A Survey on Computational Propaganda Detection (2007.08024v1)

Abstract: Propaganda campaigns aim at influencing people's mindset with the purpose of advancing a specific agenda. They exploit the anonymity of the Internet, the micro-profiling ability of social networks, and the ease of automatically creating and managing coordinated networks of accounts, to reach millions of social network users with persuasive messages, specifically targeted to topics each individual user is sensitive to, and ultimately influencing the outcome on a targeted issue. In this survey, we review the state of the art on computational propaganda detection from the perspective of Natural Language Processing and Network Analysis, arguing about the need for combined efforts between these communities. We further discuss current challenges and future research directions.

Overview of Computational Propaganda Detection

The paper, "A Survey on Computational Propaganda Detection," provides an extensive analysis of the current state of research concerning the identification and mitigation of computational propaganda, focusing on contributions from NLP and Network Analysis. Computational propaganda, which involves using technical means to propagate persuasive information, often in social media, is a phenomenon with serious societal implications. The paper emphasizes the need for an interdisciplinary approach to tackle this challenge effectively.

Propaganda and Its Computational Aspects

Propaganda, historically entrenched as a method of mass persuasion, leverages various psychological and rhetorical techniques to influence public opinion. Computational propaganda represents an evolution, enabling widespread automated dissemination via social networks, often utilizing coordinated entities like botnets and troll armies. This paper delineates these entities and their roles in expanding the scale of influence.

Key Challenges in Detection

Detecting computational propaganda is fraught with challenges, notably the scarcity of comprehensive datasets that adequately capture the nuanced nature of propaganda. Existing datasets often suffer from biases due to distant supervision, making it difficult to discern actual propaganda techniques from broader media biases. This limitation implies that the development of finer-grained techniques, such as those that identify specific propaganda tactics within text, may be more promising.

Text Analysis Perspective

In exploring text analysis methodologies, the paper highlights the progression from document-level classification datasets like TSHP-17 and QProp to more granular annotation efforts seen in the PTC corpus. The latter focuses on identifying specific propaganda techniques at the fragment level, presenting a more precise framework for training models that detect manipulation tactics within texts.

Network Analysis Perspective

From a network analysis standpoint, the evolution of malicious actors requires the development of mechanisms that go beyond profiling individual accounts to assessing coordinated group activities. Early approaches focused on node classification; however, contemporary strategies aim to detect anomalous patterns in connectivity, revealing inauthentic and coordinated behaviors at scale. Unsupervised and semi-supervised models are increasingly favored to overcome the evolving tactics of adversaries who deploy sophisticated AI for evasion.

Interdisciplinary Approaches and Future Prospects

The paper advocates for integrating NLP and network analysis to evolve computational propaganda detection. Upcoming trends may include AI-generated propaganda, necessitating detectors capable of cross-modal analyses where textual, visual, and behavioral data are examined holistically. Adopting adversarial machine learning to anticipate adversary evolution and addressing ethical concerns relating to data privacy and model transparency are pivotal for advancing this field.

In sum, while substantive progress has been made in computational propaganda detection, the persistent evolution of methodologies and adversarial tactics underscores the importance of continued interdisciplinary collaboration and innovation in this field.