagion, adoption requires multiple social contacts — a "complex contagion". Global cascade condition:
On Erdős-Rényi graphs, a cascade window exists in (φ, ⟨k⟩) space. Outside it, cascades die locally. The vulnerable cluster — nodes whose degree k ≤ 1/φ — must percolate for global spread.