This master thesis treats opinion dynamics and manipulation of mass opinions in social networks. There are several types of mathematical models, developed through the years, suitable to describe social networks dynamics and behaviours. The present work is based on a recently proposed stochastic multi-agent model which describes the interaction among individuals within social networks by means of continuous-time finite-state Markov chains. It is assumed that each individual, besides having a natural attitude to change its opinion when isolated, is influenced by its neighbors and tends to conform to their opinions according to a probabilistic model. The influence intensity is modulated by parameters that take into account the centralized curatorial function on users information diet exerted by the digital platform manager through content-aware filtering algorithms. A MATLAB simulator has been developed, in order to carry out simulations and studies upon several types of social configurations. Thanks to that simulator we are able to study different social behaviours and phenomena, including some that are already known in sociological literature, such as consensus, community cleavage and antagonism. We complement the theoretical results, worked out in previous papers, with simulation experiments that can also reveal the effect of biased influence intensity parameters, not covered by the currently available theory. Moreover, a closed-loop control system has been developed, in order to simulate phenomena of mass manipulation towards a certain opinion. The last topic is the analysis of a synthetic network made by thousands of agents representing a small community. In particular the performances, in this network, of two different strategies of opinion manipulation are discussed and compared.
Questa tesi tratta di dinamica e manipolazione delle opinioni in reti sociali. Esistono diversi modelli, sviluppati nel corso degli anni, adatti a descrivere le dinamiche e i comportamenti di reti sociali. Il modello su cui è basata questa tesi è un recente modello stocastico multi-agente che descrive l’interazione tra individui in una rete sociale attraverso catene di Markov a tempo continuo e a stati finiti. Si presume che ogni individuo, oltre alla sua naturale attitudine a cambiare opinione quando isolato, sia influenzato dai suoi vicini e tenda a conformarsi alle loro opinioni secondo un modello probabilistico. L’intensità di questa influenza è modulata da parametri che prendono in considerazione l’azione di controllo sulle informazioni percepite dagli utenti, esercitata dal manager della piattaforma, attraverso algoritmi filtranti dei contenuti. Un simulatore è stato sviluppato in MATLAB per poter portare avanti studi e analisi su diversi tipi di configurazioni sociali. Grazie al simulatore siamo in grado di studiare diversi tipi di comportamenti e fenomeni sociali, alcuni dei quali già conosciuti in letteratura, come ad esempio il consenso, l’antagonismo e il community cleavage. Siamo inoltre in grado di ampliare i risultati teorici, portati avanti in lavori precedenti, con simulazioni sperimentali che possono mostrare l’effetto, non mostrato dalle attuali teorie, di parametri di diffusione delle opinioni polarizzati. E' stato inoltre sviluppato un controllo in anello chiuso in grado di simulare fenomeni di manipolazione delle masse verso una specifica opinione. Come ultimo argomento, viene trattata e analizzata una rete sintetica formata da migliaia di individui in grado di simulare una piccola comunità. In particolare si discutono e confrontano due diverse strategie di manipolazione delle opinioni all’interno di tali reti.
Opinion dynamics and manipulation in social networks : simulations with a stochastic multi-agent model. Dinamica e manipolazione delle opinioni in reti sociali : simulazioni con un modello stocastico multi-agente
SECCHI, MARCO
2018/2019
Abstract
This master thesis treats opinion dynamics and manipulation of mass opinions in social networks. There are several types of mathematical models, developed through the years, suitable to describe social networks dynamics and behaviours. The present work is based on a recently proposed stochastic multi-agent model which describes the interaction among individuals within social networks by means of continuous-time finite-state Markov chains. It is assumed that each individual, besides having a natural attitude to change its opinion when isolated, is influenced by its neighbors and tends to conform to their opinions according to a probabilistic model. The influence intensity is modulated by parameters that take into account the centralized curatorial function on users information diet exerted by the digital platform manager through content-aware filtering algorithms. A MATLAB simulator has been developed, in order to carry out simulations and studies upon several types of social configurations. Thanks to that simulator we are able to study different social behaviours and phenomena, including some that are already known in sociological literature, such as consensus, community cleavage and antagonism. We complement the theoretical results, worked out in previous papers, with simulation experiments that can also reveal the effect of biased influence intensity parameters, not covered by the currently available theory. Moreover, a closed-loop control system has been developed, in order to simulate phenomena of mass manipulation towards a certain opinion. The last topic is the analysis of a synthetic network made by thousands of agents representing a small community. In particular the performances, in this network, of two different strategies of opinion manipulation are discussed and compared.File | Dimensione | Formato | |
---|---|---|---|
2019_10_Secchi.pdf
accessibile in internet per tutti
Descrizione: Testo della tesi
Dimensione
2.8 MB
Formato
Adobe PDF
|
2.8 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/150523