"I'm gonna KMS": From Imminent Risk to Youth Joking about Suicide and Self-Harm via Social Media

Title of the Paper

“I’m gonna KMS:” From Imminent Risk to Youth Joking about Suicide and Self-Harm via Social Media

Bibliographic Information

• Topic Area: Representation, impact, and recommendations regarding youth discussions of self-harm and suicide on social media
• Keywords: youth, social media, self-harm, suicide, mental health, Instagram, privacy, peer support, risk detection

Research Background and Problem Statement

• Identified Issues or Challenges:

  • The sharing of self-harm and suicide-related content by youth on social media platforms may lead to the spread of dangerous behaviors (e.g., suicide contagion) or the normalization of self-harm, posing a public health concern.
  • Existing studies in the Human-Computer Interaction (HCI) field often rely on public data or surveys, which may not adequately capture how youth discuss these topics in private conversations.
  • Current technologies struggle to differentiate between youth using "suicide language" humorously or as exaggerated expressions versus genuinely dangerous scenarios, resulting in high false-positive rates.

• Research Motivation and Related Work:

  • Youth suicide is a critical global public health issue.
  • Previous studies have primarily focused on the representation and impact of youth self-harm and suicide behaviors on public social media. However, private conversations remain underexplored.
  • The authors aim to analyze private Instagram messages to uncover how youth genuinely discuss self-harm and suicide, thereby aiding the development of timely and accurate risk detection and intervention measures.

Proposed Solution

• Methodology or Proposed Solution:

  • Utilize an ecological sample dataset donated by youth users, consisting of private Instagram conversations.
  • Conduct qualitative research using thematic analysis to identify key patterns, themes, and contexts in youth discussions about self-harm and suicide.

• Innovative Contributions:

  • The first study to examine youth private conversations about self-harm and suicide, uncovering deeper meanings behind their language.
  • Propose design recommendations to balance supporting healthy discussions among youth while monitoring or limiting online dangerous content, avoiding excessive censorship or false alarms.

• Implementation Steps and Key Techniques:

  • Dataset filtering: Extract 1,224 conversations related to self-harm/suicide from English-language youth Instagram messages and divide them into 2,019 sub-conversations.
  • Analytical approach: Conduct thematic analysis of conversation content, categorized by the following research questions (RQs):
    1. How do youth discuss their personal experiences with self-harm/suicide?
    2. How do youth discuss third-party self-harm/suicide behaviors?
    3. How do youth use "self-harm/suicide language" humorously or exaggeratedly?

Research Findings

• Specific Findings:

  • RQ Summaries:

    • RQ1 (Personal experiences): 23% of sub-conversations involved youth disclosing their current or past self-harm/suicide behaviors, including sharing immediate dangerous thoughts, reflecting on past actions, or celebrating recovery.
    • RQ2 (Third-party experiences): 7% of conversations involved gossiping about or intervening in others' behaviors, defending against or technically avoiding exposure to triggering content.
    • RQ3 (Humor and exaggerated expressions): 70% of sub-conversations used "self-harm/suicide language" as a form of humor or emotional expression rather than indicating genuine risk.

  • Proposed specific design recommendations for social media platforms:

    • Create "trust circles" to provide youth with a safe and supportive community.
    • Implement context-aware interventions to offer timely assistance and support resources in dangerous situations.
    • Encourage youth to adopt less violent language through educational nudges (e.g., "prompt design").

• Advantages Over Existing Solutions:

  • Provides nuanced insights into private online conversations, complementing existing studies focused solely on public data.
  • Guides social media platforms to design algorithms that better account for context and linguistic complexity, reducing false positives.

• Experimental or Evaluation Results:

  • Data revealed that exaggerated and humorous "self-harm/suicide language" is a common communication method among youth, potentially masking genuinely dangerous behaviors.
  • Youth often rely on peer support, including comforting friends in risky situations or creating mutual intervention plans.

• Limitations and Future Directions:

  • Data is solely from Instagram and limited to U.S. users, potentially excluding other platforms or broader cultural contexts.
  • No comprehensive analysis exists on the differences in "suicide language" usage across age groups or cultural backgrounds.
  • Future research is recommended to explore other platforms and incorporate more diverse youth datasets to validate the findings' generalizability.