Understanding Emotional Disclosure via Diary-keeping in Quarantine on Social Media

Quarantine, as a restriction on the movement of people, is intended to prevent the spread of disease. To contain COVID-19, unprecedented quarantine strategies are operated and dynamically adjusted corresponding to the pandemic development. With the epidemic panic, the confinement policy tends to exacerbate mental health problems (Bavel et al., 2020; Ma et al., 2021). In this vein, social media platforms have become popular sites for individuals to express their opinions, describe their lives and interact with social networks, thus mapping individuals’ responses and behaviors to their psychological traits (Hung et al., 2020; Ang et al., 2015). Owing to the prevalence of social media, researchers have paid great attention to how social media help disclose mental health, like discovering the shifts of suicidal ideation (De Choudhury et al., 2016), understanding the virtual attributes of mental health (Manikonda and De Choudhury, 2017) and proposing Mental Well-being Index drawing on social media (Bagroy et al., 2017). Compared with traditional methods in mental health (e.g., questionnaires and scales), social media could naturally, expediently, and timely collect emotional and behavioral information, becoming a promising place for recognizing psychological effects (Adikari et al., 2021; Lu et al., 2020). Although social media expressions cannot guarantee the identical emotional intensity or absolute authenticity of feelings with privacy or sensitivity concerns (Bazarova and Choi, 2014), they showcase users’ thoughts, emotions, and experiences to some extent.

Hence, this work explores mental well-being during quarantine by comprehensively understanding the characteristics of emotional disclosure via social media (RQ1), changing patterns of emotional disclosure (RQ2) and specific concerns of the public which provide guidelines to mitigate mental health disorders (RQ3).

Social media has become a widely used channel to express emotions (He et al., 2021; Kelley and Gillan, 2022; De Choudhury et al., 2013, 2016). A burgeoning body of research investigated how emotions were disclosed with the extensive use of social media during quarantine. One strand of work focused on emotional differences between various groups of individuals (Wang et al., 2020; Zhu et al., 2021). For example, Zhu et al. revealed the anxiety difference between six types of quarantine (e.g., forced quarantine at home and centralized quarantine) (Zhu et al., 2021). Another stream of research centered on the changes of emotions where interesting changing patterns usually hide (Cao et al., 2021; Lu et al., 2020; Feng and Ivanov, 2022). For example, Cao et al. uncovered emotional evolution during each phase of Wuhan lockdown including the incubation phase, explosive phase, declining phase and stable phase (Cao et al., 2021).

In previous research the measurement of emotions was mostly based on existing API (Li et al., 2022; Hung et al., 2020), psychological scale (Canet-Juric et al., 2020), or only from the perspective of words and themes (Arora et al., 2021) rather than scenario-based analysis. However, emotions are comprehensive manifestations of not only words but also contexts. Furthermore, the majority of previous studies about emotion categories were theory-driven, e.g., drawing on Plutchik’s model with basic emotions (Adikari et al., 2021) or LIWC with various emotional dimensions (Ashokkumar and Pennebaker, 2021) to determine emotion categories.

The specific emotions expressed by individuals in quarantine, utilizing a bottom-up approach to emotion categorization, remain yet to be established. Thereby, this study contributes to this area by employing an inductive approach to investigate common emotions that align closely with the context of quarantine.

