Theoretical and practical importance:
Social media interactions have become ubiquitous. However, some of these interactions are negative. In a study of 1178 US college students, participants reported that over a quarter of their interactions on social media were negative, and negative interactions were associated with social isolation (Primack et al., 2019). My own work has demonstrated that rumination over negative social media interactions is linked with depression (Parent et al., 2019). Yet other work has demonstrated that specific groups may be at elevated risk for negative social media interactions, such as members of the LGBTQ+ community (Escobar-Viera, 2020)
In my prior work (Parent et al., 2019), I demonstrated that social media use was related to depression, but that this association was mediated by valence of online interactions. That is, negative online interactions mediated the relationship between time spent online and depression. This finding was an important addition to the literature, as much prior work on social media use focused on total amount of social media use and mood disturbance, without assessing how valence of online interactions mediates this relationship.
These investigations have primarily relied on cross-sectional data. The current proposal will use prolific to gather longitudinal data to examine the mental health impact of exposure to, rumination over, and interaction with negative material online.
These findings have practical importance. If negative social media interactions are associated with increases in depressive symptoms, interventions could be rapidly deployed. These interactions may take the form of traditional approaches to wellness (e.g., mindfulness as applied to rumination about or intention to actively engage with negative online interactions). Or, transdisciplinary approaches may be used to develop entirely new ways of addressing these mental health threats. For example, AI may be developed to recognize and replace a negative comment with a prompt asking if the user actually wants to see it, or AI may identify when a user is responding to a negative post and prompt the user to disengage rather than contribute to a negative interaction. Further research could then be applied to how to help individuals to strategically engage with negative material on social media (e.g., allocating attention toward social or political causes with which individuals may wish to engage but which may generate some negative feedback, and away from negative or antagonistic material that is not ultimately important to a user).
Methodology:
A random intercepts cross-lagged panel method (RI-CLPM) will be used to explore longitudinal effects. This approach allows for the delineation of events over time. Data will be collected on four key variables: Exposure to negative social media interactions, rumination over negative online interactions, active engagement in negative interactions, and depression.
Exposure to negative social media interactions will be assessed using a single item used in prior research (Primack et al., 2019) asking about percentage of online interactions (0%-100%) that have been negative. Negative interactions will be operationally defined (e.g., having someone post something insulting on one’s social media).
Rumination will be assessed using items from the affect-biased attention in online interactions measure used in my prior study (Parent et al. 2019). An example item is, “How often do you think about negative things people have said to you online, after you go offline?” Responses are made on a 5-point scale (1 = never, 5 = Very often).
Active engagement will be assessed using items from the affect-biased attention in online interactions measure used in my prior study (Parent et al. 2019). An example item is, “How often do you respond to negative comments posted in response to things you post online?” Responses are made on a 5-point scale (1 = never, 5 = Very often).
Depression will be assessed using the Patient Health Questionnaire-9 (Kroenke & Spitzer, 2011), a well-established measure of depressive symptoms.
Analyses will be conducted using Mplus version 8. The RI-CPLM model will be established such that each variable will be regressed onto itself at the prior time point. Depression will be regressed onto active engagement, rumination, and exposure at the prior time point. Active engagement and rumination will also be regressed onto exposure at the previous time point (i.e., both direct and mediational relationships will be estimated). I have experience in conducting similar longitudinal studies. Multigroup analysis will be used to examine for group differences in the model by race/ethnicity, gender identity, and sexual orientation.
Data collection will occur biweekly. At each data collection prior, participants will complete the assessments of exposure to negative online interactions, rumination, active engagement, and depression. The first time period will also be used to assess time-invariant variables such as age. Prolific.co user IDs will be used to match participants across time; participants at baseline will be used to generate a custom allow list for all subsequent time points to match longitudinal data.
Sample size:
A Monte Carlo analysis for the planned RI-CLPM indicated that a sample size of 250 would be sufficient to conduct the planned analysis. However, I aim to intentionally sample with representation for racial/ethnic, gender identity, and sexual orientation diversity in sufficient sizes to allow meaningful comparisons across these dimensions. The planned enrollment to facilitate such comparisons is 600.
Cost:
600 participants will be recruited. £4 in compensation will be provided at time 1, and £2 for time 2-5. Plus fees, this totals £9,576.
Preregistration:
This project is preregistered on the Open Science Framework. DOI: 10.17605/OSF.IO/HYTFN
Making findings available:
My goal will be to publish the research in a high quality open access journal. Whether the work is published in an open access journal or not, I will make the study data available via the Open Science Framework. Further, I will post a summary of the findings of the study in an infographic on my Twitter (@DrMikeParent1). I have created and posted such infographics to my social media platforms in the past and have found this to be an excellent way to generate public and media attention for research.
PI info: