King's College London Digital Humanities

Event organised by the Computational Humanities research group.

To register to the seminar, please fill in this form by 28 February 2025.

4 March 2025 – 4pm GMT

Remote – Via Microsoft Teams.

In person – King’s College London, Bush House (SE), 1.10.

Brian Ball (Northeastern University London), Structure Meets Strategy in the Misinformation Age: a Simulation-Based Study

Abstract

We are now living in what some have called ‘the misinformation age’ (O’Connor and Weatherall, 2019), in which AI algorithms and social networks determine the quality of the information we can access – with a range of important consequences (e.g. for democracy). Previous work has explored the effects of different information processing strategies on the abilities of communities of rational agents to discover the truth in a timely manner (Ball et al., 2024). The present paper uses the PolyGraphs simulation framework to (computationally) investigate how network structure interacts with, and impacts upon, the effectiveness of these strategies.

We begin (in section 1) with a thorough investigation of artificial networks generated using the preferential attachment model (Barabasi and Albert, 1999) on which existing network nodes (agents) with more connections are more likely to be linked, via an information-sharing edge, to new nodes joining the network. We find (amongst other things) that the ill-effects of mis- and disinformation on the efficiency of truth-seeking inquiry are even stronger in these relatively sparse (i.e. low density) networks than in the complete networks that were explored previously; and we detect hints that concerns about accuracy extend to larger networks than previously thought (Zollman, 2007). We then (in section 2) pursue a more systematic investigation of a range of network structures, including random networks (Erdos and Renyi, 1959), as well as those that exhibit the so-called ‘small-world’ property (Watts and Strogatz, 1998), using advanced statistical techniques to tease out the relative influences of a range of structural features, including density, clustering, and path length. Finally, (in section 3) we look at a pair of larger networks representing real-world communities/populations.

Bio

Brian Ball is Head of Faculty in Philosophy at Northeastern University London. A Senior Fellow of the Higher Education Academy, he was previously a Lecturer in Philosophy at St. Anne’s and then Balliol College, Oxford. His expertise is in the theory of knowledge, the philosophy of mind, and the philosophy of language. His recent work engages with computer science and artificial intelligence.

Event organised by the Computational Humanities research group.

To register to the seminar, please fill in this form by 23 February 2025.

25 February 2025 – 3pm GMT

Remote – Via Microsoft Teams.

Vjosa Preniqi (Queen Mary University of London), Evolution of Moral Expressions in Music: Applying MoralBERT to Large-Scale Lyrics Datasets

Abstract

Lyrics serve as more than emotional expressions; they act as a powerful medium for reflecting and influencing societal, political, and cultural landscapes. From addressing racial inequality and gender discrimination to mirroring political unrest, song lyrics have historically provided a lens into the social issues of their time. Recent analyses, such as the rising gender bias in lyrics from 1960 to 2010 among male artists on Billboard charts, reveal concerning trends and ethical implications, including the promotion of violence, misogyny, and substance abuse. These findings underscore the importance of understanding the broader moral and cultural messages conveyed in music. In this presentation, I explore the application of MoralBERT, a tool that we built to detect and analyse moral foundations in English song lyrics. By leveraging the extensive WASABI dataset (1960–2010), the presentation will introduce a study that was conducted to examine the evolution of moral expression across genres while highlighting themes such as gender perspectives and societal narratives in music. The findings demonstrate how advanced natural language processing (NLP) models can uncover underlying moral dimensions in lyrics, offering insights into their societal impacts and implications for media literacy.

Bio

Vjosa Preniqi is a PhD Candidate at Queen Mary University of London, based in the Centre for Digital Music (C4DM) within the School of Electronic Engineering and Computer Science. Her research explores the connections between moral values, music content, and listening behaviours, combining insights from data science, AI, musicology and psychology. Passionate about interdisciplinary research, she bridges the gap between data science and the arts to better understand the cultural and emotional impact of music.