An online diary, also known as a web diary, is defined as a personal journal or diary that individuals choose to publish on the World Wide Web ¹¹1https://en.wikipedia.org/wiki/Online_diary. The United States witnessed the emergence of the earliest online diaries around 1994 (Ben-Amos and Ben-Amos, 2020). In China, the rise of online diary-writing began in 2000 when Lu Youqing’s cancer battle diary was serialized on a popular online literature website, attracting a significant number of followers (Hockx, 2015). The introduction of blogs to China in 2002 further popularized online diary-writing, making it widely recognized. Thus, online diaries have a significant and extensive history, serving as a common form for people to document their daily lives. During the lockdown in Wuhan, residents turned to social media like WeChat and Weibo, and started writing and sharing their experiences via ”Lockdown Diaries”, providing a glimpse into life in a sealed city (Ni et al., 2020). The diaries attracted scholarly attention, with studies focusing on various aspects such as feminist issues (Bao, 2020), and the impact of the pandemic on daily life and work. They were part of a global trend, with pandemic diaries appearing worldwide, offering access to wider audiences and facilitating interaction (Nierenberg, 2020). However, this online exposure also brought unwanted attention and censorship. Previous studies primarily focused on Wuhan, including lockdown diaries in Wuhan or nationwide quarantine diaries during the outbreak period in Wuhan (Feng and Ivanov, 2022). Nevertheless, this diary form exists nationwide, and the quarantine measures have been in place for an extended period of time.

In general, diaries that convey emotional disclosure across the country during the extended period of quarantine remain unexplored. This research gathers nationwide data spanning approximately 2 years to uncover the changing patterns of emotional disclosure as quarantine becomes a new normal, as well as to identify potential factors associated with negative emotions.

Weibo, one of the largest microblogging websites in China similar to Twitter, has become a dominant site for understanding the well-being and health trends of the public due to its characteristics of sharing, disseminating and receiving instant information (Li et al., 2020; Chen et al., 2021; He et al., 2022). Weibo was a significant and popular channel for recording daily life and expressing feelings during quarantine, where posts with hashtag ”#Quarantine Diary#” had owned more than 1.2 billion reads and more than 0.5 million discussions (including original posts and retweets) ²²2https://m.s.weibo.com/vtopic/detail_new?click_from=searchpc&q=%23%E9%9A%94%E7%A6%BB%E6%97%A5%E8%AE%B0%23.

To obtain a preliminary understanding of the emotional status in quarantine, we kept track of posts related to quarantine on Weibo in advance. We found users applied some strategies to identify topic-related posts including hashtags, Super Topic (i.e., communities consisting of people having common interests and following specific topics (Williams, 2019)) and keywords. In order to make the data comprehensive, we selected the keyword-based collection approach rather than other approaches based on hashtags or Super Topic. Moreover, the most straightforward keyword was ”Quarantine”, but posts with ”Quarantine” were miscellaneous containing official quarantine policies, suggestions after quarantine or even advertisements of primers (i.e., cosmetics). Compared with ”Quarantine”, we noticed that posts with the ”Quarantine Diary” keyword, where users tended to record and share their activities and emotions for one day or several days, getting closer to the personal quarantine lives and self-expression. Hence, we chose ”Quarantine Diary” as the keyword, so as to make the collected data align with our scenario.

We collected the data with the assistance of the WeiboSuperSpider tool (Xiao, 2019) ranging from April 29, 2020, to July 25, 2022. Since we only considered the quarantine in China, we removed the overseas posts (i.e., the quarantine locations are not in China) by identifying the keyword ”Overseas Quarantine” and the locations of posts in self-identified profiles.

In total, 58, 796 original posts were retrieved with the keyword ”Quarantine Diary” from 34, 677 distinct users.

This section described how we constructed two filtering models for obtaining emotional posts and one classification model for emotional categories, to understand the categories of emotional disclosure in quarantine.

In this section, we described how we identified the temporal information in posts and investigated temporal characteristics from long-term and period-based views.

Data utilized for this section was 31, 714 emotional posts. After browsing these emotional posts, we found the temporal information of posts would be expressed in three ways as follows: (1) ”Day” or ”D” plus a number (e.g., ”Day 9” or ”D 9”); (2) using an ordinal number (e.g., ”the first day”); (3) ”Quarantine Diary” plus a number (e.g., ”Quarantine Diary 9”). Moreover, there was multiple temporal information in some posts containing two or three types of temporal information or utilizing one type of temporal information more than once. To circumvent this challenge, we adopted a priority-based regular expression matching approach to extract the temporal information for each post. The priority was the same as the order mentioned above. If there was more than one temporal number in one type, we selected the maximum number as the temporal information. Note that the priority order and the maximum temporal approach were determined based on the accuracy of temporal information after manual coding. After the extraction operation, there existed 15, 256 posts with temporal information.

