Does counting emotion words on online social networks provide a window into people’s subjective experience of emotion? A case study on Facebook.

Abstract

Psychologists have long debated whether it is possible to assess how people subjectively feel without asking them. The recent proliferation of online social networks has recently added a fresh chapter to this discussion, with research now suggesting that it is possible to index people's subjective experience of emotion by simply counting the number of emotion words contained in their online social network posts. Whether the conclusions that emerge from this work are valid, however, rests on a critical assumption: that people's usage of emotion words in their posts accurately reflects how they feel. Although this assumption is widespread in psychological research, here we suggest that there are reasons to challenge it. We corroborate these assertions in 2 ways. First, using data from 4 experience-sampling studies of emotion in young adults, we show that people's reports of how they feel throughout the day neither predict, nor are predicted by, their use of emotion words on Facebook. Second, using simulations we show that although significant relationships emerge between the use of emotion words on Facebook and self-reported affect with increasingly large numbers of observations, the relationship between these variables was in the opposite of the theoretically expected direction 50% of the time (i.e., 3 of 6 models that we performed simulations on). In contrast to counting emotion words, we show that judges' ratings of the emotionality of participants' Facebook posts consistently predicts how people feel across all analyses. These findings shed light on how to draw inferences about emotion using online social network data. (PsycINFO Database Record

Natural emotion vocabularies as windows on distress and well-being

“…Future studies incorporating trait preoccupation with moods (e.g., rumination) may be fruitful. EV's correspondence with mood is noteworthy given possible difficulties inferring momentary well-being from emotion word frequencies 40,41 . The EV approach bypasses this issue by relying not on frequency but rather on the diversity of emotion word categories.…”

Social Media and Well-Being: Pitfalls, Progress, and Next Steps

“…For example, one prominent study counted the number of emotion words contained in people's Facebook posts to draw inferences about how they felt although no validation data supported the use of such methods to track people's emotions on social media [103]. As later research pointed out, counting emotion words does not track how people feel on Facebook [104]. The take-home point is simple: psychometrically sound measures are not a luxury: they are instrumental for valid inferences.…”

Examining the language of solitude versus loneliness in tweets

“…These lexicons are specific to emotion with some developed to address discrete emotions (e.g., anger, sadness) (Aman & Szpakowicz, 2007;Mohammad & Turney, 2010) and others to address affect dimensions (e.g., valence, arousal) (Mohammad, 2018;Warriner et al, 2013). Researchers in computational linguistics and psychological disciplines have used these openly available lexicons to examine questions such as how emotions change at the micro timescale (Kross et al, 2019) and how emotions change at the macro, developmental timescale (Hipson, 2019), as well as such varied topics as attitudes toward products, politicians, and professional sports teams (Gratch et al, 2015;Maynard & Funk, 2011;Zhang et al, 2010).…”