Pandemic emergency: A multi‐level AI framework for analyzing psychological impact through public discourse

Emergency management has become a widely discussed topic in recent years, particularly in relation to different types of emergencies such as climate change, natural disasters, public health, and armed conflicts. The Covid-19 pandemic has triggered unprecedented uncertainty, sparking a global emergency that raised concerns not only about health but also social, economic, and environmental stability. At a time when psychology is re-evaluating its role in supporting processes of change during emergencies, a qualitative approach can offer a concrete response to the needs that the population highlights. This study explores, over a period of more than a year (from April 2020 to May 2021), the discursive productions shared by the Italian population on social media in relation to the Covid-19 pandemic emergency, in order to observe the impact on public health and identify clinical implications. A text analysis was performed using an annotation system based on 24 discursive modalities (learned with specific training by two humans and one machine learning model). IRaMuTeQ was also used for content detection and descending hierarchical classification. The analysis highlighted the need for support both at a clinical level and through sociopolitical intervention, based on how individuals expressed their experiences and concerns. Specifically, people's discourse tends to rigidly adhere to the status quo, where a focus on the "here and now" precludes the development of future-oriented health goals. Also, by prioritizing institutional critique over personal accountability, individuals adopt the role of passive observers rather than proactive participants in managing their own and the community's well-being. These results highlight how machine learning models can assist clinicians in better understanding the health needs of the population during emergencies, offering greater precision and speed in operational decision-making.