To comprehensively explore how emotional status changed during quarantine, we considered it from two views: (1) long-term changing patterns, the temporal characteristics of emotions from April 2020 to July 2022 to have an overall exploration of emotional disclosure; (2) period-based changing patterns, the temporal characteristics of emotions related to quarantine durations to further understand the day-by-day emotions, including 21-day global changing patterns (i.e., based on posts whose temporal information was within 21 days, e.g., ”Quarantine Diary 6”), as well as 7-day and 14-day grouped changing patterns (i.e., based on users whose quarantine duration were 7 days or 14 days). Note that 7 days and 14 days were two typical quarantine periods adopted by the government in China (Commission, 2021) and recommended by WHO (Organization et al., 2020). We used proportions instead of numbers to capture changing patterns of emotional disclosure. By doing so, it allows us to effectively unveil relative distribution changes among emotional categories, considering the varying number of posts collected at different points in time.

The negativity bias theory (Adikari et al., 2021) indicates that negative emotions play a more significant role in the public’s behaviors and cognition. In the context of quarantine, annoyed and anxious generally exhibit a stronger expression of personal concerns and potentially have more direct impacts on emotional well-being, relationships, and life quality, while boring typically represents a mundane account of daily life. Hence, in this section, we described how we explored factors that might lead to annoyed and anxious.

The statistical results of emotional disclosure during quarantine were demonstrated in Figure 3. Based on the results of three models constructed in Section 3.2, we concluded the following findings: (1) Most posts (79.57%, N=46, 785) containing the ”Quarantine Diary” keyword tallied with our research context shown in Figure 3 (a), proving that this type of data collection approach was feasible; (2) Emotional posts were predominant (67.78%, N=31, 714) compared with emotionless ones indicated in Figure 3 (b), which demonstrated that people in quarantine were willing to express their emotions on social media; In Figure 3 (c), (3) Posts with negative emotions (45.77%, N=14, 516) in quarantine occupied almost the half of all emotional posts; (4) Annoyed (36.34%, N=11, 525) was the dominant category in amount compared with anxious (6.79%, N=2, 154) and boring (2.64%, N=837); (5) In terms of positive emotions, happy (35.96%, N=11, 405), hopeful (13.42%, N=4, 256) and appreciative (4.85%, N=1, 537) were sorted in descending order.

We focused on how emotional disclosure changes during quarantine from long-term and period-based perspectives. More specifically, we first unveiled the changing patterns from April 2020 to July 2022 to have an overall exploration of emotions in this context, and then investigated those related to quarantine durations, which further understood the day-by-day emotions during quarantine.

After keeping track of some interesting temporal characteristics of emotional disclosure, we probed further into the latent factors that might contribute to negative emotional disclosure, attempting to explore effective countermeasures. Table 2 and Table 3 summarized the topical themes and corresponding keywords with LDA after an inductive thematic analysis described in Section 3.4. The abundance of topical themes indicated there were diverse possible factors denoting negative emotional disclosure, and the reasons behind each negative emotion were also extensive. To effectively and precisely alleviate negative emotions, managers of epidemic prevention and control were suspected to pay more attention to divergent and fundamental measures instead of purely providing psychological counseling to regulate emotions in quarantine.

By identifying quarantine-specific emotional categories, this work enriches the understanding of emotional disclosure from the perspectives of how individuals respond to quarantine, and how the responses demystify the practical and design implications to aid psychological balance during quarantine.

