[GCd18a] Reducing Filter Bubbles With a Community Aware Model

Conférence Nationale avec comité de lecture : BDA, October 2018, Vol. 34, pp.1-10, Bucarest, Roumania,

Auteurs: Q. Grossetti , C. Constantin , C. du Mouza , N. Travers

Mots clés: Twitter, Communities, Filter Bubble, Recommender System, Collaborative Filtering, Microblogging Systems, Echo-Chamber

Résumé: After facing tremendous growth, main social network platforms took an important place in modern societies. In order to increase user engagement, they rely heavily on recommender systems and fears arise that personalizing content might trap users into a filter bubble exacerbating divisions within societies. After a thorough study of the detected communities in a large Twitter dataset and their role in information propagation, we present a metric to compute the strength of the filter bubble. Our results show that filter bubble effects are in fact limited for a majority of users. We find that, counterintuitively, in most cases recommender systems tend to open users perspectives. However, for some specific users, the bubble effect is noticeable and we propose a model relying on communities to provide a list of recommendations closer to the userâ€™s usage of the platform.

Collaboration: LIP6

BibTeX

@inproceedings {
	GCd18a,
	title	=	"{Reducing Filter Bubbles With a Community Aware Model}",
	author	=	" Q. Grossetti and C. Constantin and C. du Mouza and N. Travers ",
	booktitle	=	"{BDA}",
	year	=	2018,
	month	=	"October",
	volume	=	34,
	pages	=	"1-10",
	address	=	"Bucarest, Roumania",
}