Research Information System
in: Artificial Intelligence in Public Administration: Opportunities and Challenges, NOVA Publications , pp.219-236, 2026
The accuracy, reliability, and functionality of artificial intelligence (AI) systems largely depend on the quality of the data they are fed. AI algorithms perform their learning processes on large data sets; therefore, the accuracy, integrity, timeliness, and representativeness of the data used directly affect the success of the resulting model. While quality data enables the model to learn quickly, generalize, and cope with unexpected situations, incomplete, inconsistent, or biased data can cause the system to produce erroneous, unfair, and unethical results. Data quality is a multi-layered issue encompassing technical, social, ethical, and legal dimensions. In critical areas such as healthcare, AI systems trained with low-quality data can lead to incorrect diagnoses and treatment decisions, putting patient safety at risk. In educational systems, when data representation is lacking, inequalities of opportunity among students can deepen. These situations show that data quality has a direct impact not only on model output but also on human life. Studies conducted in recent years have shown that even advanced algorithms perform poorly when trained with low-quality data. This demonstrates how vital data engineering and data governance are in AI projects. This book chapter aims to address the importance of data quality in AI systems at a theoretical and practical level, to explain the contributions of quality data production to system success, and to raise awareness on this issue with examples from different sectors. The chapter discusses ways to build reliable, ethical, and sustainable AI systems based on data quality.