Towards echo chamber assessment by employing aspect-based sentiment analysis and GDM consensus metrics

Echo chambers naturally occur on social networks, where individuals join groups to share and discuss their own interests driven by algorithms that steer their beliefs and behaviours based on their emotions, biases, and cognitive vulnerabilities. According to recent research on information manipulation and interference, echo chambers have become crucial weapons in the arsenal of Cognitive Warfare for amplifying the effect of psychological techniques aimed at altering information and narratives to influence public perception and shape opinions. The research is focusing on the definition of assessment methods for detecting emerging echo chambers and monitoring their evolution over time. In this sense, this work stresses the complementary role of the existing topology-based metrics and the semantics of the viewpoints underlying groups as well as their belonging users. Indeed, this paper proposes a metric based on consensus Group Decision-Making (GDM) that acquires community members’ opinions through Aspect-Based Sentiment Analysis (ABSA) and applies consensus metrics to determine the agreement within a single community and between distinct communities. The potential of the proposed metrics have been evaluated on two public datasets of tweets through comparisons with sentiment-aware opinions analysis and state-of-the-art metrics for polarization and echo chamber detection. The results reveal that topology-based metrics strictly depending on random walks over the individuals are not sufficient to fully depict the communities closeness on topics and their prevailing beliefs coming out from content analysis.