Diversity and Inequality in Information Diffusion on Social Networks
summary
Uses a preferential attachment with unequal communities model plus estimated homophily to show when seeding for proportional representation also improves reach. Proposes two constrained heuristics — strict parity and relaxed diversity seeding — and compares them to degree, degree-discount and greedy on theory and on DBLP (≈53,000 nodes) under an independent cascade p=0.01. Theory gives a closed-form threshold on seedset size; empirical runs (≈10,000 simulations) show the fairness/efficiency trade-off is small, relaxed diversity or neighbor-seeding often matches or beats agnostic seeding once the seedset is large enough.
AI-written summary · about