Machine Learning Models Interpretations: User Demands Exploration

Automated decision making is becoming more and more popular in various domains and demonstrates high performance capabilities. The growing model complexity has limited the opportunities for understanding and justifying the model behaviour. Explainable Artificial Intelligence (XAI) has emerged to make complex models more transparent and provide insights of model behaviour. There are numerous XAI tools for implementing different types of explanations, but the majority of these tools’ outputs are quite complex and can be misused. Therefore, this research aims to make explanations more comprehensible. We plan to review existing approaches to explanation, study user needs for interpretation tools and propose the design of the tool, selecting the appropriate approach and returning explanation in a simple form.