PORTALE DELLA DIDATTICA

Ricerca CERCA
  KEYWORD

INFLUENCE MAXIMIZATION AND RESILIENCE IN OPINION DYNAMICS

Parole chiave INFLUENCE, OPINION DYNAMICS, POLARIZATION, RESILIENCE

Riferimenti GIACOMO COMO, FABIO FAGNANI

Gruppi di ricerca Analisi e controllo di sistemi dinamici

Descrizione Opinion dynamics is a family of mathematical models [1] that have been proposed over the years to study the formation of opinions in small and large communities and to analyze how certain global phenomena like consensus or polarization can show up. A popular instance is the so called voter model where dynamics consists of imitation steps where a player copies the 'opinion' of another player among those he interacts with. In this case the state is typically binary and model situations where we are considering the opinion on a fact that can either be true or not or rather a preference with respect to two alternatives. A different model goes back to the social science literature of half a century ago and is known as the French-De Groot averaging model. In this case the opinion is treated as a continuous variable and the agents keep modifying it by taking averages of the opinion of the other agents they interact with. Under mild assumptions all these dynamics converge to a consensus configuration where all agents share the same opinion [2]. A modification of both the voter and the French-De Groot model foresees the presence of stubborn agents [3,4], namely agents that do not modify their opinion as time passes by. Stubborn agents may model media or influencers in a social network. In this case the model does not typically converge to a consensus and much of the interest is in understanding how the new equilibrium depends on the topology of the network and on the displacement of the stubborn agents. New phenomena of polarization can take place in this context. Two possible research areas are:

- Intervention problems where an external planner can either add or delete edges in order to increase bias towards a particular opinion or decrease polarization. Previous works to be read are [5,6,7];

- Analysis of mixed models where part of the agents are following either the voter or the French-De Groot dynamics while the others are strategic in their choices (i.e., they maximize a utility function). The insertion of strategic agents could be seen as an intervention problem with the goal of increasing the resilience of the network. The exact formulation is part of the thesis work.



[1] M. Grabisch and A. Rusinowska,"A survey on nonstrategic models of opinion dynamics," Games, 11 (4), 2020.

[2] G. Como and F. Fagnani, "From local averaging to emergent global behaviors: The fundamental role of network interconnections," Systems and Control Letters, 95, pp. 70-76, 2016.
[3] D. Acemoglu, G. Como, F. Fagnani, and A. Ozdaglar, "Opinion fluctuations and disagreement in social networks," Mathematics of Operations Research, 38 (1), pp. 1-27, 2013.
[4] E. Yildiz, A. Ozdaglar, D. Acemoglu, A. Saberi, and A. Scaglione, "Binary opinion dynamics with stubborn agents," ACM Transactions on Economics and Computation, 1 (4) pp. 19-30, 2013.
[5] G. Como and F. Fagnani, "Robustness of large-scale stochastic matrices to localized perturbation," IEEE Transactions on Network Science and Engineering, 2 (2), pp. 53-64, 2015.

[6] A. Gionis M. Mathioudakis K. Garimella, G. De Francisci~Morales, "Reducing controversy by connecting opposing views," 27th International Joint Conference on Artificial Intelligence (IJCAI-18), pp. 5249-5253, 2018

[7] V. Amelkin and A. K. Singh, "Fighting opinion control in social networks via link recommendation," 25'th ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD), 2019.


Scadenza validita proposta 18/04/2022      PROPONI LA TUA CANDIDATURA