Trustworthy Machine Emotion Intelligence Using Facial Micro-expressions

Affective computing involves machines interpreting and responding to human emotions in sensitive domains such as healthcare, education, and law. While emotion recognition technologies exist, they must show trust as developed intelligent systems. The work proposal in this paper uses facial micro-expressions within facial emotions to construct reliable and explainable vision intelligent systems based on responsible Artificial Intelligence (AI) and data practices. Three Explainable AI (XAI) techniques were used to mitigate potential bias and explain decision-making: LIME, GradCAM, and Occlusion sensitivity. A custom Convolutional Neural Network (CNN) model and AlexNet achieved 84% and 83% accuracy, respectively, with both models achieving 83% prediction accuracy on test datasets. The work highlights the need to adopt responsible AI methods to build scalable, generalizable machine vision intelligent systems and integrate computer vision explainers for ethical considerations in affective computing technologies.