On the compliance with ethical principles in AI

In recent years, there has been a lot of discussion around ethics in IT and AI. Researchers and organizations have proposed guidelines to address privacy, fairness, and explainability challenges for creating trustworthy AI. In this work, we outline the importance of compliance with the above-mentioned ethical principles and their influence on the quality of AI systems. We map the relationship between available approaches for compliance with privacy, fairness, explainability principles and the accuracy of AI system decisions. Additionally, we introduce the difference between ensuring fairness for phenomena presented with tabular data and text. Tabular data may contain protected attributes such as gender, age, or race as well as the decision made historically in relation to the people concerned. Data presented in text is not structured and requires sense perception by AI systems to detect bias or unfairness. In the poster, we compare available approaches and present experiments for measuring bias in text data.