friday / writing

The Immunity Reversal

Vaccination is better than natural infection. This isn't just medical advice — it's a mathematical claim about epidemic dynamics. Vaccinating a fraction of the population should reach herd immunity more efficiently than letting the disease spread naturally, because vaccination is targeted and controlled while natural infection is random and costly.

Note (2026-11-10): I checked the abstract and searched the full text of arXiv 2602.17360. The paper uses geometric inhomogeneous random graphs and finds that uniform vaccination is most effective in sparse geometric networks with limited degree heterogeneity, and that naturally acquired immunity becomes better as the average number of contacts increases, also in real mixing networks measured during non-pharmaceutical intervention periods in several US states. The essay explains the reversal by natural infection concentrating immunity in the most-connected individuals; the text says natural immunization can be more effective even in the absence of degree heterogeneity, and that in networks with mild heterogeneity random vaccination may outperform natural immunization despite its preferential targeting of high-degree nodes, so the hub mechanism is not the paper's.

Balogh, Odor, and Karsai (arXiv:2602.17360) show this reverses when non-pharmaceutical interventions reshape the contact network. Social distancing measures reduce the average number of contacts per person, changing the network through which both disease and immunity spread. In sparse networks (few contacts per person), uniform vaccination is indeed most effective — it distributes immunity efficiently across the thin network.

But as average contacts increase, a reversal occurs. Naturally acquired immunity — gained through controlled exposure to disease — outperforms vaccination at achieving herd immunity. The crossover happens because vaccination distributes immunity uniformly, ignoring the network's structure, while natural infection concentrates immunity in the most-connected individuals (who are most likely to get infected). In a dense network, concentrating immunity in hubs is more efficient than distributing it uniformly, because hubs disproportionately transmit disease.

The interaction with non-pharmaceutical interventions creates the paradox. Social distancing measures that reduce contacts make the network sparse, which favors vaccination. Removing social distancing makes the network dense, which favors natural immunity. The optimal immunization strategy depends on the current state of the contact network — which is itself being shaped by policy decisions about distancing.

The structural lesson: the effectiveness of an intervention depends on the system it operates in, and that system is not static. Non-pharmaceutical interventions change the contact network, which changes which immunization strategy is optimal, which should change the immunization policy — but the immunization policy and the contact network are coupled. Optimizing one component (vaccination strategy) without accounting for the other (network-shaping interventions) can produce worse outcomes than not optimizing at all.

Balogh, Odor, & Karsai, "Non-Pharmaceutical Interventions Reshape Network Immunization Outcomes," arXiv:2602.17360 (2026).