In modern daily life, social media (SM) is ubiquitous; on individuals’ phones, laptops, tablets, and other devices, these websites and applications enable people to create posts, share and view content, and communicate with others. Unlike other social interactions, those on SM are always available to individuals, regardless of whether others in their networks are currently nearby or online (Nesi et al., 2018). These behaviors on SM can also be more public and permanent than the words and actions used during in-person interactions (Nesi et al., 2018). Due to its ubiquity, constant availability, publicity, and permanence, SM has unique impacts on the well-being of individuals beyond those of other social interactions (Davila et al., 2012; Nesi et al., 2018).

For late adolescents and emerging adults, these impacts may be more pronounced (Coyne, et al., 2013; Perrin, 2015; Scott et al., 2017). Firstly, almost all individuals in this age group use SM and report its use as one of their most frequent activities (Coyne et al., 2013), thus creating the possibility of more frequent or accumulative impacts. In addition, late adolescence and emerging adulthood is a period characterized by increasing independence in one’s life in the transition to increased responsibility, moving out of their caregiver’s home, the workplace, and or college attendance (Arnett, 2007)—with individuals in this age group using SM, the creation of a personal profile and the increasing engagement in independent activity on SM may be a manifestation of the emerging adult’s independence that holds great value to them. Social relationships in this period are still immensely important, and are pursued with greater social-cognitive maturity than their early-adolescent counterparts; SM represents an important medium through which emerging adults create and maintain these relationships (Arnett, 2007; Scott et al., 2017).

When previous work has examined the impacts of SM use in late adolescents and emerging adults, much evidence has emerged that SM use can be maladaptive for this age group. These impacts manifest in a variety of facets of well-being, including greater body dissatisfaction, greater depressive symptoms, greater exposure to cyber-bullying, greater negative affect, and lesser positive affect (Añez et al., 2018; Best et al., 2014; Holland and Tiggemann, 2016, Marino et al., 2018; Rasmussen et al., 2020). However, previous work has also revealed that not all types of SM use show these associations; it is rather the motivation with which an individual uses SM and the activities they engage in on it that determine whether SM use is maladaptive (Griffioen et al., 2023; Keum et al., 2023). In particular, the use of SM with the motivation of seeking reassurance or feedback on oneself or one’s appearance (e.g., making posts with the aim of receiving attention through likes, views, and comments, and feeling negatively about oneself if the sought-after attention does not come) is associated with detriments to well-being. In addition, using SM for the purpose of social comparison (e.g., viewing others’ posts, seeing their popularity, and feeling inferior in comparison) may be maladaptive (Davila et al., 2012; de Vries et al., 2015; Fuller-Tyszkiewicz et al., 2019; Holland and Tiggemann, 2016, Marino et al., 2018; Nesi & Prinstein, 2015). For example, Nesi and Prinstein (2015) illustrated that both social comparison and feedback-seeking behaviors on SM were predictive of depressive systems, even when total technology use and prior depressive symptoms were taken into account. These motivational and activity-specific factors, more than the amount of SM use, seem to be key to determining the impacts of use on well-being.

In contrast to the maladaptive uses of SM, more recent work has begun to explore the possible adaptive uses of SM that can foster well-being instead. In this literature, too, findings suggest that any benefits of SM use depend on the motivation driving it and the activities engaged in. Specifically, previous work suggests that SM use can be adaptive when it is used for self-expression, such as sharing one’s creative projects or interests, or when it is used to extend, maintain, or deepen social network connections, such as by communicating directly with others in messages (Allen et al., 2014; Davila et al., 2012; Keum et al., 2023; Weinstein, 2018). For example, Weinstein (2018) examined how different types of SM use as gleaned from narrative accounts of online experiences related to high-schoolers’ affective well-being; results revealed, among other findings, that participants who described using social media to share their ideas and make content, or to maintain closeness with friends outside of in-person interactions, reported greater positive affect and lower negative affect than other participants. Other positive outcomes associated with these more adaptive uses of SM include greater self-esteem, greater life satisfaction, and greater perceived social connectedness and support (Allen et al., 2014; Best et al., 2014; Keum et al., 2023). Thus, there is much evidence that SM use impacts the well-being of the late adolescents and emerging adults who use it, but whether these impacts are harmful or beneficial is determined by the specific characteristics of their SM use.

