Quantified Authenticity and subjective well-being
In Study 1, we analyzed the data of 10,560 Facebook users who had completed a personality assessment and reported on their Life Satisfaction through the myPersonality application31,32. To estimate the extent to which their Facebook profiles represent authentic expressions of their personality, we compared their self-ratings to two observational sources: predictions of personality from Facebook Likes (N = 9237)33 and predictions of personality from Facebook status updates (N = 3215)34. These are based on recent advances in the automatic assessment of psychological traits from the digital traces they leave on Facebook35. For each of the observable sources, we calculated Quantified Authenticity as the inverse Euclidean distance between all five self-rated and observable personality traits. Our measure of Quantified Authenticity exhibits a desirable level of variance, ranging all the way from highly authentic self-expression to considerable levels of self-idealization (see ridgeline plot of Quantified Authenticity calculated for self-language and Self-Likes in Supplementary Fig. 3, see Supplementary Tables 1 and 2 for zero-order correlations among variables).
To test the extent to which authentic self-expression is related to Life Satisfaction, we ran linear regression analyses predicting Life Satisfaction from the two measures of Quantified Authenticity (Likes, status updates). The results support the hypothesis that higher levels of authenticity (i.e. lower distance scores) are positively correlated with Life Satisfaction (Table 1, Model 1 without controls). These effects remained statistically significant when controlling for self-reported personality traits. Additionally, we included a control variable for the overall extremeness of an individual’s personality profile (deviation from the population mean across all five traits), as people with more extreme personality profiles might find it more difficult to blend into society and therefore experience lower levels of well-being36 (see Table 1, Model 2 with controls; the results are largely robust when controlling for gender and age, see Supplementary Table 3; see Supplementary Figs. 1 and 2 for interactions between individual self-reported and predicted personality traits).
To further explore the mechanisms of Quantified Authenticity, we conducted analyses that distinguished between normative self-enhancement (i.e., rating oneself as more Extraverted, Agreeable, Conscientiousness, Emotionally Stable, and Open-minded than is indicated by one’s Facebook behavior) from self-deprecation (i.e., rating oneself lower on all of these traits). While normative self-enhancement has a negative effect on well-being, normative self-deprecation has no effect. These findings suggest that self-enhancement specifically, rather than overall self-discrepancy/lack of authenticity, is detrimental to subjective well-being (see Supplementary Fig. 4).
To test the robustness of our effects, we regressed Life Satisfaction on three additional measures of Quantified Authenticity (i.e., calculated using Manhattan Distance, Cosine Similarity, and Correlational Similarity; see SI for details on these measures). In both comparison sets (likes and status updates), we found significant and positive correlations between the various ways of estimating Quantified Authenticity (see Supplementary Tables 1 and 2). The standardized beta-coefficients across all four metrics of Quantified Authenticity and observable sources are displayed in Fig. 1. Despite variance in effect sizes across measures and model specifications, the majority of estimates are statistically significant and positive (11 out of 16). Importantly, no coefficients were observed in the opposite direction. These results suggest that those who are more authentic in their self-expression on Facebook (i.e., those who present themselves in a way that is closer to their self-view) also report higher levels of Life Satisfaction.
In exploratory analyses, we considered whether authenticity might benefit individuals of different personalities differentially. In order to examine this, we regressed Life Satisfaction on the interactions between Quantified Authenticity and each of the five personality traits (e.g., Quantified Authenticity × Extraversion). The results of these interaction analyses did not provide reliable evidence for the proposition that individuals with socially desirable profiles (i.e., high openness, conscientiousness, extraversion, agreeableness, and low neuroticism) benefit from authentic self-expression more than individuals with less socially desirable profiles (see Table 1, Model 3). While the interactions of the five personality traits with Quantified Authenticity reached significance for some traits and measures, the results were not consistent across both observable sources of self-expression (Likes-based and Language-based). Consequently, we did not find reliable evidence that having a socially desirable personality profile boosts the effect of authenticity on well-being. Instead, individuals reported increased Life Satisfaction when they presented authentic self-expression, regardless of their personality profile.
The findings of Study 1 provide evidence for the link between authenticity on social media and well-being in a setting of high external validity. However, given the correlational nature of the study, we cannot make any claims about the causality of the effects. While we hypothesize that expressing oneself authentically on social media results in higher levels of well-being, it is also plausible that individuals who experience higher levels of well-being are more likely to express themselves authentically on social media. To provide evidence for the directionality of authenticity on well-being, we conducted a pre-registered, longitudinal experiment in Study 2 (see Fig. 2 for an illustration of the experimental design).
Experimental manipulation of authentic self-expression on well-being
We recruited 90 students and social media users at a Northeastern University to participate in a 2-week study (Mage = 22.98, SDage = 4.17, 72.22% female). The sample size deviates from our pre-registered sample size of 200. The reason for this is that the behavioral research lab of the university was shut down after the first wave of data collection due to the COVID-19 pandemic.
