cs.HCAug 29, 2026
SaveHow Mental Health Self-Disclosure Becomes Visible: Evidence from Eight Conditions on Reddit
Organizations: College of Information Science University of Arizona Tucson, Arizona, USA · Department of Computer and Information Science and Engineering University of Florida Gainesville, Florida, USA
Abstract
People share mental health diagnoses on social media, yet how such language becomes visible around their self-disclosure, and whether community engagement tracks it, remain unexamined across conditions. We analyze 89,605 Reddit posts from 739 users across eight conditions, removing each user's diagnosis disclosure and aligning their surrounding posts to that anchor. Within the pre-disclosure year, language-visible burden was highest in the month before disclosure for six conditions, earlier for post-traumatic stress disorder and furthest from it for borderline personality disorder, and remained visible afterward rather than resolving. The theme Seeking Clinical Explanations showed the largest early-to-late difference before disclosure in five conditions, yet engagement rarely tracked what users wrote: only 9 of 360 language--engagement correlations survived correction. Disclosure is therefore a waypoint in an unevenly visible process, and we offer implications for community practice and platform design where engagement metrics do not reflect clinical need.
Figures & tables
Figure 1. Analysis overview: one disclosure anchor is selected and removed per author, surrounding posts are positioned in eight event-time windows, post-level measures are aggregated into author-window and condition-window summaries (RQ1–RQ2), and RQ3 correlates 15 language indicators with three Reddit engagement fields in the same author-window. Two-stage overview of data analysis. The first stage shows an author's surrounding posts positioned in eight event-time windows around a removed disclosure anchor and presents the event-time equation. The second stage maps RQ1 to the burden dimension-share score, RQ2 to burden-dimension and primary-theme measures, and RQ3 to same-window correlations between language indicators and Reddit engagement fields.
| Condition | Authors | Pre-disclosure maximum window | Mean score at that window | Pre-disclosure range | Leading late-pre burden dimensions | Language–engagement correlation summary |
|---|---|---|---|---|---|---|
| ADHD | 389 | Immediate-pre | 0.009 | 0.007 | Organizational Lapses (0.013); Hyperactive Impulsivity (0.008) | 5 Pearson correlations meet the BH-adjusted threshold; the largest absolute correlation is the Imagining Self-Removal as an Escape From Distress indicator with comment count ( , ). |
| Autism | 79 | Immediate-pre | 0.016 | 0.013 | Social Camouflaging (0.020); Sensory Overload (0.016) | — |
| Depression | 63 | Immediate-pre | 0.028 | 0.022 | Low Mood (0.023); Self-Devaluation (0.023) | — |
| OCD | 55 | Immediate-pre | 0.047 | 0.033 | Reassurance Seeking (0.064); Distressing Interference (0.049) | — |
| PTSD | 61 | Near-pre | 0.026 | 0.013 | Involuntary Reexperiencing (0.028); Avoidant Shutdown (0.026) | — |
| Bipolar disorder | 39 | Immediate-pre | 0.049 | 0.040 | Impulsive Risk and Crash (0.078); Manic Activation (0.048) | — |
Table 2. Condition Profiles Around Self-Disclosure. Note. Authors counts all retained authors per condition ( ); the window-based columns draw on the 728 authors with posts inside the eight event-time windows. The score and range columns report the burden dimension-share score. Leading late-pre burden dimensions reports the two largest pooled means across the near-pre and immediate-pre windows. An em dash indicates that no Pearson correlation meets the BH-adjusted threshold.
Figure 2. Pre-disclosure peak timing by condition. The filled bubble marks the window with the highest condition-window mean of the burden dimension-share score, and bubble size indicates that mean; pale gray dots show the other pre-disclosure windows. Figure 3 shows all eight windows. Peak timing map showing four pre-disclosure windows for eight mental health conditions. A filled colored bubble marks the highest condition-window mean of the burden dimension-share score for each condition, while pale gray dots show the other pre-disclosure windows. The far-pre window contains the highest condition-window mean for BPD, the near-pre window contains the highest mean for PTSD, and the immediate-pre window contains the highest mean for the other six conditions.