However, the previous research on adaptive and maladaptive SM use discussed above has often assessed usage with global measures, such as total time using SM, or retrospective self-report measures, such as asking participants to report how often they engaged in certain SM activities over the past month (Griffioen et al., 2023; Marino et al., 2018; Nesi et al., 2018; Tibber et al., 2020). Although such measures have generated useful knowledge, they may be subject to retrospective reporting biases (Beal, 2015), and do not illuminate how the relationship between SM use and well-being may unfold in daily life from moment to moment. Whereas global and retrospective measures of social media use can investigate between-person differences in well-being based on SM use, intensive longitudinal methods, such as daily diaries or ecological momentary assessments (EMA), are needed to explore the process of this relationship and how SM use and well-being covary in real time. With these methods, because many observations are collected from each participant, individuals can be used as their own points of comparison while accounting for overall individual differences in SM use. Thus, intensive longitudinal methods would enable the examination of whether individuals experience different levels of well-being at times when they engage in certain SM activities, compared to themselves at times when they do not engage in those activities.

To address this gap, the current study investigated the association between a range of specific SM activities and concurrent affective well-being in daily life using EMA in a sample of emerging adults. We hypothesized that firstly, the activities likely associated with maladaptive SM use for comparison and feedback-seeking, such as viewing others’ posts and checking likes and comments on one’s own posts, would relate to greater concurrent negative affect and lower positive affect. Secondly, we hypothesized that the more adaptive forms of SM use related to self-expression and connection with others, such as posting original content and direct messaging with others, would predict greater positive affect and lower negative affect. We expected that these patterns would emerge both at the within-person and the between-person levels. Because the bulk of previous work has examined the relationship between SM use and well-being with global measures and between-person comparisons, any notable differences in within-person and between-person patterns of results in the current study will suggest the need for further examination of this relationship in daily life to tease apart the processes behind global between-person differences.

Method

Sample

The current study conducted a secondary data analysis with participants from the Stony Brook Temperament Study, an ongoing longitudinal study based in Stony Brook, NY, that began when participants were 3 years old (Klein & Finsaas, 2017). Participants who had completed the EMA period within the study’s age 18 wave of data collection were included, as this EMA period contained the measures of interest. This data collection occurred from 2019 through 2022. The COVID-19 pandemic began partway through data collection; 77% of participants completed their EMA period after the pandemic began.

Participants who had very low response rates (completed fewer than 20% of EMA surveys; 31 participants) were excluded from analyses, as were any specific measurement occasions that participants completed extremely quickly (survey completed in under 30 s; 132 measurement occasions). These exclusions left an analytical sample of 349 participants and 18,737 total observations. All participants were 18 years of age. As such, some participants were currently attending high school (43.79%), some were attending college (43.79%), some were working (3.27%), and some were not currently enrolled in school or working (9.80%). In terms of sex, 183 (52.44%) participants were female and 166 (47.56%) were male. The majority of the sample identified as White (92.26%), followed by Black/African American (4.87%), Asian (2.29%), Native American/American Indian (.29%), and Other (.29%). Across these groups, 11.75% of the sample identified as Hispanic.

Procedure

The current study utilized data from the 14-day EMA period of the Stony Brook Temperament Study’s age 18 wave of data collection. During the EMA period, participants completed up to 5 surveys per day, creating a maximum of 70 observations per person.

Participants received survey prompts pseudo-randomly throughout the day, with the specification that all survey prompts were a minimum of 2.5 h apart. The hours throughout the day in which participants completed surveys was determined according to the participant’s reported wake-up time on weekdays. This wake-up time was reported to study staff in a training session for the study’s EMA period, and was coded as either before 7 am, between 7:30 and 9:30 am, or after 9:30 am. For the first survey of each day, participants had 1 h to respond after the survey prompt. For the remaining four surveys throughout the rest of the day, participants had 2 h to respond after the survey prompt. Participants were compensated $1 USD for each of the first 40 surveys they competed. For the remaining 30 surveys, participants earned $2 USD for each additional survey completed. Thus, participants could earn up to $100 USD for completing all surveys.

