BACKGROUND
The COVID-19 pandemic has caused a global disruption, starting with a public health emergency, followed by a significant loss of human life and a severe economic and social fallout. As physical distancing regulations were introduced to manage outbreaks, individuals, groups and communities took to social media to express their thoughts and emotions reflecting different behaviours. This has led to increased interaction on social media thereby recording diverse behaviours of people as the pandemic progressed.
OBJECTIVE
This research aims to explore the human behaviours recorded on the digital traces of social media during the pandemic. The investigation is focused on examining the emotions, emotion state and intensity changes, topical associations and different groups of people using social media conversations that uncover informed insights on behaviours during the COVID-19 pandemic.
METHODS
Our study explores emotion classifications, intensities, transitions, profiles and alignment to key themes and topics, across the four stages of the pandemic; declaration of a global health crisis, first lockdown, easing of restrictions, and the second lockdown. This study employs a human-centric artificial intelligence (AI) based framework comprising of natural language processing, emotion modelling, unsupervised clustering methods that are collectively used to investigate the social media conversations. The investigation was carried out using 73,000 Twitter conversations related to Australia from January to September 2020.
RESULTS
The outcomes of this study enabled to analyse and visualise different emotion behaviours and concerns reflected on social media during the COVID-19 pandemic. First, the topic analysis showed the diverse and common concerns people have expressed during the four stages of the pandemic. It was noted that starting from personal level concerns, the concerns expressed over social media has escalated to broader concerns over time. Second, the emotion intensity and emotions state transitions showed that ‘fear’ and ‘sad’ emotions were more prominently expressed at first, however, they transition into ‘anger’ and ‘disgust’ over time. Negative emotions except ‘sad’ were significantly higher (P < .05) in the second lockdown showing increased frustration. Emotion state changes during these stages were visualised to comprehend the change in emotions over time. Third, based on the concerns expressed social media users were categorized into profiles. The profiles in the first lockdown differed from the profiles in the second lockdown showing the shift of concerns as the pandemic progressed.
CONCLUSIONS
This study showed diverse emotion behaviours and concerns recorded on social media during the COVID-19 pandemic. While this study establishes the use of social media to discover informed insights during a time where physical communication is impossible the outcomes also contribute towards post-pandemic recovery and understanding people’s emotions better during crises. The study exploits AI and social media to enhance our understanding of human behaviours in global emergencies, leading to improved planning and policymaking for future crises.