Psychosocial dynamics of suicidality and nonsuicidal self-injury: a digital linguistic perspective

Abstract

Self-harm-encompassing suicidality and nonsuicidal self-injury (NSSI)-presents a critical public health concern, particularly as it is a major risk factor of death by suicide. Understanding the psychosocial dynamics of self-harm is imperative. Accordingly, in a large-scale, naturalistic study, we leveraged modern language analysis methods to provide a comprehensive perspective on suicidality and NSSI, specifically in the context of borderline personality disorder (BPD), where self-harm is particularly prevalent. We utilised natural language processing techniques to analyse Reddit data (i.e., BPD forum posts) of 992 users with self-identified BPD (combined N posts = 66,786). The present findings generated further insight into the psychosocial dynamics of suicidality and NSSI, while also uncovering meaningful interactions between the online BPD community and these behaviours. By integrating advanced computational methods with psychological theory, our findings provide a nuanced understanding of self-harm, with implications for clinical practice, clinical and personality theory, and computational social science.

Fig. 1. GLMM emotion plots: changes in mean emotion language in proximity to recent suicidality and nonsuicidal self-injury (NSSI) events.

The figure shows changes in mean emotion language category scores (derived from LIWC) per week (i.e., aggregated weekly) surrounding suicidality and NSSI events. The dotted lines illustrate the point at which engagement in the event occurred (i.e., time point 0), thus dividing the figures by pre- and post-event. The shaded areas surrounding the means represent the error margins (95% confidence intervals). The means (and confidence intervals) have been estimated from the generalised linear mixed models (GLMMs), and thus are reflective of the repeated measures nature of the data (i.e., person-centered) while also controlling for random user effects. The indicators assigned to the suicidality and NSSI keys show the statistical significance of the overall fixed effects of time in proximity to the events: ***p < 0.001, **p < 0.01, *p < 0.05, ^†p < 0.10. Time point labels: −3 = three weeks before event, −2 = two weeks before event, −1 = one week before event, 1 = one week after event, 2 = two weeks after event, 3 = three weeks after event.

Fig. 2. GLMM social plots: changes in mean social language in proximity to recent suicidality and nonsuicidal self-injury (NSSI) events.

The figure shows changes in mean social language category scores (derived from LIWC) per week (i.e., aggregated weekly) surrounding suicidality and NSSI events. The dotted lines illustrate the point at which engagement in the event occurred (i.e., time point 0), thus dividing the figures by pre- and post-event. The shaded areas surrounding the means represent the error margins (95% confidence intervals). The means (and confidence intervals) have been estimated from the generalised linear mixed models (GLMMs), and thus are reflective of the repeated measures nature of the data (i.e., person-centered) while also controlling for random user effects. The indicators assigned to the suicidality and NSSI keys show the statistical significance of the overall fixed effects of time in proximity to the events: ***p < 0.001, **p < 0.01, *p < 0.05, ^†p < 0.10. Time point labels: −3 = three weeks before event, −2 = two weeks before event, −1 = one week before event, 1 = one week after event, 2 = two weeks after event, 3 = three weeks after event.

Fig. 3. Comparisons of community support generated in self-harm versus non-self-harm posts (N = 9106 posts).

This figure presents a visual display of the RQ2 t-test results, in which post scores and number of replies received were compared between posts disclosing past/recent engagement in suicidality/NSSI versus all other manually coded posts (i.e., posts coded as not disclosing suicidality/NSSI). Post scores reflect the number of upvotes a post receives subtracted by the number of downvotes, thereby reflecting the overall “rating” of the post. Number of replies received are in relation to submissions only (i.e., not responses to comments). The figure is organised by the type of the event disclosed (i.e., suicidality or NSSI) and whether the post disclosed past or recent engagement in the event. Error bars represent the standard errors.