In these surveys, participants reported their positive and negative affect within the last hour and engagement in five SM activities within the last hour, among other variables. On average, participants completed 55.42 (SD = 14.76; 79.17%) of the total 70 possible surveys, with the median number of surveys completed being 62 (88.57%).

Measures

Outcomes

Affect

Positive and negative affect were each calculated as the mean of four individual affect items presented in each EMA survey. Positive affect represented the average score of the items “happy,” “excited,” “cheerful,” and “content/peaceful.” Negative affect represented the average score of the items “sad/down/depressed,” “anxious/worried/nervous,” “irritated/annoyed/angry,” and “upset.” Participants were asked to rate how much they had experienced each of these feelings in the last hour on a five-point scale from “not at all” to “extremely.”

Predictors

SM Activities

In each EMA survey, participants also reported whether they had engaged in a list of five activities on any SM platform within the last hour. These activities included “posting a story, video, picture, or status update,” “viewing another person’s posts, story, or profile,” “liking or commenting on another person’s story or post,” “checking likes, views, or comments,” and “sending a direct message.” Participants could endorse as many options as applied. Reports of engagement in each activity were coded as binary yes or no variables for each activity.

Statistical Approach

To investigate the relationship between these SM activities and affective well-being, we utilized three-level multilevel linear models to account for the data structure of moments nested within days, which were nested within individuals. We conducted separate parallel models for positive affect and negative affect as outcomes. These models included fixed effects for each of the five momentary SM activity variables at the moment-level, as well as the person-means for each of the five SM activity variables at the person-level. With all SM activity predictors included in models simultaneously, we were able to discern each activity’s unique association with positive and negative affect, with the associations of other activities partialled out. No day-level fixed effects were included in the models. Both models included random effects for the day-level and person-level intercepts and utilized an unstructured random effects covariance structure. We attempted to include random effects for the moment-level slopes of the five SM activities, but with these additional parameters, the models did not converge; thus, the random slope effects were not retained in the models.

Results

Descriptive Statistics

Full descriptive statistics for all model variables are presented in Table 1. All descriptive statistics were calculated at the person level; that is, each person’s mean level was calculated across observations, and these individual means were averaged to describe the analytical sample. On average, participants reported moderate levels of positive affect (M = 2.82, SD =0.75) and low levels of negative affect (M = 1.44, SD = 0.45). The ICC for positive affect was 0.54, and for negative affect was 0.44; these values represent the proportions of variance in a variable that is due to stable individual differences, with the remaining variance due to momentary fluctuations within individuals and error.

Table 1 Person-level Descriptive Statistics
Full size table

Participants reported having used SM within the last hour at about half of their measurement occasions (44.13%). In their use of SM, participants reported engaging in direct messaging with the highest frequency (on average, reported on 21.28% of all surveys), followed by viewing others’ posts (21.12%), liking or commenting on others’ posts (10.61%), making their own posts (4.86%), and lastly, checking posts for likes, views, and comments (4.68%).

Analytical Models

Complete results from both analytical models are presented in Table 2. Examining SM activities’ relationship with positive affect, results showed significant fixed effects at the moment-level in the positive direction for making posts (B = .109, SE = .024, p < .001) and for direct messaging (B = .054, SE = .015, p < .001). Thus, at times when participants reported making their own posts or direct messaging, they reported higher positive affect than at times when they had not engaged in any SM activities. There were also significant fixed effects in the negative direction at the moment-level for liking or commenting on others’ posts (B =  − .062, SE = .020, p = .002) and for viewing others’ posts (B =  − .037, SE = .015, p = .012), such that at times when participants reported liking or commenting on, or viewing others’ posts, they reported lower positive affect than at times when they did not report any SM activities. However, the relationship between checking likes, views, and comments and positive affect at the moment-level was not statistically significant (B =.041, SE = .027, p = .122). Lastly, there was also a significant fixed effect of viewing others’ posts at the person-level (B =  − .678, SE = .274, p = .014), such that participants who reported viewing others’ posts in a greater proportion of surveys across the study period also reported lower positive affect on average across the study period. The fixed effects for all other SM activities at the person-level were not statistically significant. When gender interaction effects with each of the within-person SM activity predictors were included in this model, none of these interaction effects was significant.

