Towards the Development of Explainable Machine Learning Models to Recognize the Faces of Autistic Children

Machine learning with image classification has shown promise in supporting the detection of autism in children. To extend this line of work, our study compared the use of two algorithms to explain why an image is categorized as autistic or not. First, we trained and tested models on the Autistic Children Facial Image Data Set to identify the three that produced the highest accuracy. Following the identification of the three best models, we used two methods to examine explainability: Local Interpretable Model-agnostic Explanations (LIME) and Randomized Input Sampling for Explanation of Black-box Models (RISE). Overall, the best model produced an accuracy of 92% and LIME resulted in more explainable models than RISE. Albeit promising, researchers must conduct more studies to examine the generalizability of the results prior to the adoption of this model in practice.