Modern artificial intelligence provides highly efficient systems that can easily outperform humans in various kinds of tasks. The core of such methods is largely based on modern machine learning. Perhaps the main problem with the use of such methods is that while they are efficient, they rarely explain why and how the related results are produced. Thus we cannot use these methods in tasks relating to, for example, very high-stakes decisions or delicate ethical issues. Explanations are crucial. This project contributes to finding methods that explain decisions based on machine learning devices. Furthermore, we develop novel frameworks that attempt to preserve the efficiency of machine learning, but also produce explanations of different decisions up to different levels of accuracy. The explanations are language-based - formulas of modern formal logic or logic-based deductions.

2022

2024