All participants completed two intervention stages during which they were asked to post on their social media profiles in a way that was: (1) authentic for 7 days and (2) self-idealized for 7 days. The order in which participants completed the two interventions was randomly assigned. This experimental set-up allowed us to study the effects of authentic versus idealized self-expression on social media in between-person (week 1) and within-person analyses (comparison between week 1 and week 2). All analyses were pre-registered prior to data collection37. Given the reduced sample size, the effects reported in this paper are all as expected in effect size, but only partially reached significance at the conventional alpha = 0.05 level. Consequently, we also consider effects that reach significance at alpha = 0.10 as marginally significant.
All participants completed a personality pre-screen (IPIP)38 prior to beginning the study, and received personalized feedback report at the beginning of the treatment period (t0). Both the authentic and self-idealized interventions (see Methods for details) asked participants to reflect on that feedback report and identify specific ways in which they could alter their self-expression on social media to align their posts more closely with their actual personality profile (authentic intervention) or to align their posts more closely with how they wanted to be seen by others (see Supplementary Information for treatment text and examples of responses). The operationalization of the treatment follows our conceptualization of Quantified Authenticity in Study 1 in that it does not prescribe the direction of personality change (e.g. towards higher levels of extraversion). Instead, this design leaves it up to participants what posting in a more desirable way means in relation to their current profile.
Participants self-reported their subjective well-being as Life Satisfaction39, a single-item mood measure, and positive and negative affect40 a week after the first intervention (t1), and a week after the second intervention (t2). This design allowed us to examine the causal nature of posting for a week in which participants posted authentically (“authentic, real, or true”), compared to a week in which they posted in a self-idealized way (“ideal, popular or pleasing to others”). Specifically, we hypothesized that individuals who post more authentically over the course of a week would self-report greater subjective well-being at the end of that week, both at the between and within-person level.
We examined the effect of authentic versus self-idealized expression at the between person level at t1 (see t1 in Fig. 3) using independent t-tests. Contrary to our expectations, we did not find any significant differences between the two conditions for any of the well-being indicators. This suggests that individuals in the authentic vs. self-idealized conditions did not differ from one another in their level of well-being after the first week of the study. However, when examining the effect within subjects using dependent t-tests we found that participants reported significantly higher levels of well-being after the week in which they posted authentically as compared to the week in which they posted in a self-idealized way. Specifically, the well-being scores in the authentic week were found to be significantly higher than in the self-idealized week for mood (mean difference = 0.19 [0.003, 0.374], t = 2.02, d = 0.43, p = 0.046) and for positive affect (mean difference = 0.17 [0.012, 0.318], t = 2.14, d = 0.45, p = 0.035), and marginally significant for negative affect (mean difference = −0.20 [−0.419, 0.016], t = −1.84, d = 0.39, p = 0.069). There was no significant effect on Life Satisfaction (mean difference = 0.09 [−0.096, 0.274], t = 0.96, d = 0.20, p = 0.342).
These findings are reflected in Fig. 3 which showcases the interactions between condition and time point. The graphs highlight that subjective well-being was higher in the weeks in which participants were asked to post authentically (red bars) compared to those in which they were asked to post in a self-idealized way (blue bars). While there was no difference in subjective well-being across conditions at t1, subjective well-being measures differed significantly between the authentic and self-idealized conditions at t2. We found no significant difference between conditions on Life Satisfaction (mean difference = 0.29 [−0.226, 0.798], t = 1.11, d = 0.23, p = 0.270), however, we found a significant difference between conditions such that the group which received the authenticity treatment had greater positive affect (mean difference = 0.45 [0.083, 0.825], t = 2.43, d = 0.51, p = 0.017), lower negative affect (mean difference = −0.57 [−1.034, −0.113], t = −2.47, d = 0.52, p = 0.015), and higher overall mood (mean difference = 0.40 [0.028, 0.775], t = 2.14, d = 0.45, p = 0.036).
The findings of the experiment provide support for the causal relationship between posting authentically, compared to posting in a self-idealized way, on the more immediate affective indicators of subjective-wellbeing, including mood and affect, but not on the more long-term, cognitive indicator of life satisfaction. This findings aligns with our pre-registration in that we had predicted mood and affect measures to be more sensitive to the treatment compared to Life Satisfaction, which is a broader global assessment one’s overall life39 and less likely to change in the course of a week.
Additionally, the fact that we did not find significant effects in our between-subjects analysis in the first week of the study suggests that authentic self-expression might be difficult to manipulate in a one-off treatment as social media users are likely used to expressing themselves on social media both authentically and in a self-idealized way. Thus, when only one strategy is emphasized, participants might not shift their behavior. This is supported by the finding that participants did not differ significantly in their subjective experience of authenticity on social media at t1 (mean in authentic condition at t1 = 5.56, mean in self-idealized condition at t1 = 5.55, t = 0.05, d = 0.01, p = 0.958; Participants responded to a single item, which read “This past week, I was authentic on social media” on a 7-point scale where 1 = strongly disagree and 7 = strongly agree), indicating that the between-subjects manipulation was unsuccessful in getting people to shift their behaviors more toward self-idealized or authentic self-expression compared to their baseline. However, the contrast of the two strategies highlighted in the within-subjects part of the study seems to have successfully shifted participants’ behavior. When compared within person, students did indeed report higher levels of experienced authenticity in their posting during the week in which they were instructed to post authentically (mean difference = 0.30 [0.044, 0.556], t = 2.33, d = 0.49, p = 0.022).