Figure 3. Burden-dimension hit rates by event-time window. Each cell is the condition-window mean of author-level post hit rates for one condition-specific dimension. Heatmap of condition-window means for condition-specific burden-dimension hit rates across event-time windows. Rows are grouped by condition, and each cell averages authors' within-window post hit rates.
Figure 4. Pooled late-pre-minus-early-pre differences for the ten primary-theme indicators. Column labels abbreviate the primary-theme names in Table 8 , ordered by mean difference across conditions. Grid of horizontal bars showing pooled late-pre-minus-early-pre differences for ten primary-theme indicators. The Seeking Clinical Explanations theme has the largest absolute difference in five condition profiles; the exhaustion theme does so for anxiety, the personal-defectiveness theme for depression, and the uncertainty theme for OCD.
Figure 5. The 9 of 360 same-window Pearson correlations with BH-adjusted . Each row names the condition, the language indicator, and the engagement field; theme names are abbreviated as in Table 8 . Dotplot of same-window Pearson correlations that pass Benjamini–Hochberg q less than or equal to .05. Points right of zero indicate higher engagement as a language indicator increases, while points left of zero indicate lower engagement.
Figure 6. The mental health self-disclosure visibility framework: three dimensions of observability derived from RQ1–RQ3 (top) and their implications for online communities and platforms (bottom). Concept diagram of the mental health self-disclosure visibility framework. A header box connects to three columns: temporal visibility, interpretive visibility, and engagement-bounded visibility. Each column contains an analytical-lens box summarizing what that dimension describes and an implications box listing the corresponding design and community implications for disclosers, peer supporters, and platforms or moderators.
Appendix figures & tables4 assets
Supplementary material from the paper’s appendix.
Appendix
| Self-disclosure terms (41 phrases) | |
|---|---|
| Diagnosis statements (19) | “was / got / am / have been diagnosed with,” “officially / clinically / formally / previously diagnosed with,” “received a diagnosis of,” “received diagnosis,” “was given a diagnosis of,” “have / has / had a diagnosis of,” “has an official / clinical diagnosis of,” “carry a diagnosis of,” “was clinically / medically determined to have” |
| Clinician-referenced statements (22) | “diagnosed me with,” “diagnosed by a psychiatrist / psychologist / doctor / therapist,” “my doctor / psychiatrist / psychologist / therapist diagnosed me with,” “my doctor / psychiatrist / psychologist said / says I have / had,” “my therapist told me” |
| Mental health terms (33 terms) | |
| General terms | mental health; mental disorder; mental illness |
| Study-condition terms | ADHD ; attention deficit ; attention-deficit hyperactivity disorder ; anxiety disorder ; autism spectrum disorder ; asperger’s syndrome ; autistic ; bipolar disorder ; manic depression ; cyclothymia ; depression ; depressive disorder ; MDD ; OCD ; obsessive-compulsive disorder ; PTSD ; post-traumatic stress disorder |
| Other queried terms | schizophrenia; paranoid schizophrenia; schizo; panic disorder; social phobia; eating disorder; anorexia; bulimia; personality disorder; idiopathic developmental intellectual disability; conduct disorder; anger issues; aggressive behavior |
Table 3. Search terms for identifying candidate self-disclosure posts.
