Generative Artificial Intelligence in the Lives of Young Adults: Exploring Motivations and Mental Health.

Men, Black youth, and higher SES youth used generative AI more and for more purposes than other demographic groups.

• Sample consisted of U.S. young adults ages 18-25 (N = 1003; 56.3% women, 39.4% men, 4.3% another gender)
• Cross-sectional online survey design was used
• Sociodemographic comparisons were made by gender, race/ethnicity, and socioeconomic status (SES)
• No specific effect sizes were reported for these demographic differences in the abstract

Using generative AI for social-emotional support was associated with higher depressive symptoms.

• Standardized beta coefficients ranged from 0.22 to 0.29 (ps < 0.004)
• This pattern was observed across the full sample path models
• The association suggests that reliance on genAI for relational support 'may exacerbate (or reflect pre-existing) internalizing symptoms'
• The cross-sectional design prevents determination of causal direction

Using generative AI for social-emotional support was associated with higher anxiety symptoms.

• Standardized beta coefficients ranged from 0.22 to 0.30 (ps < 0.008)
• Effect sizes were comparable in magnitude to those observed for depressive symptoms
• Among men specifically, social-emotional support motivations predicted greater loneliness (β = 0.33; p < .001)

Using generative AI for learning and exploration was associated with lower depression, anxiety, and loneliness.

• Lower depressive symptoms: βs = -0.20 to -0.40 (ps < 0.008)
• Lower anxiety symptoms: βs = -0.26 to -0.26 (ps < 0.003)
• Lower loneliness: βs = -0.28 to -0.42 (ps < 0.001)
• These associations were described as potentially 'adaptive' in the discussion

Among women, task automation and dating/sexuality motivations for using generative AI were associated with poorer mental health.

• Standardized beta coefficients ranged from 0.12 to 0.19 (ps < 0.05) for these gender-specific associations
• This pattern was identified through multiple group comparisons by gender
• No equivalent associations for task automation or dating/sexuality motivations were reported for men
• The authors describe these as 'distinct vulnerabilities by gender'

No significant moderation effects on mental health associations were observed by race/ethnicity or socioeconomic status.

• Multiple group comparisons tested moderation by gender, race/ethnicity, and SES
• Gender moderation was significant, but race/ethnicity and SES moderation were not
• This occurred despite finding that Black youth and higher SES youth used genAI more and for more purposes

The study identified four distinct motivations for generative AI use: social-emotional support, task automation, learning/exploration, and dating/sexuality.

• These four motivation categories were used as predictors in path models
• Path models tested associations between each motivation type and internalizing symptoms
• The study design was cross-sectional and observational
• Sample was recruited via online survey from U.S. young adults ages 18-25