First, this work revealed the resilience and adaptivity of the public in the scenario of quarantine to some degree. As shown in Section 4.1, the proportion of positive emotions was slightly higher than negative ones, and happy (35.96%) played the most significant role in positive emotions, indicating the emotional well-being of the population on average was challenging but not as bad as imagined. Meanwhile, our findings in Section 4.2 further demonstrated that negative emotions diminished over time. Indeed, a type of natural recovery capability (i.e., resilience and adaptivity) was corroborated. It also dovetailed with the cognitive dissonance theory (Festinger, 1962), pointing out that in order to reduce negative psychological effects, people are more likely to change their cognition if there is no way of changing the environment. In this regard, adding the recovery capability as a reference index during the psychological treatment process in quarantine, would be promising to better help mental health management.

Second, this work also uncovered severe psychological stress in people during quarantine. As demonstrated in Section 4.1, the proportion of posts with negative emotions was 45.77% among all posts. The massive negative content on social media could form a vicious circle like pyramid schemes. From the perspective of receivers, it is common that people have more engagement in negative content on social media (Peng et al., 2019; Naveed et al., 2011). Negativity bias theory (Adikari et al., 2021) indicates that people’s behavior and cognition are more likely to be affected by negative emotions. From the other perspective of posters, on account of social compensation theory (Valkenburg and Peter, 2007), people with psychological distress have more intense motives to disclose on social media who desire more social connection and affiliation (Lee et al., 2013; Tian, 2013; Hollenbaugh and Ferris, 2014). Furthermore, recall that the most predominant negative category was annoyed exhibited in Section 4.1, rather than anxious. The more outstanding performance of annoyed compared with anxious bears out the fatigue phenomenon during quarantine due to endless epidemic (WIlliams and Dienes, 2021; Ross et al., 2021). Angry speech is proven to be more likely to receive abusive language, in turn, (Mechkova and Wilson, 2021). Then, a new round of negative outcomes emerges on social media. Thus, detecting and moderating negative emotional disclosure on social media is of great imperative, especially with the extensive negative exposure to social media during quarantine. In light of the fatigue phenomenon, how to adjust social support measures accordingly (e.g., maybe from previously disseminating professional information about the virus and debunking misinformation to providing more emotional accompany), deserves more in-depth exploration. Moreover, the emotion analysis approach in our work, which focuses on identifying quarantine-specific emotional categories, investigating emotional changing patterns and exploring latent factors of emotions, could be considered as an adjunct in mental health diagnosis during quarantine or subsequent epidemic era. Taking a broader perspective, our work offers compelling evidence supporting the feasibility and significance of conducting emotion analysis in affective computing by annotating context-specific emotions. By focusing on the emotions experienced within particular contexts, we can gain valuable insights into individuals’ mental well-being and tailor interventions and support accordingly.

In the specific setting of quarantine, we notice a novel form of emotional disclosure: people in quarantine spontaneously write diaries on social media for mundane recording, experience sharing and opinion expression. By uncovering the diary-keeping behaviors on social media, we reflect on the design implications of how diaries during extraordinary times could be supported via social media, and the practical implications of health intervention strategies based on emotional disclosure.

The intimacy with data is well exhibited via social-media-based diary-keeping in our work. The narration technique on social media, leveraged by users to delineate their daily life, provides a complementary perspective for application-centered numerical tracking techniques (e.g. Personal Informatics apps (Cho et al., 2022)) in the way of person-centered instructive description. The functions on social media like sharing, forwarding, commenting and mentioning (i.e., @) provide a platform for users to communicate with each other, stimulating and boosting the social and cultural potential. The changing patterns in Section 4.2 provided evidence that employing large-scale observation of quarantine diaries could, to some degree, serve as a viable approach for identifying and comprehending emotional disclosure. On this note, emotion-detecting systems and system-level interventions (e.g., reverse recommendation of positive information when detecting extreme negative emotions) could come into being by future designers and researchers. Note that we do not instigate users to open their private and specific data on social media but just encourage the narrative way of diaries. On the long view, the present identification approach of diary-self-tracking is based on the keyword (e.g., Quarantine Diary) and temporal label (e.g., Day 1) which is manually tracked by users. To this end, system-facilitated construction of diaries (e.g., providing an elaborately-designed diary template) may be considered for future researchers. Also, integrating diary apps or other self-tracking functions into social media, and whether the diary interface should become more strict like numerical recording or more flexible like paper journaling, are a worth-exploring research direction. Nevertheless, some studies have shown that self-disclosure on social media, especially negative sharing, can engender a lot of risks (e.g., privacy) (Bazarova and Choi, 2014) and need to be carefully created and negotiated (Andalibi and Forte, 2018). Social media posts are also subject to content moderation and surveillance, so it is hard to assume that people share experiences and opinions of quarantine without concern.

