Quality of Models for Man and Machine

This open access book investigates the renewed role of models in the solutions supporting our organisational reality. With AI and other advanced IT as digital actors, we are more and more delegating decisions that help to create new social reality to autonomous artificial systems. This raises several critical questions, for example, how can we keep our skills in building and discussing abstractions to create a unified view of the existing and emerging social and digital reality? How can we build autonomous information systems using these abstractions that we understand and trust, i.e., systems that reflect our values, and that are semantically transparent and able to explain their decisions? In recent years, the main focus on AI has been machine learning and generative AI with limited explainability. In the future, AI will combine traditional symbolic techniques (including conceptual modelling and knowledge graphs) with sub-symbolic techniques (neural networks) in neuro-symbolic computing. Supporting modelling of the people, by the people, for the people becomes, in this situation, necessary to empower humans to stay as active actors in the loop of AI-enhanced decision-making and action. To understand how work in neuro-symbolic computing can progress, we should clarify what it means for both the symbolic and machine learning models to be good. What is the quality of the different types of models, and how do these notions of quality align with human notions of quality? The quality of conceptual models (as an example of symbolic models) has been investigated for many decades. In this book, we will combine works from semiotics, information systems, conceptual modelling, AI and neuroscience to investigate the quality of models and data for both human and machine actors, laying a firm foundation for the future converging neuro-symbolic computing.