Working remotely carries certain challenges, including the expression of emotions. Many workers use emojis to express emotion or adhere to social norms in their conversations with colleagues via asynchronous communication, such as instant messaging services, emails or forum discussions.
In a study that researchers recently published in PLOS One, variations in the use of emojis forecasted whether software developers working remotely dropped out from the online work platform GitHub.
The investigators collected online event log data from 2018 from GH (GitHub) Archive (62,852,221 posts). They found that about 5.53% of GitHub posts from 264,808 users, contained at least one emoji. Emoji users were matched with a random sample of nonemoji users.
Emojis were more common among pull request comments (a pull request takes place in software development when a developer is ready to begin the process of merging new code changes with the main project repository), compared with commit comments (used to save changes to a local repository after staging in GitHub), and they were likely used for work and emotional expression, the researchers reported.
They conducted ordinary lease square (OLS) regressions with emotional emojis as an outcome variable, controlling for primary programming language and tenure with GitHub, and found that emoji use was associated with the working status of individual developers. Increasing average working hours or number of posts was linked with emoji posts and proportion of emoji posts. More even spread-out of working activities over the week was linked with more emoji posts, but a lower proportion of emoji posts. The intensity of emotions expressed with emojis indicated developers’ current working status.
In 2019, users of emojis worked more than twice the number of days on average compared with those who did not use emojis (62.39 vs 23.13 P <<.001 Welch 2-sample t-test). Emoji users worked more hours daily, on average, compared with nonusers (1.71 vs 1.45 hours P <<.001 Welch 2-sample t-test).
Controlling for programming language use, the researchers found that nonemoji users were 3 times more likely to drop out in 2019 compared with emoji users (49.2% vs 13.0% across activity levels; P <<.001 Welch 2-sample t-test). Nearly 5% of nonusers whose days worked in 2018 were in the top fifth percentile dropped out in 2019. For emoji users, that figure was 1.4% (P <<.001 Welch 2-sample t-test).
Examining solely emoji users and controlling for activity levels by number of working days, they found, utilizing standard machine learning models, that developers who used emojis in posts were less likely to drop out, particularly when they tend to use emojis in posts, but a high proportion of posts containing emojis signals dropout as well. The accuracy of prediction models was mostly above 0.7 across activity levels.
The researchers said expressing emotion, including via emojis, is an effective way to manage stress and allows coworkers’ and practitioners’ opportunities to intervene appropriately.
“In online work platforms where people don’t express emotions face-to-face, it is important to encourage the workers to express their emotions in any possible way,” the researchers said.
The study’s limitations included inability to establish causality, no observation of offline or private messaging, lack of consideration of context, and generalization outside of the developers’ use of online collaborative platforms.
The researchers said that their main goal was to verify the predictive power of emotions remote workers expressed, rather than to optimize the accuracy of dropout prediction.
They also said that employers should pay attention to negative emotions — and emojis from remote workers — as signals of negative work experiences and to encourage workers to use a large variety of emojis, for multiple purposes.
“Using a variety of emojis may reduce the stress of work and distract workers from obsessive passion and potential burnout.”
Lu X, Ai W, Chen Z, et al. Emojis predict dropouts of remote workers: an empirical study of emoji usage on GitHub. PLOS One. 17(1): e0261262. doi:10.1371/journal.pone.0261262