Artificial Intelligence, Machine Learning, and Analytics for Social Media

Permanent URI for this collectionhttps://hdl.handle.net/10125/112445

Browse

Now showing 1 - 6 of 6

The Color of Engagement: How Visual Hue, Hedonic Cues, and Audio Arousal Shape Fan Attention in Short-Form Sports Videos
(2026-01-06) Yoo, Taewoong; Dasari, Prasanna; Chang, Yonghwan
We analyze 100 fan-generated TikTok videos from the 2025 NCAA Men’s Basketball Tournament to examine how visual hue, branding, hedonic cues, and audio arousal relate to engagement. Using computer vision and audio signal processing, we test theory-driven hypotheses within a model framework. Results show directional evidence for a U-shaped association between hue and engagement; this pattern appears stronger when hedonic cues are present and weaker under high audio arousal, consistent with accounts of affective amplification and multimodal overload. Given the small sample and measurement limits, we treat non-significant effects as exploratory. We contribute an ecologically grounded approach for studying multimodal attention in short-form sports media and offer practical guidance: design for sensory balance (color intensity with moderated audio) rather than saturation. Future work should scale data, incorporate saliency-weighted and temporal visual features, and examine algorithmic mediation.
A Comprehensive Study on the Role of Mental Health Advocates on Instagram and Their Effects on User Engagement
(2026-01-06) Yazdizadeh, Honeyeh; Farivar, Samira
This paper examines the role of mental health advocates on Instagram, focusing on how professional status and content type impact engagement. It analyzes a large sample of posts created by licensed professionals (e.g., psychologists) and unlicensed influencers (e.g., wellness coaches, advocates), including branded and non-branded content. Using Natural Language Processing (NLP) and Top2Vec Topic Modelling, the study identifies key mental health-related themes and examines which terms and topics are more frequently linked to higher engagement metrics like likes and comments. The analysis reveals differences in engagement patterns based on advocate credentials and whether content is commercially affiliated. Terms related to anxiety, depression, and trauma are more prominent among high-engagement content. Licensed professionals emphasize clinical support, while unlicensed influencers frequently use motivational or emotionally resonant language. These findings offer insights for mental health communicators, brands, and platform designers aiming to foster more credible and engaging mental health discourse on social media.
An Agent-Based Model to Simulate Individual Reframing of News Media Posts on Social Media
(2026-01-06) Shao, Lumin; Bernardi, Roberta; Zhang, Jie
News frames have long been considered a powerful device for shaping people’s understanding of news issues. Recent studies have shown that social media users actively reframe media content. However, there is a limited understanding of the dynamics through which social media users reframe news over time. To address this research gap, this study develops an agent-based model to simulate individual reframing behaviour on social media, exploring the role of sentiment in the individual reframing of news posts. In the model, we design novel heterogeneous agents (innovative agent and persuadable agents), develop semantic distance-based rules to agents’ confidence in external information. The agent-based model is validated by comparing the simulation results with real data (discussions about COVID-19 vaccine) collected from Weibo. Finally, in the discussion section, we analyse the simulation results, highlight the contributions and limitations of this study, and propose avenues for future research.
Examining the Efficacy of Multi-Theoretical Social Science-informed Deep Learning Models in Predicting Mob Outcomes
(2026-01-06) Al-Khateeb, Samer; Burright, Jack; Brainard, James; Murray, Rebecca; Agarwal, Nitin
Organized and coordinated events, whether conducted in physical spaces (e.g., fun flash mobs, parkour, or even deviant mobs), cyberspace (e.g., collective hacking, organized propaganda campaigns), or in both spaces (e.g., electronic to face-to-face (e2f) events), represent various forms of collective action aimed at improving a group's status or condition and usually to achieve a common goal. Understanding such events requires a combination of technological and sociological approaches due to the complexity of the relationships that could exist, form, and dissolve among participating individuals. In this research, we integrate our knowledge of five social science theories that can explain such events with technical skills. We use this combination to estimate theoretical factors (both event-related and individual-related) using data collected from Meetup.com. We then train four classifiers using a deep neural network to predict the mob's outcome and rank the importance of each factor in determining the mob outcome. Results suggest that using factors related to individuals and aggregated per event made the model better classify mob outcomes (success or failure) than using event-related factors. Also, combining these factors did not affect the performance. However, all models performed better than the base model, which used raw data.
Affective Foraging: Knowledge Graph-Assisted Analysis of Emotion and Topic Information Patches in Online Discourse on the 2025 US Tariffs
(2026-01-06) Kok-Shun, Brice Valentin; Chan, Johnny; Peko, Gabrielle; Sundaram, David
The tariffs introduced by the second Trump administration in early 2025 sparked significant public discourse on social media. This paper presents a computational framework for analyzing the emotional and thematic content of this discourse, focusing on YouTube as a key platform. We introduce a novel methodology for emotion detection and topic modelling using a combination of GenAI and embedding-based models to analyse 5874 transcripts and 866673 comments. The knowledge graph created from this data revealed clusters of related emotions and topics which were subsequently used to surface patterns of affective foraging. These patterns highlight the ways in which users navigate emotional content and seek information in a complex digital landscape. Our results contribute to the field of Information Systems by adding an emotional dimension to Information Foraging Theory, proposing the concept of Affective Foraging that integrates Digital Emotion Regulation with the latter to depict “informavores” as emotionally inclined, irrational decision-makers.
Introduction to the Minitrack on Artificial Intelligence, Machine Learning, and Analytics for Social Media
(2026-01-06) Mentzer, Kevin; Yates, David; Gerhart, Natalie

Browse

Recent Submissions