Ethical gaps in the ethics of artificial intelligence

The development of neural networks and their active implementation in various spheres of the economy have raised the challenge of ethical evaluation and legal regulation of the development and application of artificial intelligence (AI) systems. Over the last decade, more than 200 ethical codes and recommendations have been created worldwide, which are actively discussed within the framework of scientific research. The majority of these documents are based on a principle-based approach that sets the boundaries of ethics in the field of AI. This article examines the ethical gaps in artificial intelligence ethics through the development of a multi-level system of strategies to overcome them. The subject of the research is specific mechanisms and strategies for transforming abstract ethical principles into functioning practices at each level. This article discusses the main approaches to addressing issues in artificial intelligence ethics. A systemic approach is proposed as a methodological basis, integrating technical solutions, organizational changes, and legal mechanisms. The applicability of the proposed approach is demonstrated through the analysis of key ethical issues in AI: systematic reproduction of biases, the "black box" problem, distribution of responsibility, the declarative nature of ethical and legal norms, and the challenges of responsibility attribution. For each problem, specific methods of resolution are outlined – from the creation of counterfactual fairness algorithms to checklists for assessing the ethics of AI systems. Conclusions are drawn that a three-level "filter" system is necessary for systematically addressing ethical gaps in AI ethics, which involves differentiating approaches at the development level, organizational changes, creating new principles for state regulation, and a shift in the paradigm of responsibility distribution. The main conclusion of the research is the need for a comprehensive systemic approach, where the rigor of regulation is proportional to the potential harm from the use of the AI system.

artificial intelligence ethics, black box, neural networks, ethical codes, counterfactual fairness, AI principles, ethical gaps, AI ethics, AI audit, interpretability problem