Today almost every aspect of our existence is ruled by machine learning systems, which autonomously learn from data and examples provided in order to generate predictions. The issue emerges when user's data profiled are used to filter on the basis of his interests the content to show. The user is trapped inside a bubble of information coherent with his own vision of reality, this could affect the formation of public debate and the propagation of disinformation. Deep learning is considered the most advanced branch of machine learning. Deepfake content, produced through deep learning technology, has the characteristic of appearing extremely realistic and therefore capable of deceiving the recipient, who is unable to recognize the artificiality of the content. This thesis aims to analyze the impacts and links between filter bubbles, disinformation and deepfake, as well as the main approaches adopted by platforms and institutions to fight the phenomenon as a whole.
Ad oggi quasi ogni aspetto della nostra esistenza è governato da sistemi di machine learning in grado di apprendere autonomamente dai dati e dagli esempi forniti per produrre previsioni. La criticità emerge nel momento in cui i dati dell’utente profilato vengono impiegati per filtrare sulla base dei suoi interessi i contenuti da mostrare. L’utente si ritrova confinato all’interno di una bolla di informazioni coerenti con la propria visione della realtà in grado di incidere sulla formazione del dibattito pubblico e sulla propagazione di disinformazione. Il deep learning rappresenta il ramo più avanzato del machine learning. I contenuti deepfake, prodotti mediante deep learning, hanno la caratteristica di apparire estremamente realistici e dunque idonei ad ingannare il destinatario che non è in grado di riconoscere l’artificialità del contenuto. Il presente elaborato intende analizzare gli impatti e il legame tra bolle filtro, disinformazione e deepfake, oltre che analizzare i principali approcci adottati da piattaforme e istituzioni per il contrasto del fenomeno nel suo complesso.
Dalle filter bubbles ai deepfake: nuove forme di manipolazione della realtà. Un'analisi degli impatti e delle misure per il contrasto
BALDO, SARA
2023/2024
Abstract
Today almost every aspect of our existence is ruled by machine learning systems, which autonomously learn from data and examples provided in order to generate predictions. The issue emerges when user's data profiled are used to filter on the basis of his interests the content to show. The user is trapped inside a bubble of information coherent with his own vision of reality, this could affect the formation of public debate and the propagation of disinformation. Deep learning is considered the most advanced branch of machine learning. Deepfake content, produced through deep learning technology, has the characteristic of appearing extremely realistic and therefore capable of deceiving the recipient, who is unable to recognize the artificiality of the content. This thesis aims to analyze the impacts and links between filter bubbles, disinformation and deepfake, as well as the main approaches adopted by platforms and institutions to fight the phenomenon as a whole.File | Dimensione | Formato | |
---|---|---|---|
Baldo_Sara.pdf
accesso aperto
Dimensione
1.5 MB
Formato
Adobe PDF
|
1.5 MB | Adobe PDF | Visualizza/Apri |
The text of this website © Università degli studi di Padova. Full Text are published under a non-exclusive license. Metadata are under a CC0 License
https://hdl.handle.net/20.500.12608/67057