Applying explainable artificial intelligence methods to models for diagnosing personal traits and cognitive abilities...

Area of Science:

• Psychology
• Computer Science
• Social Media Analysis

Background:

• Understanding the link between online behavior and psychological traits is crucial.
• Previous research often focused on correlations rather than predictive indicators.

Purpose of the Study:

• To analyze psychological attributes in social media users' behavior on VK.
• To determine the implications of these attributes on user-interaction models.
• To identify key social media behaviors indicative of personality and cognitive traits.

Main Methods:

• Utilized artificial intelligence (AI) to analyze 753,252 posts from 1358 VK users.
• Integrated Big Five personality traits and intelligence test results.
• Employed integrated gradients, partial dependence plots, and Shapley values for feature analysis.

Main Results:

• Emotional tone of posts predicts Extraversion, Agreeableness, and Openness.
• Social engagement metrics correlate with Logical thinking.
• High Neuroticism linked to provocative content; religion theme relates to Consciousness and Agreeableness.

Conclusions:

• Social media behavior, particularly emotional tone and engagement, offers significant insights into psychological traits.
• AI-driven analysis can identify behavioral indicators for personality and cognitive diagnostics.
• This study shifts focus from correlation to predictive behavioral indicators in social networks.