Ethical Considerations in Affective Computing

This work delves into the critical role of explainable artificial intelligence (XAI) in enhancing ethical considerations within affective computing (AC), spanning across three pivotal stages: pre-XAI, in-XAI, and post-XAI. At the in-XAI stage, the focus is on elucidating the model’s internal mechanics, including feature utilization, determination of feature importance, and the decision-making algorithms, leveraging deep learning and various feature extraction methods. Conversely, the post-XAI stage concentrates on providing end users with transparent and comprehensible explanations of the model’s emotional predictions, employing techniques like LIME, SHAP, and natural language explanations. These methodologies collectively aim to bolster transparency, fairness, and trust, thereby ensuring ethical integrity in AC systems.