Data Management

The speed and accuracy of data management are essential advantages offered by AI systems. A further advantage could be if the data are transformed to insightful solutions facilitating business performance and end-user applications. The current chapter addresses how to possibly generate such solutions, translating user needs into explainable data architectures. Understanding how to generate, train, test, and optimise AI-generated behaviour is also in the focus hereby. Machine behaviour could be navigated by exposing AI systems to specific training data. While substantial human effort was needed to annotate, characterise and interpret information, the enhancement of autonomous capabilities could mark a new era in data management. In this respect, classification algorithms for text, voice, and images are trained to optimise accuracy on a specific set of human-labelled datasets. Most importantly, the selection, labelling, and management of a particular dataset and the chosen features can reshape not only the behaviour of an AI system. Rather, the user behaviour could be modified by the way the system is trained. However, data management may experience some bias, but this is a call to rethink AI systems, in order to preclude biased responses. We further recommend remediation of the AI systems currently available on the market. As seen from the outcomes of the field studies reported hereby, informativeness, accuracy, and competence are crucial parameters determining proper system functioning, and thus its adoption by users. Therefore, by fine-tuning algorithms and data architectures, an effective approach for data management is expected to be created to appropriately meet the user expectations and business demands for transformational AI solutions.