**From --- Johnna Blair, Chi-Yang Hsu, Ling Qiu, Shih-Hong Huang, Ting-Hao (Kenneth) Huang, and Saeed Abdullah. 2021. Using Tweets to Assess Mental Well-being of Essential Workers During the COVID-19 Pandemic. In CHI Conference on Human Factors in Computing Systems Extended Abstracts (CHI’21 Extended Abstracts), May 8–13, 2021, Yokohama, Japan. ACM, New York, NY, USA.**
The Covid-19 pandemic has had a significant impact on mental health and well-being of the population as a whole, leading to large-scale lifestyle changes, social isolation, and increased stress. However, this has been especially pertinent to essential workers—facing high workloads, insufficient safety supplies, and risk to their own health and the health of their families when returning home9 12 . This, combined with reduced in-person support, suggests a highly complex and challenging situation for essential workers.
Approximately 55 million Americans have been deemed “essential workers” during the pandemic10 . The majority of essential workers in the US work within the healthcare industry (30%)—including clinicians and any hospital staff, agriculture and food production (20%), and the commercial service industry (12%)—such as retail or grocery store workers10 . Workers in these industries are also disproportionately women, minorities, immigrants, people over 50, and low-income, which can put them at a further disadvantage12 . For low-income workers or single income families, this can involve making the difficult decision between personal safety and a paycheck9 .
Not only are essential workers' physical health at risk, but their mental health as well. Healthcare professionals, in particular, are now at higher risk for mental health conditions than the general public—reporting enough symptoms on average to constitute a depression diagnosis11 . These workers have experienced higher levels of stress, anxiety, and tiredness, perceive lower feelings of control over their lives, and are less able to prepare themselves for future stressful events. Those in other essential industries face similar stressors that may also put them at risk for poor mental health.
How did essential workers use Twitter during the pandemic?
Social media platforms allow users to connect with others about shared experiences, from coping with mental health conditions2 3 4 8 to voicing workplace concerns1 13 . Platforms like Twitter have also become tools to voice shared societal concerns, allowing individuals to organize online movements for change—such as unionization13 and higher minimum wage demands from retail and food service workers across the US1 . These connections help users feel less isolated and allow them to bond with other people in similar situations that they may lack access to in their offline lives2 3 .
At the same time, social media use can be used to infer general well-being. Characteristics like when and how often people tweet can be used to map out irregular sleep patterns, a common occurrence with depression and anxiety5 6 . Similarly, the stress, anxiety, or depression experienced by users can reflect in the sentiment of their tweets, leading to more negatively associated posts. This same process can be used to assess shared experiences, both positive and negative, at a societal-level. By analyzing Twitter use and tweet sentiment, along with more specific keywords, it has been shown that we can even differentiate users with depression from non-diagnosed populations7 . Given what can be learned through Twitter use, we asked what these same processes can tell us about a key group of users—essential workers—and their experiences and well-being before and during the pandemic, as compared to the average Twitter user.
Using key phrases such as “I am an essential worker” to identify these users, we gathered all their tweets authored from January 2019 through September 2020. Overall, 4055 accounts were analyzed—1752 essential worker accounts and 2303 random Twitter accounts, to represent the average Twitter user, as a comparison. We focused on general usage, tweet sentiment, and the use of Covid-19 and mental health keywords, to explore what Twitter can initially tell us about essential worker experiences.
When looking at simply when and how often users post tweets, essential workers tweeted less frequently than the average Twitter user and did so later in the day—with most of their activity between 5pm and 1am, as compared to average Twitter users who were more active between 9am and 3pm. On average, the ratio of tweets posted by essential workers and average Twitter users both increased during the pandemic, but essential workers still appeared less active by comparison.
Considering findings from other Twitter studies, fewer tweets could suggest lower sociability and essential workers’ late-night activity could suggest a disrupted sleep schedule. Both of these are commonly associated with poor mental well-being. However, these behaviors were consistent for essential workers before and during the pandemic, meaning that we cannot claim the pandemic as a single driving factor. These differences could be the result of job-related factors, such as having later or longer work schedules and less free time to devote to Twitter.
