Social media are pervasive in our daily lives. These platforms are gaining momentum thanks to powerful artificial intelligence models. Serious negative consequences of social media have been repeatedly highlighted in recent years, pointing at various threats to society and its more vulnerable members, such as teenagers, in particular, ranging from much-discussed problems such as digital addiction and polarization to manipulative influences of algorithms and further to more teenager-specific issues. The thesis proposes a collective well-being-oriented social media companion by examining the background and context surrounding these issues, such as problems related to the AI-human value alignment and how to deploy Educationally managed social media and well-being metrics that are flexible enough to take into account specific needs and the coexistence of different stakeholders such as teacher and content creators. The research includes agent-based simulation to examine the impact of content and people recommendation strategies on opinion dynamics, otherwise invisible network-based threats directly into the user feed, such as polarisation and echo chambers. A deep learning architecture based on a graph neural network for classifying tweets is introduced, and a dataset annotation protocol that combines objective and subjective features to understand factors affecting the potential misleadingness of images because methods for addressing social media threats must support multimodal content. Overall, the results indicate that machine learning models can serve as detectors for labelling content in the proposed companion and that there is room left for including additional types of information (e.g. social media context) given the flexibility of the proposed architecture. The game-based digital media literacy educational intervention inspired by the wisdom of crowds phenomenon can increase the perception of social media's influence on participants. This contribution is motivated by the desire to improve the impact of social media on our society. The presence of a trade-off between users' rights and duties or freedom VS safety introduces ethical issues that require the formulation of a comprehensive and shared view of the values of the social media community, so more multidisciplinary work is still required.
I social media sono pervasivi nella nostra vita quotidiana. Queste piattaforme stanno guadagnando slancio grazie a potenti modelli di intelligenza artificiale. Le gravi conseguenze negative dei social media sono state ripetutamente evidenziate negli ultimi anni, indicando varie minacce per la società e i suoi membri più vulnerabili, come gli adolescenti, in particolare, che vanno da problemi molto discussi come la dipendenza digitale e la polarizzazione alle influenze manipolative degli algoritmi e oltre a questioni più specifiche per gli adolescenti. La tesi propone un compagno di social media orientato al benessere collettivo esaminando lo sfondo e il contesto che circonda questi problemi, come i problemi relativi all'allineamento AI-valore umano e come distribuire social media gestiti in modo educativo e metriche di benessere flessibili sufficiente per tener conto delle esigenze specifiche e della coesistenza di diverse parti interessate come insegnanti e creatori di contenuti. La ricerca include la simulazione basata su agenti per esaminare l'impatto delle strategie di raccomandazione di contenuti e persone sulle dinamiche di opinione, minacce altrimenti invisibili basate sulla rete direttamente nel feed degli utenti, come la polarizzazione e le camere di eco. Viene introdotta un'architettura di deep learning basata su una rete neurale a grafo per la classificazione dei tweet e un protocollo di annotazione del set di dati che combina caratteristiche oggettive e soggettive per comprendere i fattori che influenzano la potenziale falsità delle immagini perché i metodi per affrontare le minacce dei social media devono supportare il contenuto multimodale. Nel complesso, i risultati indicano che i modelli di apprendimento automatico possono fungere da rilevatori per etichettare i contenuti nel compagno proposto e che c'è ancora spazio per includere ulteriori tipi di informazioni (ad esempio il contesto dei social media) data la flessibilità dell'architettura proposta. L'intervento educativo di alfabetizzazione mediatica digitale basato sul gioco, ispirato al fenomeno della saggezza della folla, può aumentare la percezione dell'influenza dei social media sui partecipanti. Questo contributo è motivato dal desiderio di migliorare l'impatto dei social media sulla nostra società. La presenza di un compromesso tra diritti e doveri degli utenti o libertà VS sicurezza introduce questioni etiche che richiedono la formulazione di una visione completa e condivisa dei valori della comunità dei social media, quindi è ancora necessario un lavoro più multidisciplinare.
(2023). Raising Teenagers' Awareness of Social Media Threats: A Theoretical and Empirical Study. (Tesi di dottorato, Università degli Studi di Milano-Bicocca, 2023).
Raising Teenagers' Awareness of Social Media Threats: A Theoretical and Empirical Study
LOMONACO, FRANCESCO
2023
Abstract
Social media are pervasive in our daily lives. These platforms are gaining momentum thanks to powerful artificial intelligence models. Serious negative consequences of social media have been repeatedly highlighted in recent years, pointing at various threats to society and its more vulnerable members, such as teenagers, in particular, ranging from much-discussed problems such as digital addiction and polarization to manipulative influences of algorithms and further to more teenager-specific issues. The thesis proposes a collective well-being-oriented social media companion by examining the background and context surrounding these issues, such as problems related to the AI-human value alignment and how to deploy Educationally managed social media and well-being metrics that are flexible enough to take into account specific needs and the coexistence of different stakeholders such as teacher and content creators. The research includes agent-based simulation to examine the impact of content and people recommendation strategies on opinion dynamics, otherwise invisible network-based threats directly into the user feed, such as polarisation and echo chambers. A deep learning architecture based on a graph neural network for classifying tweets is introduced, and a dataset annotation protocol that combines objective and subjective features to understand factors affecting the potential misleadingness of images because methods for addressing social media threats must support multimodal content. Overall, the results indicate that machine learning models can serve as detectors for labelling content in the proposed companion and that there is room left for including additional types of information (e.g. social media context) given the flexibility of the proposed architecture. The game-based digital media literacy educational intervention inspired by the wisdom of crowds phenomenon can increase the perception of social media's influence on participants. This contribution is motivated by the desire to improve the impact of social media on our society. The presence of a trade-off between users' rights and duties or freedom VS safety introduces ethical issues that require the formulation of a comprehensive and shared view of the values of the social media community, so more multidisciplinary work is still required.File | Dimensione | Formato | |
---|---|---|---|
phd_unimib_876570.pdf
accesso aperto
Descrizione: Tesi di "Lomonaco" "Francesco" - "876570"
Tipologia di allegato:
Doctoral thesis
Dimensione
11.12 MB
Formato
Adobe PDF
|
11.12 MB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.