The finding emerged from an experimental paradigm called Socializapp, an Instagram-mimicking application paired with ecological momentary assessment that tracked participants’ psychological states twice a day for a full week in their everyday environments.
The research team, led by Carlos Schmidt and Daniel Vega and based at hospitals and universities in Spain, recruited 65 women aged 18 to 30 who all held active Instagram accounts. The participants spanned a clinically meaningful severity continuum: 21 women with borderline personality disorder and recurrent recent self-injury recruited from clinical services, 21 women with recurrent recent self-injury but no psychiatric diagnosis recruited from the community, and 23 healthy controls with no history of self-injury or psychological treatment. Only Android users were included because of the app’s technical requirements. Fifty-five participants completed at least 80 percent of the scheduled assessments and formed the final analytical sample of 17 in the comorbid group, 18 in the subclinical group and 20 healthy controls.
Socializapp was modeled on the Cyberball ostracism procedure but transplanted into a social networking environment. Participants believed they would interact with ten other study members by posting photographs and exchanging likes and comments over seven days. In reality, the other profiles were simulated and all feedback was predetermined by the research team, co-developed with a youth advisory panel. Each participant was told to avoid uploading identifiable personal photographs such as selfies, and the app included a risk alert system that notified the principal investigator when predefined triggers occurred, alongside a help icon linking to risk-management resources. After the study ended, all participants received a full debriefing revealing the simulated nature of the interaction.
The experimental manipulation unfolded in three fixed phases across the week. Days one to three constituted the Inclusion condition, in which posts received a normal flow of peer feedback. Days four and five introduced Partial Exclusion, with visibly reduced engagement, and days six and seven brought full Exclusion, with feedback largely withdrawn. The researchers chose this graded three-level design rather than Cyberball’s binary inclusion-versus-exclusion contrast because ostracism on real social media typically involves diminished engagement rather than complete silence. Twice-daily sessions were prompted by app alerts, and each session involved viewing the feedback environment, completing a brief momentary questionnaire, and posting a new photograph.
The momentary assessments captured perceived exclusion with three items asking how ignored, excluded and rejected participants felt, which showed good internal consistency. Perceived approval was measured with a single item asking to what extent participants thought others liked their photo. Affect was tracked with two items each for negative states, sadness and anger, and positive states, happiness and calm. A binary item asked whether participants were thinking about self-injuring right now, and another asked whether they had self-injured in the last 24 hours. Compliance was high: 667 of 770 scheduled assessments were completed, an overall rate of 86.62 percent, and compliance did not differ significantly across groups, conditions or study days.
Statistical modeling using linear and generalized linear mixed-effects models confirmed that the manipulation registered as intended. Perceived exclusion rose stepwise and perceived approval fell stepwise from Inclusion through Partial Exclusion to full Exclusion across all participants, regardless of whether they had self-injury histories. Full Exclusion also increased negative affect and reduced positive affect, while Partial Exclusion produced only weaker, trend-level affective changes. Behaviorally, comments, which the researchers describe as a more effortful and self-exposing form of digital interaction, declined under both exclusion conditions, whereas likes dropped significantly only under Partial Exclusion, a pattern the authors interpret as liking possibly functioning as a low-effort compensatory response under full exclusion.
The central clinical result concerned self-harm motivation. Among the 35 participants with a history of non-suicidal self-injury, full Exclusion was associated with more than a threefold increase in the odds of reporting a momentary urge to self-injure relative to the Inclusion baseline, with an odds ratio of 3.53 and a 95 percent confidence interval of 1.52 to 8.18. Partial Exclusion showed only a marginal, non-significant trend in the same direction. The researchers caution that urges were infrequent events during the brief seven-day window, so this result should be treated as exploratory.
Notably, the surge in urges did not consistently translate into behavior. Across the week, seven participants in the subclinical group and five in the comorbid group reported self-injury, and the urge-to-behavior transition varied markedly between individuals, indicating that factors beyond the immediate social-evaluative context govern whether motivation becomes action. Meanwhile, borderline severity left a distinct signature on perception. When the researchers examined responses during the objectively favorable Inclusion condition, participants with higher borderline trait scores perceived significantly greater exclusion and markedly lower approval than those with lower scores, even though the feedback they received was objectively positive. The data therefore more clearly support a sustained negative interpretation bias tied to borderline pathology than amplified reactivity to actual ostracism, for which the cross-level moderation tests were statistically underpowered and thus inconclusive.
The authors frame these results as an initial validation of Socializapp and emphasize several limitations. Conditions were presented in fixed sequential blocks, so condition effects cannot be fully separated from time-related processes such as habituation or expectancy, and no formal credibility check was administered. Monte Carlo simulations showed strong power for within-person effects but limited power for between-person interactions, meaning all cross-level findings are preliminary. The sample comprised exclusively young women, restricted to Android users, and the seven-day protocol may have been too short to capture low-frequency clinical events reliably. Because participants could not post personal photographs and the experimental comments were mild compared with real social media interactions, the paradigm may underestimate real-world effects.
Subject of Research: Experimental smartphone-based social media ostracism and its real-time effects on non-suicidal self-injury urges in young women
Article Title: Non-suicidal self-injury and digital ostracism: a novel smartphone-based paradigm capturing real-time responses to social media exclusion
Article References: Schmidt, C., Nicolaou, S., Otero, D., Marco-Pallarés, J., Castilla, D., Zaragoza, I., Soler, J., Pascual, J. C., & Vega, D. (2026). Non-suicidal self-injury and digital ostracism: a novel smartphone-based paradigm capturing real-time responses to social media exclusion. Borderline Personality Disorder and Emotion Dysregulation, 13(1), Article 35. https://doi.org/10.1186/s40479-026-00364-0
Image Credits: AI Generated
DOI: 10.1186/s40479-026-00364-0
Keywords: non-suicidal self-injury, borderline personality disorder, digital ostracism, social media exclusion, ecological momentary assessment, smartphone paradigm, self-harm urges, emotion dysregulation, online engagement, experimental psychology, Non-suicidal, self-injury
News Source: Glenn Wilkins. (November 5, 2026). Research report: Non-suicidal self-injury and digital ostracism: a novel smartphone-based paradigm capturing real-time responses to social media exclusion. Scienmag.
