The COVID-19 pandemic presented an unprecedented global health crisis, prompting widespread lockdown measures. These measures, while necessary to curb the virus's spread, led to significant psychological distress, including anxiety and depression, as evidenced by studies on quarantine's impact (Qui et al., 2020; Brooks et al., 2020). Simultaneously, internet and social media usage surged to unprecedented levels (Effenberger et al., 2020; Fischer, 2020). This study explores the hypothesis that social media, in the pandemic context, could act as a buffer against anxiety by facilitating collective resilience processes. The researchers posit that social media's role is nuanced, potentially leading to either powerlessness or resilience depending on how individuals process information. This is particularly relevant given the looming threat of the pandemic, characterized by dynamic and uncertain risk factors, exacerbating the gap between perceived and actual knowledge (Fischhoff, 2020; Neuhoff, 1998; Riskind and Rector, 2018; Chater, 2020; Yu and Dayan, 2005; Radecki and Jaccard, 1995). The study builds upon the looming vulnerability model (LVM) and the concept of Virtual Collective Consciousness (VCC) (Marzouki et al., 2012; Marzouki and Oullier, 2012, 2015; Alperstein, 2019), suggesting that social media can enhance social support functions (Cohen and Willis, 1985; Meng et al., 2017) and ultimately buffer stress.

The introduction extensively reviews existing literature on the psychological impacts of quarantine and lockdown measures, highlighting the prevalence of anxiety and depression. It also cites studies on the increased social media use during the pandemic. Key theoretical frameworks, such as the Looming Vulnerability Model (LVM) and the concept of Virtual Collective Consciousness (VCC), are introduced to provide a foundation for understanding social media's potential role in shaping individual responses to the pandemic. The stress-buffering model of social support is also discussed, outlining how social media might provide self-esteem support, informational support, diffuse support, and instrumental support.

This study employed a cross-sectional online survey design, utilizing a snowball sampling technique given the global lockdown. Data was collected over nine weeks from March to May 2020, yielding 1408 responses. Week 4 was excluded due to a small sample size (n=4). The survey included questions on social media use, self-perceived knowledge about the pandemic, perceived threat, and anxiety levels. The researchers analyzed the data using several methods: 1. **Structural Equation Modeling (SEM):** A SEM model was used to examine the relationship between perceived knowledge (comprising social media use, self-perceived knowledge, and perception of threat) and anxiety. The model was tested for each week (excluding weeks 4 and 7). Goodness-of-fit indices (CFI and RMSEA) were used to assess model fit. 2. **Text Corpus Analysis:** Participants were asked to provide up to five keywords associated with their experience. This corpus (N=862) underwent cluster analysis (using bisecting K-means) and correspondence analysis to identify thematic clusters (Global Knowledge, Anxiety, Symptoms, Coping). A co-occurrence matrix was also generated to visualize thematic clustering. 3. **Discriminant Analysis:** Discriminant function analysis was performed to determine if the study variables (fear, social media usage, self-knowledge, and perceived threat) discriminated between the thematic clusters. 4. **Sentiment Analysis:** LIWC software was used to analyze the sentiment (positive and negative valence) and cognitive categories (cognitive processing, space, time, death) in the text corpus. 5. **Entropy Analysis:** Entropy was calculated to measure corpus diversity and reflect the collective mindset over time. The Balch (2000) algorithm was used for calculating entropy index.

The SEM analysis revealed a consistent, though variable, relationship between perceived knowledge and fear across the weeks. A trade-off pattern was observed between self-knowledge and threat perception, conditioned by social media use. As self-knowledge's impact decreased, threat perception's impact increased, and vice versa. Text corpus analysis identified four main clusters: Global knowledge, Anxiety, Symptoms, and Coping. The relative presence of 'Global Knowledge', 'Anxiety', and 'Coping' clusters showed a dynamic evolution over time. Discriminant analysis showed that in weeks 6, 7, and 8, the study variables significantly discriminated between the clusters, indicating an association between thematic clusters and combinations of fear, social media use, self-knowledge, and perceived threat. Sentiment analysis showed a cyclical pattern, with a turning point in week 5. This week showed a peak in cognitive processing words and a decrease in negatively valenced words, suggesting a shift from emotional responses to cognitive appraisal. Entropy analysis indicated a progressive increase in lexical diversity, especially after week 5, reflecting increased cognitive processing and a move away from a solely negative collective perception. The initial negative consensus shifted to a more diverse, coping-oriented narrative from week 5.

The findings support the hypothesis that social media can buffer anxiety during a pandemic. The trade-off between self-knowledge and threat perception, influenced by social media use, highlights the dynamic interplay between information seeking, cognitive appraisal, and emotional responses. The shift observed around week 5 signifies a collective reappraisal of the threat, facilitated by social support and information exchange on social media platforms. This aligns with the stress-buffering model of social support and the concept of VCC, demonstrating how social media can foster collective resilience. The increase in cognitive words and decrease in negatively valenced words after week 5 reflect a move towards a more adaptive and coping-focused response. The results align with uncertainty reduction theory, showing how information seeking through social media reduces uncertainty and fosters consensus. The study extends existing research on the relationship between media consumption, perceived knowledge, and actual knowledge, showing the complexity of information processing within social media contexts.

This study demonstrates social media's significant buffering effect on pandemic-related anxiety. The dynamic interplay between emotional responses and cognitive appraisal, facilitated by social media, is crucial for understanding collective resilience. Future research should explore the mediating effects of social media use with other variables determining coping strategies, compare the effects of different social media platforms, and consider cultural impacts on collective resilience processes. The study provides valuable insights for developing public health communication strategies during future lockdowns, emphasizing the importance of reducing uncertainty, using social media to foster social norms, and guiding users toward organized online socialization.

The study's limitations include the use of a non-probabilistic sample and the absence of socio-demographic variables. The absence of these variables was a deliberate choice to minimize participant burden during lockdown. The study focuses on the structural relationships between variables rather than providing exact incidence estimations.