Explainable AI-based suicidal and non-suicidal ideations detection from social media text with enhanced ensemble...

Area of Science:

• Artificial Intelligence
• Computational Social Science
• Mental Health Technology

Background:

• Identifying suicidal ideation on social media is critical for suicide prevention.
• Traditional AI models lack transparency in their decision-making processes.
• Explainable AI (XAI) offers interpretability, crucial for understanding AI-driven classifications.

Purpose of the Study:

• To develop a novel framework for distinguishing between suicidal and non-suicidal ideation on social media.
• To integrate Explainable AI (XAI) with ensemble machine learning methods for improved accuracy and interpretability.
• To enhance the reliability of AI systems for monitoring and intervening in online suicide-related discussions.

Main Methods:

• Utilized an ensemble technique combining multiple machine learning algorithms.
• Incorporated Explainable AI (XAI) to analyze and interpret model classifications.
• Evaluated the framework on diverse social media datasets against state-of-the-art methods.

Main Results:

• The proposed framework demonstrated superior accuracy in detecting suicidal content compared to existing methods.
• Achieved high performance metrics: F1-score of 95.5% for suicidal ideation and 99% for non-suicidal ideation.
• The XAI component provided clear insights into the features driving the model's classifications.

Conclusions:

• The developed framework offers a more reliable and interpretable approach for identifying suicidal ideation online.
• This research bridges the gap between AI performance and explainability in mental health applications.
• The findings support the use of interpretable AI for timely intervention in suicide prevention efforts.