| Output reference | Measurement object | Equation reference codes | Grounding references |
|---|---|---|---|
| Table 2 ; Figure 2 | Condition profile and burden dimension-share score | Event-time indexing in Eq. ( A1 ); burden-dimension hit and dimension-share score in Eqs. ( A7 )–( A8 ); author-window aggregation and intervals in Eqs. ( A15 )–( A16 ) | Longitudinal event-time analysis ( Singer and Willett, 2003 ; De Choudhury et al., 2013 ) ; proxy measurement and construct validity ( Grimmer and Stewart, 2013 ; Jacobs and Wallach, 2021 ; Messick, 1995 ) |
| Figure 3 | Burden-dimension hit-rate heatmap | Burden-dimension hits in Eq. ( A7 ); author-window aggregation in Eq. ( A15 ) | Clinical-domain grounding for the study-specific dimensions ( American Psychiatric Association, 2022 ; Baron-Cohen et al., 2001 ; Blevins et al., 2015 ; Bohus et al., 2009 ; Goodman et al., 1989 ; Hirschfeld et al., 2000 ; Hull et al., 2017 ; Kessler et al., 2005 ; Kroenke et al., 2001 ; Spitzer et al., 2006 ) |
| Figure 4 | Primary-theme indicator difference | Primary-theme hit in Eq. ( A14 ); author-window aggregation in Eq. ( A15 ); pooled mean in Eq. ( A17 ) | Thematic analysis and computational grounded theory ( Braun and Clarke, 2006 ; Nelson, 2020 ) |
| Table 4 ; Figure 5 | Same-window correlations with Reddit engagement fields | Pearson correlation in Eq. ( A20 ); presence share in Eq. ( A21 ); burden dimension-share score, affect indicators, and primary-theme indicators in Eqs. ( A8 ), ( A11 )–( A13 ), and ( A14 ); BH adjustment in Eq. ( A24 ) | Research on Reddit response and support quality ( Andalibi et al., 2017 ; De Choudhury and De, 2014 ) ; Benjamini–Hochberg adjustment ( Benjamini and Hochberg, 1995 ) ; affect lexicon traditions ( Hutto and Gilbert, 2014 ; Mohammad and Turney, 2013 ; Pennebaker et al., 2015 ) |
| Table 5 | Author-level sensitivity correlations | Author-level correlation in Eq. ( A22 ); BH adjustment in Eq. ( A24 ) | Observational social-media data research ( Olteanu et al., 2019 ; Ruths and Pfeffer, 2014 ) ; Benjamini–Hochberg adjustment ( Benjamini and Hochberg, 1995 ) |
| Exploratory lagged check in Section 4 | Next-window correlation with Reddit engagement fields | Lagged correlation in Eq. ( A23 ); BH adjustment in Eq. ( A24 ) | Longitudinal contrast without causal interpretation ( Singer and Willett, 2003 ) ; observational social-media data research ( Olteanu et al., 2019 ; Ruths and Pfeffer, 2014 ) |
Table 6. Appendix Measurement-to-Output Reference Map. Note. This table links each manuscript table or visualization to the appendix equations and citations used to define or calculate the relevant measurements.
| Condition | Dimensions and definitions | Sources |
|---|---|---|
| ADHD | Inattention: difficulty sustaining attention or keeping the mind from wandering. Task Follow-Through: difficulty starting or finishing tasks. Organizational Lapses: forgetting, losing track, missing deadlines, or difficulty organizing. Hyperactive Impulsivity: fidgeting, difficulty sitting still, impulsivity, or interrupting. | Adult ADHD Self-Report Scale (ASRS) items ( Kessler et al., 2005 ) . |
| Autism | Social Opacity: difficulty interpreting social cues, rules, or intentions. Social Camouflaging: hiding autistic traits or consciously compensating in social settings. Sensory Overload: sensory sensitivity, overstimulation, meltdown, or shutdown. Change Intolerance: reliance on predictability or difficulty with transitions and change. | Autism-Spectrum Quotient (AQ) social and attention-switching content, the Diagnostic and Statistical Manual of Mental Disorders , Fifth Edition, Text Revision (DSM-5-TR), and camouflaging research ( American Psychiatric Association, 2022 ; Baron-Cohen et al., 2001 ; Hull et al., 2017 ) . |
| Depression | Low Mood: sadness or sustained low-mood language. Anhedonia: reduced interest or pleasure, including emotional numbness when it conveys diminished pleasure. Functional Fatigue: low energy, exhaustion, or difficulty with everyday activities. Self-Devaluation: self-devaluation, failure, guilt, shame, or describing oneself as a burden. Self-Harm Ideation: self-harm, suicidal thinking, or not wanting to live. | Patient Health Questionnaire-9 (PHQ-9) items and DSM-5-TR depressive descriptions ( American Psychiatric Association, 2022 ; Kroenke et al., 2001 ) . |
