Music and AI Workshop

Sound Design, User Research, Data Science, Human-Centered AI.

Github page. 

Youform (online survey data collection), Mentimeter (real-time data collection), Python (Jupyter Notebook for data cleaning, analysis, and visualization).

This project was part of the research initiative 'Sound Environment in Retail: A Cross-Industry Study on Sound Design and Music Strategies'. The goal of this work is to explore how background music and environmental sounds in retail spaces influence employee-customer interactions, as well as their impact on satisfaction and well-being.

This workshop aimed to explore the impact of sound environments on retail customer experience with representatives from the retail and wholesale sectors. A key question addressed was: 'Is generative AI the solution for in-store music, or does it risk overengineering and alienating customers?' Through hands-on activities, participants discussed how music shapes atmosphere and consumer behavior, while considering the role of AI-generated content in this context.

Pre-Workshop Survey: An online survey was distributed to retail and wholesale professionals to assess their attitudes towards AI-powered music solutions. G-tests were used to analyze categorical data, with Benjamini-Hochberg FDR corrections applied to control for multiple comparisons.

On-Site Quiz: A quiz was conducted during the workshop to evaluate the participants' ability to distinguish between sounds and images from specific stores, and human-created vs. AI-generated music. Cochran's Q test and McNemar's test were employed to assess differences in performance.

Survey data was analyzed in real-time during the workshop to provide immediate insights and inform discussions. 

Pre-workshop survey results (n=42) provided insights into preferences for AI-powered in-store music solutions. Participants strongly preferred human-made music over AI-generated music. Music in stores was significantly preferred to silence. While there was some interest in personalization, participants also expressed concerns about data privacy.

The on-site quiz (n=13) revealed significant performance differences between image-based and sound-based questions. The performance matrix above visually illustrates this, showing a clearer diagonal pattern for image-based questions, indicative of higher accuracy (see Figures 2-3).

These findings highlight the potential of AI-powered music solutions to enhance in-store experiences. However, it's crucial to balance personalization with privacy considerations to ensure consumer trust and satisfaction in such contexts.