: ..,
..,
..
: :
: 121
: -
: 2026
: .., .., .. : // . - 2026. - . 121. - .227-246.
: , -, SCARDO-, , , ,
(.): coordinated influence, AI-agents, SCARDO model, mean-field approximation, optimal control, modular networks, ranking algorithms
: , . SCARDO- , , , . , , , . , , , "" . , 30% , , , .
(.): This paper concerns the problem of influencing agent opinions through coordinated informational attack by social bots embedded in a social network. Our mathematical framework is based on a mean-field approximation for the SCARDO model and regards the heterogeneity of authentic agents, the modularity of the social network, and penalties for excessive use of bots imposed by the platform itself. Computational experiments show that ranking algorithms that minimize interactions between similar agents facilitate the achievement of the Botmaster's goal. Thus, to improve the stability of a social network, one may use ranking algorithms that, by contrast, promote the formation of information cocoons, which act as a "shield" against peer influences. We also determine the optimal number of bots ensuring the best value of the objective functional - approximately 30 percent of the total number of agents. This finding is consistent with empirical estimates of the number of bots on real online platforms, as well as analytical estimates of the effective number of bots optimizing cooperative behavior in the population.
PDF
: 126, : 57, : 29.