Table 2 Multilevel Models Predicting Positive or Negative Affect from SM Activities
Full size table

Surprisingly, in regard to the model predicting negative affect, no statistically significant fixed effects emerged. At both the moment-level and at the person-level, none of the SM activities showed statistically significant relationships with negative affect. Again, no gender interaction effects with the SM activity predictors were significant.

Discussion

The current study examined the relationship between engagement in specific SM activities and affective well-being in daily life in a sample of emerging adults. We hypothesized that activities that are likely related to maladaptive uses of SM, such as viewing others’ posts and checking likes and comments on one’s own posts, would relate to lower positive affect and greater negative affect, whereas activities related to adaptive SM use, such as making posts and direct messaging, would related to greater positive affect and lower negative affect.

In our examination of relationships with positive affect, these hypotheses were supported strongly at the moment-level. Firstly, viewing others’ posts, as well as liking and commenting, was associated with lower positive affect. These activities may involve individuals voluntarily or involuntarily engaging in the maladaptive motivation to use SM for social comparison, and thus suffering decrements in their positive emotions due to feelings of inadequacy compared to others (Nesi & Prinstein, 2015; Weinstein, 2018). Liking and commenting on others’ posts may especially indicate deep engagement with and internalization of others’ posts, thus creating more opportunity for social comparison. Although we did not find the expected negative relationship between positive affect and checking for likes and comments on one’s own posts, which is an activity that would likely reflect maladaptive feedback-seeking, we suspect that this null effect may be due to the low frequencies of this activity being reported in the sample (5%).

Secondly, making one’s own posts and direct-messaging others were both associated with greater positive affect at the moment-level. Again, these results align with previous work on the adaptivity of motivations for SM use. Making one’s own posts, which likely involves utilizing one’s creativity and expressing one’s interests, may be driven by the motivation for self-expression, an adaptive use of SM (Best et al., 2014; Weinstein, 2018). Direct messaging, on the other hand, likely involves interacting with a close other or deepening relationships with newer individuals, which may illustrate the adaptive motivation to use SM for increasing social connectedness (Allen et al., 2014, Weinstein, 2018). These findings also align with the characteristics of emerging adulthood; expressing one’s own interests and curating one’s profile, as well as maintaining and deepening social connections, may be a reflection of the importance of independence and social relationships in this developmental period. Together, these relationships with positive affect in both correlational directions point to the need for specificity in measurement in regard to what individuals are doing on SM, as some activities may heighten, and others may lessen desirable outcomes.

However, these relationships with positive affect did not emerge at the person-level. These null results are notable; although between-person differences have been found in previous work, much of this research has relied on global and retrospective measures of SM use, which may not be sensitive to the moment-to-moment covariation between SM use and affective well-being. The current study, which repeatedly assessed both of these variables in the natural context of daily life over short, comparable time scales, thus contributes a new layer of understanding to the processes behind the relationship of SM use and well-being. When measured on short time scales, our findings suggest that the primary associations between SM use and well-being occur within persons, with certain SM activities related to increases or decreases in affective well-being compared to oneself. Future work involving measurement of these constructs in daily life will be important for further exploring the within-person and between-person components of this relationship.

Surprisingly, none of the hypothesized relationships between SM activities and negative affect emerged as significant, either at the moment-level or at the person-level. Firstly, these null effects imply that in this case, positive and negative affect are not simply two ends of a single bipolar construct, because increases and decreases in positive affect were not matched with opposite decreases and increases in negative affect; thus, this pattern points for the need to measure positive affect and negative affect separately.

