A Reinforcement Learning-Based Decision Support Framework for Personalized Aesthetic Treatment Recommendations

This paper presents a Reinforcement Learning (RL) based decision support framework for the generation of personalised aesthetic treatment recommendations for women in Jakarta, Indonesia. The study was carried out using a Design Science Research (DSR) approach, quantitative analysis and Structural Equation Modelling (SEM). Data were collected from 300 women between the ages of 20 and 40 who are active on social media and have an interest in skincare and aesthetic treatments. The framework is built upon the integration of demographic profile, skin condition, psychological profile, social media exposure and user feedback via a Q-Learning based recommendation mechanism. The findings show that social media exposure has a positive impact on treatment intention and that influencer credibility has a positive impact on the level of trust in AI recommendations. Recommendation accuracy is positively correlated with customer satisfaction which in turn has a positive effect on repeat purchase intention. The prototype achieved an increase in recommendation accuracy from 65.5% to 90.0% over 50 training episodes and a System Usability Scale (SUS) score of 86.7. The results provide some initial evidence of the framework's learning capability and perceived usability. The work contributes to the development of an application oriented framework for adaptive and human centric aesthetic decision support, with a focus on personalisation, informed decision making and responsible AI.