Su, D., Picard, R., Liu, Y. "AMAI: Adaptive Music for Affect Improvement." Proceedings of the 44th International Computer Music Conference (ICMC), Daegu, Korea, August 2018.

Abstract

This paper introduces Adaptive Music for Affect Improvement (AMAI), a music generation and playback system whose goal is to steer the listener towards a state of more positive affect. AMAI utilizes techniques from game music in order to adjust elements of the music being heard; such adjustments are made adaptively in response to the valence levels of the listener as measured via facial expression and emotion detection. A user study involving AMAI was conducted, with N=19 participants across three groups, one for each strategy of Discharge, Diversion, and Discharge → Diversion. Significant differences in valence levels between music-related stages of the study were found between the three groups, with Discharge → Diversion exhibiting the greatest increase in valence, followed by Diversion and finally Discharge. Significant differences in positive affect between groups were also found in one before-music and after-music pair of self-reported affect surveys, with Discharge → Diversion exhibiting the greatest decrease in positive affect, followed by Diversion and finally Discharge; the resulting differences in facial expression valence and self-reported affect offer contrasting conclusions.