The Effects of Algorithmic Bias on Political Discourse

Abstract

It is no secret that in everyday life, politics involves winner and loser communication. People on the left and right fight for attention over the online events. Over the past decade, web-based platforms have become crucial to political communication, providing citizens with different views on shared news and events posted in real-time. In order to automate such intricate tasks, ranking algorithms determine locally relevant content based on user requests. As a result, populated filter bubbles systematically distort user experiences and shape impressions on the political events. Here it is demonstrated that rising liberal and conservative concerns about increasing levels of perceived a priori and emergent algorithmic bias in the GDELT GKG dataset are not unfounded. By successfully engaging in story entry and forming strategies, it is shown that, compared with non-political results, large media sources and well-known information sources are significantly impacted by linguistic and news attention biases, and political bias biases emerge through the direct manipulation of inherent bias. Greater PPL values further insulate traditional media sources due to the symbolic nature of their coverage, enclosing new media sources in a viscous circle of liberal disadvantage. To ensure fair and balanced coverage of the event, the practices recommendations advocate the adoption of a neutral story input when ranking politically disputed events in open news retrieval competition task. Algorithmic decisions increasingly mediate fundamental aspects of everyday lives. It is not exaggeration to say that every unfolding public event is performed on a complex network of various technologies. The ever-growing amount of web-based platforms has become crucial to setting the media agenda and therefore the framing of political communication. There, a wealth of different articles, opinions, and news created and shared by common people can be accessed in real-time. In a series of ongoing efforts to augment the technology implementation, the EU Commission is negotiating a digital services act package aimed at modernizing the existing independent legal framework of what defines a non-neutral information system service. Algorithmic bias seems here an unavoidable byproduct. In the worst case, artificial intelligence systems may come to dictate the deployment and possibly the outcome of every unfolding event.

Keywords: Algorithmic Bias, Political Discourse, Filter Bubbles, Online Communication, Media Agenda, News Retrieval, Digital Services, Content Ranking