A Systems-Theoretic Approach for the Analysis of AI Ethics

This short research paper explores the ways in which ideas, theories, and techniques from the realm of safety research can be utilized in Human-Computer Interaction (HCI), specifically regarding the ethical aspects of artificial intelligence. Decades of dedicated effort have been invested in the domains of aviation, healthcare, and nuclear industry to enhance accident investigation and prevention methodologies. Safety-related issues can be considered as a subset of a significantly larger and more multidimensional group of ethical issues. Ultimately, both safety and ethical reflections are about avoiding unwanted consequences. A systems-theoretic causal model is one of the many theoretical approaches used in safety research. Systems-theoretic thinking that expands linear reasoning originates from engineering. By applying the systems-theoretic causal model it is possible to create causal mechanisms for explaining empirical sociotechnical phenomena. This makes it possible to perceive different types of interaction mechanisms at the system level and to compare different situations and system structures with each other. A potential benefit systems-theoretic modelling of ethical action is its ability to combine the terminology of engineering, moral philosophy and social sciences into the same framework. This allows for the development of new interdisciplinary research methods for the systematic analysis and design of ethical AI systems.