Zombie Politics: Evolutionary Algorithms to Counteract the Spread of Negative Opinions

This paper is about simulating the spread of opinions in a society and about finding ways to counteract that spread. To abstract away from potentially emotionally laden opinions, we instead simulate the spread of a zombie outbreak in a society. The virus causing this outbreak is different from traditional approaches: it not only causes a binary outcome (healthy vs infected) but rather a continuous outcome. To counteract the outbreak, a discrete number of infection-level specific treatments is available. This corresponds to acts of mild persuasion or the threats of legal action in the opinion spreading use case. This paper offers a genetic and a cultural algorithm that find the optimal mixture of treatments during the run of the simulation. They are assessed in a number of different scenarios. It is shown, that albeit far from being perfect, the cultural algorithm delivers superior performance at lower computational expense.

💡 Research Summary

The paper “Zombie Politics: Evolutionary Algorithms to Counteract the Spread of Negative Opinions” presents a novel abstraction of opinion dynamics by modeling the diffusion of harmful or radical viewpoints as a zombie‑style epidemic. Unlike classic epidemiological models that treat infection as a binary state, the authors introduce a continuous infection variable (the C‑Virus) ranging from 0 (completely healthy) to 1 (fully infected). An individual becomes a “zombie” – i.e., an active spreader – only when this variable exceeds a predefined threshold (0.75). This continuous formulation mirrors real‑world opinion propagation, where most people hold a latent degree of agreement with an extreme view, but only a subset actively disseminates it.

To counteract the outbreak, the authors define a discrete set of ten treatment types, each effective only within a specific infection‑level band. In practice these treatments correspond to calibrated interventions such as mild persuasion, public statements, or legal warnings. A daily treatment plan is therefore a 10‑dimensional vector of probabilities (values in (0,1)) indicating how often each treatment is applied to the population. The central research problem is to discover, at each simulation step, the optimal mixture of treatments that minimizes the mean infection rate (mir) while prolonging the survival of healthy individuals (measured by the last day a human is encountered, lHd).

Because the epidemic evolves continuously, the optimization problem is dynamic: the fitness landscape changes every day as the infection spreads and the treatment is applied. The authors cast this as a “dynamic continuous disease treatment problem,” analogous to a dynamic knapsack where the capacity and item values shift over time. They evaluate two evolutionary strategies: a conventional Genetic Algorithm (GA) and a Cultural Algorithm (CA).

The GA encodes a candidate treatment mixture as a chromosome of ten real‑valued genes. Standard operators are used: crossover produces offspring by averaging parent genes, mutation replaces a gene with a uniformly drawn value, and tournament selection (size 5) chooses parents. The best individual is cloned each generation to enforce monotonic improvement.

The CA builds on the GA but adds a belief space (meta‑storage). After each generation the fittest individual is stored in this belief space. When generating offspring, one parent is randomly replaced by the belief‑space solution, thereby injecting historically successful knowledge into the population. This mechanism is intended to accelerate convergence and avoid premature stagnation, especially when the number of generations per day (gd) is limited.

Experiments are conducted on a well‑mixed population of 100 agents, initialized with random infection levels. Two virulence settings are examined: low (v = 0.3) where contact with a zombie leads to infection with probability 0.3, and high (v = 1) where every contact results in infection. For each virulence, three gd values (1, 5, 25 generations per day) are tested, and each configuration is replicated ten times. Performance is assessed via mean infection rate (mir) and last human day (lHd), with a baseline simulation that applies no treatment.

Results show that the CA consistently outperforms the GA across all scenarios. In low‑virulence settings the advantage is modest, but under high virulence the CA achieves markedly lower mir and significantly longer lHd, even when only a single generation per day is allowed. Moreover, the CA requires roughly 30 % less computational time than the GA, demonstrating its efficiency. The authors attribute the CA’s superiority to the belief‑space mechanism, which preserves high‑quality solutions and guides the search despite the rapidly changing fitness landscape.

The paper acknowledges several limitations. The underlying network is assumed to be a fully mixed mean‑field, ignoring realistic small‑world or scale‑free structures that heavily influence epidemic spread. Treatment effectiveness is constrained to ±10 % around a nominal value, and the number of treatment types is fixed at ten. Real opinion dynamics involve multiple competing narratives, temporal delays, and heterogeneous influence patterns, none of which are captured in the current model. Consequently, the findings serve primarily as a proof‑of‑concept for dynamic evolutionary optimization rather than a ready‑to‑deploy policy tool.

In conclusion, the study demonstrates that evolutionary computation—particularly cultural algorithms that incorporate historical knowledge—can effectively navigate dynamic, continuous‑state epidemic models and identify near‑optimal intervention mixes under tight computational budgets. Future work should extend the model to realistic social network topologies, incorporate multi‑opinion interactions, and validate the approach against empirical data from online platforms or real‑world political campaigns. Such extensions would bridge the gap between the elegant theoretical framework presented here and actionable strategies for mitigating the spread of harmful opinions in modern societies.