When it came to what users chose to tweet about, we looked at their tweets about two different topics—Covid-19 and mental health. We used keywords such as Covid, Covid 19, Coronavirus, mask(s), safe, pandemic, sick, and risk to gather pandemic-related tweets. Tweets with keywords like sad, struggle, stress, stressed, anxiety, anxious, depression, depressed, coping, and mental health were used to examine mental health related tweets. Overall, essential workers posted fewer tweets about the pandemic as compared to average Twitter users. Regarding mental health however, essential worker tweets contained a higher ratio of these keywords than those from average Twitter users. This could hint at some of the leading concerns of essential workers, such that they choose to talk more openly about mental health—either their own or on a societal level, and less so about the pandemic itself—something that they already deal with in their offline lives in ways beyond the average Twitter user.
Most surprising, despite the level of stress and high demand put on essential workers, their tweets were more positive than the average user, when analyzed for overall sentiment. This trend was also consistent before and during the pandemic. Not only were essential worker tweets more positive than the average user, but the positive sentiment of essential worker tweets before the pandemic did not significantly drop during the pandemic. Instead, their positivity remained relatively consistent from 2019 through 2020.
At this point, we cannot definitively say where this positivity comes from, but it could suggest that characteristics of essential jobs and the people who hold them may genuinely make workers happier or more positive. For example, some essential jobs—such as healthcare workers—can provide increased job security, higher pay, a higher perceived sense of purpose, or higher social status as compared to those in other job roles, which could affect how workers appear. However, this is not the case for all essential jobs, such as retail and service jobs, which generally provide lower pay, less stability, and increasingly poor treatment from the public during the pandemic.
On the other hand, the type of job they hold may prompt essential workers to put on a more positive online presence. Instead, it may be an issue of differentiating between how users actually feel versus what they choose to put online for others to see. The essential worker accounts sampled were more likely to be personal, individual user accounts, used less anonymously than Twitter users on average. They may avoid posting about negative or polarizing topics for fear of work-related consequences, potentially swaying the average sentiment of their tweets more positively. Conversely, the average Twitter users randomly sampled in this study could possess characteristics that lead them to post more negatively-associated content by comparison.
It could also be that essential workers simply turn to Twitter with different goals than the average user. Considering that essential workers use Twitter later in the evening, they might be motivated to seek out positive connections with others following long, stressful work days, instead of venting their negative experiences. While we cannot say for certain at this stage, this mix of online behaviors could be a deliberate choice to escape pandemic-related topics and focus on other needs, such as supporting each other and trying to remain positive during a very difficult time.
Our initial look at Twitter during the pandemic was based on a small dataset and as such, we focused on exploratory analysis and findings regarding possible positive or negative shifts in sentiment. However, subsequent work leveraging a larger Twitter dataset has reported similar findings on generic twitter use during the pandemic14 . More specifically, as the pandemic progressed through to the release of vaccinations, users’ tweets contained fewer negative emotional words while talking about the pandemic and showed an overall drop in negativity. Though similar to our work, the answer to why this is the case is still undetermined.
How can we support essential workers moving forward?
This initial study helps address how essential workers use Twitter differently than the average user, but more importantly, it provides us with new questions to ask about the underlying factors driving these unique characteristics and surprising positivity. At the same time, it highlights how Twitter and other platforms can be used as a positive outlet and a tool for social support and potential social change, despite the negativity that is often expected in online spaces. From a technological perspective, this can provide new avenues for research focused on making these spaces more conducive to positive support in times of social isolation.
In future work, it is important to consider essential workers not as a monolith, but to examine individual experiences and identify needs existing across different industries and job types. Through a more nuanced lens, we can capture a more true-to-life picture of essential worker well-being. By getting to the core of what is driving this trend of positivity, we can identify groups of essential workers in need of more support during crisis situations, now and in the future.
In doing so, we can call attention to important issues faced by those in disadvantaged positions with high-risk, high-demand jobs that provide workers with less income, social status, or agency, to help mitigate their increased stress. Assessing large scale well-being and common stress points for at-risk individuals, based on what they share in these online spaces, can help inform future policies to further support their needs through offline means, such as reinforced support networks and a potential societal shift in how we view “essential work” and the people that keep our society functioning day to day.
