Propostas Submetidas

DEI - FCTUC
Gerado a 2024-03-29 13:37:11 (Europe/Lisbon).
Voltar

Titulo Estágio

Dynamics of Opinions or Beliefs in Social Networks

Áreas de especialidade

Sistemas de Informação

Sistemas Inteligentes

Local do Estágio

DEI-FCTUC

Enquadramento

In a social learning paradigm, each individual forms his own opinion/belief about a certain observed phenomenon (e.g., political, social, etc.) via his own rational along with integrating the opinions/beliefs of his social ties (family, friends, role models, etc.). More concretely, each agent continuously updates his individual knowledge about the observed phenomenon through the exchange of information (social interaction) with his social influences. A critical challenge in this setting is to comprehend the fundamental dynamical laws underlying the belief formation across communities of individuals. In fact, this sets the main formal ground to understand how inconsistent beliefs emerge, i.e., what structural attributes of a community could foster the emergence of fake news.

It is clear that the underlying network structure of the social ties plays critical role in the long run emergence and downfall of opinions as it entails the main avenues of information flow across individuals. In this regard, one particularly relevant paradigm for social learning is that of weakly-connected graphs/networks [1,2]: which conform to graphs wherein information can flow mostly from certain opinion-maker nodes to other follower nodes, but not (easily on) the other way around. For example, an influencer may have a large number of followers, whose individual opinions are not necessarily followed by the celebrity. Another example is that of media networks, which promote the emergence of opinions by feeding data to users without processing their feedback. Under this model, there are two categories of sub-networks: the sending sub-networks, which feed information to the receiving ones without necessarily integrating their information back. This scenario is ubiquitous over social networks.

A question of particular interest that stems from this paradigm is: what are the characterizing attributes of a community or set of communities that foster the emergence of a particular opinion? What communities or networks are more prone to the emergence of inconsistent beliefs or fake news? What kind of opinions are more likely to be dominant over a certain community?

[1] Vincenzo Matta, Augusto Santos, Ali H. Sayed, “Exponential Collapse of Social Beliefs over Weakly Connected Heterogeneous Networks”, IEEE ICASSP 2019.

[2] Vincenzo Matta, Virginia Bordignon, Augusto Santos, Ali H. Sayed, “Interplay between Topology and Social Learning over Weak Graphs”, IEEE Open Journal of Signal Processing, 2020.

Objetivo

The goal of this internship is to enhance our understanding about the fundamental laws that govern the dynamics of social learning and emergence or downfall of opinions along with the role played by the underlying network structure. We are particularly interested in characterizing the relevant network structural attributes that engender the emergence of consistent or inconsistent beliefs. Further, we would like to establish a dual relationship between certain set of ideas/beliefs with the nature of the underlying network: What ideas tend to be more virulent at a particular environment?

In order to address these questions, we explore non-Bayesian dynamical models, e.g., the distributed social learning models in [1,2,3], along with epidemical models. Besides the qualitative analysis of these models, we resort to the training of artificial neural networks to: i) ascertain the risk of the emergence of fake news in a community given the current community characteristics; ii) estimate the likelihood of an idea to become viral in a particular network. Further, we will explore the methods developed on real datasets associated with social networks (e.g., twitter).

[3] Anusha Lalitha, Tara Javidi, Anand Sarwate, “Social Learning and Distributed Hypothesis Testing”, IEEE Transactions in Information Theory, September 2018.

Plano de Trabalhos - Semestre 1

[Some tasks might overlap; M=Month]
T1 (M1-M2): State of the art analysis: Exposition to the non-Bayesian dynamical models, networked dynamical systems, Deep Learning tools applied to the social learning context.
T2 (M3-M5): Propose Deep Learning based tools for fake news risk identification .
T3 (M6): Writing the Intermediate Report.

Plano de Trabalhos - Semestre 2

[Some tasks might overlap; M=Month]
T4 (M7-M8): Implementation of the proposed methods across synthetic datasets.
T5 (M9-M10): Mining of real datasets and application of the tools developed.
T6 (M11): Writing the thesis.

Condições

Note: There is a possibility of having a 6-month scholarship within the scope of the project UNPOP https://ces.uc.pt/en/investigacao/projetos-de-investigacao/projetos-financiados/unpop-desmontar-o-populismo.

Type of scholarship: Bolsa de Investigação para Licenciado.
Value of the scholarship: 752,38€.

Observações

Please contact the advisors for any questions or clarifications needed.

Advisors: Augusto Santos (augustosantos@dei.uc.pt, DEI-CISUC) and Cristiano Gianolla (cgianolla@ces.uc.pt, CES-UC)

This is a joint collaboration between DEI-CISUC and the Centre for Social Studies, University of Coimbra.

Orientador

Augusto Santos / Cristiano Gianolla
augustosantos@dei.uc.pt 📩