This study delves into the phenomenon of coordinated inauthentic behaviour (CIB) on Facebook and its consequential impact on the dissemination of misinformation. It underscores the pressing need to tackle misinformation and preserve the integrity of online discourse. The objective of this research is to distinguish between individual posts on Facebook pages and categorize them as instances of CIB or not. To achieve this, we utilized a dataset comprising Facebook posts collected between 2017 and 2019, coupled with the implementation of various supervised machine learning models, including Logistic Regression, K-Nearest Neighbors, Naive Bayes, Random Forest, and a single-layer Neural Network, along with a range of vectorization techniques for analysis, Bag-of-Words, Term Frequency-Inverse Document Frequency and Word2Vec embeddings. The findings provide insights into the detection of CIB, revealing nonlinear patterns and distinguishable relationships within the networks, making individual detection feasible. In conclusion, the study contributes to the comprehension of CIB on social media platforms emphasising the importance of addressing misinformation to maintain the integrity of online engagements.
Questo studio esplora il fenomeno del comportamento inautentico coordinato (CIB) su Facebook e il suo impatto sulla diffusione della disinformazione. Sottolinea la necessità pressante di affrontare la disinformazione e preservare l'integrità digitale. L'obiettivo di questa ricerca è distinguere tra i singoli post delle diverse pagine di Facebook quali categorizzare come casi di CIB e quali no. Per raggiungere questo obiettivo, utilizziamo un dataset composto da post di Facebook raccolti nell'arco di tre anni, associati all'implementazione di vari modelli di apprendimento intelligente, tra cui Logistic Regression, K-Nearest Neighbors, Naive Bayes, Random Forest, e un Single-layer Neural Network, insieme a una gamma di tecniche di vettorizzazione per l'analisi. I risultati forniscono spunti sulla rilevazione del CIB, rivelando pattern non lineari e relazioni distinguibili all'interno delle reti, rendendo la rilevazione individuale fattibile. In conclusione, lo studio contribuisce alla comprensione del CIB sulle piattaforme di social media e sottolinea l'importanza di affrontare la disinformazione per preservare l'integrità digitale.
Classifying coordinated inauthentic behaviour on Facebook
LILJA, KRISTIN LIV JULIA ULFSDOTTER
2023/2024
Abstract
This study delves into the phenomenon of coordinated inauthentic behaviour (CIB) on Facebook and its consequential impact on the dissemination of misinformation. It underscores the pressing need to tackle misinformation and preserve the integrity of online discourse. The objective of this research is to distinguish between individual posts on Facebook pages and categorize them as instances of CIB or not. To achieve this, we utilized a dataset comprising Facebook posts collected between 2017 and 2019, coupled with the implementation of various supervised machine learning models, including Logistic Regression, K-Nearest Neighbors, Naive Bayes, Random Forest, and a single-layer Neural Network, along with a range of vectorization techniques for analysis, Bag-of-Words, Term Frequency-Inverse Document Frequency and Word2Vec embeddings. The findings provide insights into the detection of CIB, revealing nonlinear patterns and distinguishable relationships within the networks, making individual detection feasible. In conclusion, the study contributes to the comprehension of CIB on social media platforms emphasising the importance of addressing misinformation to maintain the integrity of online engagements.File | Dimensione | Formato | |
---|---|---|---|
2024_04_Lilja.pdf
non accessibile
Descrizione: Text of the thesis.
Dimensione
1.8 MB
Formato
Adobe PDF
|
1.8 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/10589/218924