- 9 a b The Lancet. 2020. The plight of essential workers during the COVID-19 pandemic. Lancet (London, England)395, 10237 (2020), 1587. https://dx.doi.org/10.1016%2FS0140-6736(20)31200-9
- 12 a b Hye Jin Rho, Hayley Brown, and Shawn Fremstad. 2020. A Basic Demographic Profile of Workers in Frontline Industries. https://cepr.net/a-basic-demographic-profile-of-workers-in-frontline-in…
- 10 a b Celine McNichols and Margaret Poydock. 2020.Who are essential workers? A comprehensive look at their wages, demographics, and unionization rates. Economic Policy Institute: Working Economics Blog(may2020). https://www.epi.org/blog/who-are-essential-workers-a-comprehensive-look…
- 11Ann Pearman, MacKenzie L Hughes, Emily L Smith, and Shevaun D Neupert.2020. Mental Health Challenges of United States Healthcare Professionals During COVID-19. Frontiers in Psychology 11 (2020), 2065. https://doi.org/10.3389/fpsyg.2020.02065
- 2 a b Sairam Balani and Munmun De Choudhury. 2015. Detecting and characterizing mental health related self-disclosure in social media. In Proceedings of the 33rdAnnual ACM Conference Extended Abstracts on Human Factors in Computing Systems. 1373–1378. https://doi.org/10.1145/2702613.2732733
- 3 a b Natalie Berry, Fiona Lobban, Maksim Belousov, Richard Emsley, Goran Nenadic, and Sandra Bucci. 2017. # WhyWeTweetMH: understanding why people use Twitter to discuss mental health problems. Journal of medical Internet research19, 4 (2017), e107. https://www.jmir.org/2017/4/e107/
- 4Johnna Blair and Saeed Abdullah. 2018. Supporting constructive mental health discourse in social media. In Proceedings of the 12th EAI International Conference on Pervasive Computing Technologies for Healthcare. 299–303. https://doi.org/10.1145/3240925.3240930
- 8Nelya Koteyko and Dimitrinka Atanasova. 2018. Mental health advocacy on Twitter: positioning in Depression Awareness Week tweets. Discourse, Context and Media (2018). https://doi.org/10.1016/j.dcm.2018.04.007
- 1 a b Steven Ashby. 2017. Assessing the Fight for Fifteen Movement from Chicago. Labor Studies Journal 42, 4 (2017), 366–386. https://doi.org/10.1177/0160449X17742794
- 13 a b Aaron S Veenstra, Narayanan Iyer, Mohammad Delwar Hossain, and Jiwoo Park. 2014. Time, place, technology: Twitter as an information source in the Wisconsin labor protests. Computers in Human Behavior31 (2014), 65–72. https://doi.org/10.1016/j.chb.2013.10.011
- 5Glen Coppersmith, Mark Dredze, and Craig Harman. 2014. Quantifying mental health signals in Twitter. In Proceedings of the workshop on computational linguistics and clinical psychology: From linguistic signal to clinical reality. 51–60. https://aclanthology.org/W14-3207.pdf
- 6Munmun De Choudhury, Scott Counts, and Eric Horvitz. 2013. Social media as a measurement tool of depression in populations. In Proceedings of the 5th Annual ACM Web Science Conference. 47–56. https://doi.org/10.1145/2464464.2464480
- 7Munmun De Choudhury, Michael Gamon, Scott Counts, and Eric Horvitz. 2013. Predicting depression via social media. In Proceedings of the International AAAI Conference on Web and Social Media, Vol. 7. https://www.aaai.org/ocs/index.php/ICWSM/ICWSM13/paper/viewPDFInterstit…
- 14Alessandro N. Vargas, Alexander Maier, Marcos B. R. Vallim, Juan M. Banda, and Victor M. Preciado. 2021. Negative perception of the COVID-19 pandemic is dropping: Evidence from Twitter posts. Frontiers in Psychology, (2021), 4067. https://doi.org/10.3389/fpsyg.2021.737882