| OCD | Intrusive Doubt: unwanted obsessive thoughts, persistent doubt, or difficulty stopping a concern. Compulsive Rituals: repeated checking, compulsions, or acts intended to reduce discomfort. Reassurance Seeking: seeking reassurance or complete confirmation. Distressing Interference: language that describes obsessive-compulsive experiences as distressing or disruptive. | Yale–Brown Obsessive Compulsive Scale (Y-BOCS) obsession, compulsion, interference, and distress structure, interpreted with DSM-5-TR OCD descriptions ( American Psychiatric Association, 2022 ; Goodman et al., 1989 ) . |
| PTSD | Involuntary Reexperiencing: flashbacks, nightmares, triggers, or involuntary reexperiencing. Avoidant Shutdown: avoiding reminders, shutting down, dissociating, or emotional numbing. Negative Trauma Cognitions: trauma-related guilt, shame, blame, or persistent beliefs that the world is unsafe. Hyperarousal: hypervigilance, exaggerated startle, sleep disruption, or angry outbursts. | PTSD Checklist for DSM-5 (PCL-5) items and DSM-5-TR PTSD domains ( American Psychiatric Association, 2022 ; Blevins et al., 2015 ) . |
| Bipolar disorder | Decreased Sleep Need: reduced need for sleep or remaining awake without feeling tired, rather than generic insomnia alone. Accelerated Thinking: accelerated thinking, rapid speech, or unusually many ideas. Manic Activation: manic or hypomanic activation, euphoria, or unusually high energy. Impulsive Risk and Crash: risky or impulsive action during activation, or an abrupt drop in mood or energy after a period of activation. | Mood Disorder Questionnaire (MDQ) activation items and DSM-5-TR bipolar descriptions ( American Psychiatric Association, 2022 ; Hirschfeld et al., 2000 ) . |
Table 7. Condition-Specific Burden Dimensions. Note. Definitions describe the post language each dimension is intended to capture. Literature informed the domains, but this study’s dimensions and search terms do not reproduce instrument subscales or scores.
| Primary theme and post-level assignment rate | Definition | Representative quotation |
|---|---|---|
| Seeking Clinical Explanations = 8,254; 9.21% | Users invoke diagnoses, therapy, medication, or other clinical language to interpret their experiences and make them more legible and legitimate. | “I’m starting treatment soon … I recently got diagnosed after a long uphill battle.” |
| Anticipating Rejection or Abandonment in Close Relationships = 5,035; 5.62% | Users describe close relationships as vulnerable to rejection, betrayal, abandonment, or sudden emotional harm. | “How do you not spiral when something directly triggers your fear of abandonment?” |
| Interpreting Social Experiences as Evidence of Personal Defectiveness = 3,800; 4.24% | Users interpret social interactions as confirmation that they are defective, unlikable, or fundamentally unable to belong. | “I don’t feel like I fit in. I’ve never really felt like I fit in anywhere.” |
| Feeling Too Exhausted for Daily Functioning = 3,452; 3.85% | Users describe persistent depletion that makes ordinary daily functioning feel difficult or unsustainable. | “I’ve been so fatigued all the time and always seem to have brain fog.” |
| Trying to Eliminate Uncertainty Through Checking and Reassurance Seeking = 2,151; 2.40% | Users describe checking, seeking reassurance, or repeatedly reviewing thoughts and events in attempts to resolve uncertainty. | “I’m open to some small reassurance on this issue. However, there might not be any practical reassurance.” |
| Imagining Self-Removal as an Escape From Distress = 1,467; 1.64% | Users describe withdrawal, disappearance, self-harm, or suicide as imagined ways of escaping overwhelming distress. | “I do not trust myself to not act upon my suicidal thoughts. I need help and I know that.” |
Table 8. Primary-Theme Codebook and Post-Level Assignment Rates. Note. The post-level assignment rate is the share of the 89,605 surrounding posts with an automatic theme hit. A post may match more than one theme and most posts match none, so the rates neither sum to 100 percent nor partition the corpus. Themes are ordered by this rate.