Beyond that, the lack of statistically significant results with negative affect could be because the situations measured in the current study are not the ones that are strongly related to negative affect. Other situations might be more strongly related to negative affect, such as when another person responds insultingly to one’s post, or one feels that their post has been ignored. Alternatively, individual difference variables, such as personality and appraisal processes, could produce divergent patterns of relationships between SM activities and well-being, thus not producing detectable effects overall across all individuals. For example, the personality trait neuroticism has been linked with a greater general tendency to appraise social situations as negative or stressful, whereas the trait agreeableness has been linked with lower levels of that tendency (Luo et al., 2023). Other work has noted personality differences in types of SM use (Orchard & Fullwood, 2010), so it may be the case that these differences extend to SM use’s relationship with well-being as well. Lastly, although past studies examining between-person differences in SM use and well-being have, for example, found increased depressive symptoms due to maladaptive SM use (Davila et al., 2012; de Vries et al., 2015; Nesi & Prinstein, 2015), it may be the case that measuring these variables in daily life may illuminate different processes in this relationship than global measurements do. Future work should continue to examine the interplay between these variables on small time scales and outside of global or retrospective reports.

Future work with ongoing data collection in the Stony Brook Temperament Study will also be able to provide valuable insights into the relationship between SM use and affective well-being. As noted above, the current study’s data were collected from participants when they were 18 years of age, an important developmental period for growing independence and social development. However, it is unclear how the relationship between SM use and well-being might continue to develop and change within persons as individuals age. Using a measurement burst design (Zavala et al., In press), the Stony Brook Temperament Study will conduct additional EMA periods at the study’s next timepoints; these future “bursts” of EMA data collection will allow us to test whether the dynamics of SM activity and affect are more trait-like in nature and remain constant at later ages, or shift after this transitional developmental period is complete.

A few limitations of the current study bear mentioning. Firstly, due to the longitudinal single-cohort study design of the Stony Brook Temperament Study, all participants were from the same age cohort, and were the same age at data collection. So, although emerging adults are an especially important group in which to examine the impacts of SM use, it is unclear if our findings would emerge with other age groups, or with emerging adults of different cohorts. In addition, our sample consisted predominantly of White, non-Hispanic participants living in Long Island, NY. Thus, more work is needed to examine whether these results generalize to a more diverse population.

Secondly, much of the current study’s data were collected after the COVID-19 pandemic began. In this period, stay-at-home orders, illness, safety precautions, and virtual classes reduced the number of opportunities individuals had to socialize in person. This characteristic of the pandemic period may have increased individuals’ reliance on SM for social interaction, and may have strengthened the relationship between affect and SM use in our sample. Future studies will be needed to explore this relationship in emerging adults who do have access to in-person socialization. However, the current study’s results did not differ when pandemic period (pre-pandemic or during pandemic) was included as a covariate in the multilevel models, so any impact of the pandemic may not have changed the overall pattern of the relationship between SM use and affect.

Thirdly, although we assume that the mechanism of the relationship under study is causal, such that SM use determines the fluctuations in affect, the current study’s examination of this relationship is correlational. Thus, because affect and SM activities were assessed concurrently over the past hour, we cannot be sure of the casual direction of their relationship. For example, it may be the case that experiencing higher positive affect makes one more likely to engage in direct messaging on SM platforms, and not that one experiences higher positive affect as a result of the direct messaging. Although it would be ideal to generate knowledge on the possible directionality of this relationship, applying cross-lagged models in the analysis of EMA data would require that items be framed appropriately for pinpointing behaviors in time which was not the case in the current study (Lucas, 2023; Muthén & Asparouhov, 2024; Neubauer & Schmiedek, 2020). Fitting cross-lagged models as a secondary analysis raises concerns regarding possible errors in interpretation and replicability of results and no current consensus exists for estimating cross-lagged effects while accounting for the nuances of EMA data not designed for that purpose. Future work that is specifically designed to make targeted explorations of temporal precedence in the relationship between affective well-being and SM use may be able to better address this important question.

Despite these limitations, the current study expands the current literature on the relationship between affective well-being and SM use in a developmental period with high levels of SM use and increasing independence. In sum, these results suggest specificity—in which SM activities link with affective well-being, in the correlational direction of those links, and in links mostly with positive affect. Importantly, the current study suggests that these effects unfold within-persons in daily life, but may not be detectable in terms of individual differences.