Drawing on the outcomes of social-media-based diary-keeping, health intervention strategies are engendered. As shown in Section 4.2, the discrepancy of each group in emotional disclosure demonstrated individuals might have different psychological statuses and concerns at the same period. Such different patterns enable governments to comprehend the people in quarantine from multiple perspectives when making healthcare policies, and specialized healthcare measures for each group are recommended. Meanwhile, health intervention strategies (e.g., psychological crisis consultation time) are suspected to dynamically adjust and focus on several days, such as the 3rd and 4th days when all negative emotions were at a high level. A bounce came up in annoyed near the end of quarantine, indicating the health management department cannot ignore the negative emotions close to the end. A plausible interpretation is that approaching the end of quarantine, people are suddenly informed to remain in quarantine with the annoyed increasing. These findings empirically validated the feasibility of understanding emotional disclosure from a quantitative perspective. The changing patterns can also enlighten mental health regulations during similar public health emergencies.

Prior work ascribed great importance to self-disclosure and social support as critical approaches to improving living standards and self-efficacy. Negative emotional disclosure demonstrates the latent concerns and dissatisfaction of the public and functions as an impetus to provide available direction for social support. This section further discusses what kind of social support could be provided to alleviate negative emotions based on self-disclosure.

This work comprehensively signified the intricate characteristics of factors that might account for negative emotions and revealed the potential need for social support. Driven by the social support theory (Cohen and Wills, 1985), and the effect mechanism of self-disclosure and psychological well-being (Luo and Hancock, 2020), typical social support categories consist of instrumental (i.e., the enacted supportive behaviors) and emotional (i.e., the perceived impression) support. In essence, our observations of potential social support needs in Section 4.3 were in line with the theory-driven category. On the one hand, instrumental social support to deal with practical difficulties like short supply and poor infrastructure from governments and related departments, is imperative. Also, given that uncertainty is a remarkable factor during quarantine, construing a channel on social media to interpret policies and measures, may be beneficial to mitigate negative emotions and avoid more serious situations caused by misinformation. On the other hand, social media together with official organizations, could be taken into account for emotional social support. The emotional needs like freedom restriction and solitude and feeling the epidemic is endless, verified online accompany is vital due to abrupt reduction of physical interactions. On this note, tailored accompanying communities consisting of experienced people, professionals or digital volunteers, and interactive chatbots on social media are recommended for future designers to cure people in quarantine who are grappling with negative emotions.

This work adopted a quantitative approach to comprehensively investigate the categories of emotional disclosure, the changing patterns of emotional disclosure and factors that might give rise to negative emotions during quarantine, which enlighten the practical and design implications for better relieving mental health issues. Nonetheless, this work has the following limitations: (1) While social media can serve as a means of communication, posts shared on social media may not completely capture or convey the full range of emotions or explicit and implicit needs of the public. Individuals moderating their content and engaging in self-surveillance on social media could present a curated version that doesn’t reflect true emotions or comprehensive needs. (2) Due to the presence of multiple or repeated types of temporal information within posts, the exact accuracy of the extracted temporal information cannot be guaranteed, even though we have employed rules based on manual encoding and extensive experience.

When keeping a diary on social media becomes prevalent but related interface designs are still limited, we encourage a customized diary interface while combining the features of personalization and flexibility on social media in the future. Additionally, future work could focus on the difference between social-media-based diaries and paper journaling to obtain valuable inspiration for a user-friendly interface.