AI ethics loses its significance without ethics research

There is a great need for ethics regarding the development, implementation, and use of artificial intelligence – a need for AI ethics. This form of ethics is typically framed in terms of certain ethical principles, such as fairness, transparency, and accountability.

Identifying relevant principles is a first step toward a well-functioning framework for AI ethics. But what do abstract concepts like “fairness” actually mean in the concrete contexts where people live, work, and use AI? If ethical principles are to make a real difference – rather than merely creating an ethical halo around the technology – they must be more closely linked to these practical contexts. But where, exactly, do we find the context for AI ethics?

Can the context of AI ethics be identified with the characteristics of the AI ​​technology itself – for instance, the fact that AI systems are trained on data? Does AI fairness mean that AI systems should be trained impartially to ensure that certain groups are not disadvantaged? Does AI transparency refer to openness regarding how AI systems are trained and how they function? Does AI accountability concern clarity about who is responsible when errors occur? Is this what AI ethics is about – these specific aspects of the technology?

Certainly, AI ethics can be said to concern these aspects of the technology – but in reality, it is impossible to demarcate the context of AI ethics. AI systems are being developed on a broad front for countless applications: in healthcare, education, pharmaceutical development, finance and banking, commerce, transport, and much more. The ethical challenges within these diverse domains tend to be unique and unexpected. AI ethics must keep up with the development and adapt to each new domain of ​​application. Yet even healthcare is not a single, well-defined domain; there are numerous distinct uses of AI within the healthcare sector alone, and more are on the way. The technology appears to have an almost limitless range of possible applications and is therefore transformative. Consequently, AI ethics lacks a given context.

Furthermore, mere adherence to principles is ethically insufficient. Implementing an AI system in a specific domain can be unethical even if the system satisfies all the core principles of AI ethics, as the application may still lead to unacceptable consequences in that particular domain. For instance, by affecting schoolchildren’s ability to read, write, and think independently, in the case of educational AI applications. There is, therefore, a gap between the core principles of AI ethics and the specific contexts in which the technology is applied – contexts that often give rise to unexpected challenges requiring an ethically considered response.

This raises a critically important question: How can we ensure that AI ethics keeps pace with the multifarious applications of the technology? This question is explored in an article by Jennifer Viberg Johansson and Jessica Nihlén Fahlquist, published in the journal, AI and Ethics. One might think that the authors simplify the issue by narrowing its scope to AI ethics as applied within healthcare. Yet, as noted, even healthcare does not represent a single, clearly defined domain for AI. AI ethics has no fixed context – not even within healthcare. The question of how AI ethics keeps pace with AI applications is just as significant and challenging in healthcare as it is everywhere else in society. How do the authors answer the important question?

The answer provided by Jennifer Viberg Johansson and Jessica Nihlén Fahlquist may come as a surprise to those who believe that ethics mainly consists of principles that should be followed in practice – with ethicists acting as experts who could recite these principles in their sleep. In fact, ethics is, among other things, an ongoing research endeavor that continuously updates ethical principles and concepts in relation to currently relevant contexts. It is through such ethics research that AI ethics can keep pace with the diverse range of AI applications. By integrating empirical studies of the contexts of specific AI applications with normative ethical analysis, it becomes possible to identify and address the real ethical challenges that arise in these contexts.

In short: Without ongoing ethics research, AI ethics is blind and helpless in the face of the ethical challenges brought about by the development of AI.

Read the authors’ arguments on how AI ethics can keep pace with technological applications by combining empirical research with normative analysis: Why applied AI ethics in healthcare requires the integration of empirical inquiry and normative ethical analysis.

Viberg Johansson, J., Nihlén Fahlquist, J. Why applied AI ethics in healthcare requires the integration of empirical inquiry and normative ethical analysis. AI Ethics 6, 533 (2026). https://doi.org/10.1007/s43681-026-01370-2

This post in Swedish