The State of Ethical AI in Practice: A Multiple Case Study of Estonian Public Service Organizations

Abstract

Despite the prolific introduction of ethical frameworks, empirical research on AI ethics in the public sector is limited. This empirical research investigates how the ethics of AI is translated into practice and the challenges of its implementation by public service organizations. Using the Value Sensitive Design as a framework of inquiry, semi-structured interviews are conducted with eight public service organizations across the Estonian government that have piloted or developed an AI solution for delivering a public service. Results show that the practical application of AI ethical principles is indirectly considered and demonstrated in different ways in the design and development of the AI. However, translation of these principles varies according to the maturity of the AI and the public servant’s level of awareness, knowledge, and competences in AI. Data-related challenges persist as public service organizations work on fine-tuning their AI applications.