Abstract
This study explores the concept of emotivation in human-robot interaction, defined as the interplay between emotional perception and motivational drive, both in guiding the robot’s behavior and in promoting the user’s transition toward a more positive emotional state. Within this framework, we present a humanoid robotic system designed not only to recognize human emotions but also to act upon them with the goal of human emotional uplift and sustained well-being. The system integrates multimodal data (EEG signals from an Empatica EPOC X helmet and audio-visual input from the robot) processed through two deep learning models: a hybrid-fusion classifier for facial and vocal expressions, and a Feature-Based Convolutional Neural Network (FBCNN) for EEG data. A meta-model combines its outputs to classify emotional states into four categories: neutral, happy, angry, and sad. Crucially, this work focuses on defining which robot behaviors can be considered emotively motivated — that is, capable of responding meaningfully to the user’s current emotional state while also being motivated by the goal of shifting negative emotions toward positive ones, and sustaining those positive states over time. The robot dynamically selects behavioral responses such as engagement, encouragement, and affective reinforcement, based on the emotional context. Results demonstrate the effectiveness of the emotivational framework in promoting positive emotional transitions, highlighting the importance of emotional sensitivity and motivational purpose in fostering empathic, emotionally supportive human-robot relationships. Future work will further develop adaptive strategies to enhance emotional resilience and companionship through long-term interactions.