Exploring the Alignment of AI-Based Mood Labelling with Human Responses: Implications for Music-Based Mental Health...

Area of Science:

• Psychology
• Computer Science
• Music Therapy

Background:

• Music therapy enhances emotional well-being and quality of life.
• Artificial intelligence (AI) is increasingly applied to music analysis and generation.
• Understanding AI's alignment with human emotional responses is vital for therapeutic applications.

Purpose of the Study:

• To compare AI-based mood analysis with human physiological responses (skin conductance).
• To investigate the accuracy of AI in identifying emotions in music for therapeutic contexts.

Main Methods:

• 21 participants' skin conductance responses were recorded.
• An AI tool analyzed a negatively valenced classical music excerpt for mood.
• AI-generated mood probabilities were compared with participant physiological data.

Main Results:

• A discrepancy was found between the AI's dominant emotion and the music's intended emotion.
• Significant alignment was observed between AI mood probability and participants' skin conductance.
• AI emotion recognition shows potential for supporting music selection in interventions.

Conclusions:

• AI-based emotion recognition systems show potential for music therapy applications.
• Careful human validation is necessary to ensure AI tools align with therapeutic goals.
• Further research is needed to refine AI for accurate emotional assessment in music therapy.