Bio

Over the past eighteen months, I have had the opportunity, as a researcher, to develop a step-by-step framework for working with AI and ethics. In this blog post, I would like to share my findings and look ahead to what is still needed.

The research started from the observation that the Ethics Guidelines for Trustworthy AI, as proposed by the European Union, are not always easy to apply to concrete cases. When I introduce AI developers to the guidelines and the accompanying assessment list, I receive comments such as:

“When should you ask which question during the development process?”
“This is a great deal of work.”
“Could you not simply create a flowchart for applying ethics to AI?”

These comments reveal something about both the developers and the guidelines themselves. Let me begin with the developers.

Developers are trained to make technical choices. Ethical choices are, in a certain sense, of a different order. Ethics often involves a grey area in which there is no single objectively correct decision. Ethical reasoning can therefore feel like a foreign way of thinking that has to be introduced into the development process, and not everyone is convinced of its purpose or value. Moreover, ethics does not receive the same degree of attention in every educational programme.

The guidelines themselves are well structured. They are grounded in fundamental human rights and formulate seven key requirements that translate these principles into a more concrete working instrument. Although this top-down exercise has been carefully developed, ethics always requires contextual grounding.

The application of principles differs from one context to another. A healthcare setting, for example, will understand human agency very differently from a highly automated industrial environment. An important interpretative exercise therefore remains necessary, and this requires specific skills.

There are also additional ethical dimensions that can be included and that may further strengthen the innovative potential of ethics. The approach known as ethics of accompaniment has developed an important model in this regard. It could be described as a bottom-up approach.

All relevant actors are involved in the ethical process. The process begins by identifying what is ethically at stake in a particular case. The most important values are then made explicit and translated into technical criteria through values in design. This approach, too, requires the necessary skills.

The central question therefore remains: how can we encourage the meaningful integration of ethics?

The guidelines are not mandatory. Applying them requires considerable effort, specific capabilities are needed and not everyone is convinced that ethics can contribute to innovation. In my view, their effective application depends on two main factors.

Organisational Culture

The first factor is organisational culture. How does an organisation approach ethics?

This often begins with its vision and mission. A well-developed vision translates values into everyday practice. During the development of an AI application, both the organisational culture of the developer and that of the client play an important role.

An AI application will not be successfully adopted when the values it embodies conflict with those of the organisation in which it is implemented.

Imagine, for example, that an organisation wants to increase sustainability by introducing a particular tool. If this objective results in even a minimal loss of performance, it may create conflict, especially when high performance is deeply embedded in the organisation’s DNA.

Another aspect of organisational culture is the importance assigned to ethics itself. How seriously does the organisation take ethical relationships and decision-making? Is ethics genuinely part of the organisational culture, and is the organisation willing to invest in it? Or is ethics considered subordinate to other priorities and used primarily to comply with legislation or to improve the organisation’s public image? In the latter case, we may speak of ethics washing.

The Individuals Involved

Alongside organisational culture and its norms and values, the individuals involved in the development process also have an important role to play.

The developer immediately comes into view. How does the developer understand ethics? How comfortable are they with the fact that ethical questions do not always have black-and-white answers? Do they possess the skills required to identify ethical issues and translate general principles into a specific context?

What is often underestimated, however, is that other people also influence the decision-making process. The extent to which ethics can be implemented frequently depends on decisions made above the developer’s level.

In addition to guidelines, the skills discussed above are therefore essential: both the willingness and the ability to act ethically. This is why it is important to train as many actors as possible to engage in ethical reflection. Virtue ethics, for example, can play a valuable role in this process.

The way ethics is approached is therefore multidimensional. It does not depend solely on a good method — such as a step-by-step framework — for introducing ethical questions. It also depends on organisational culture and on the individual commitment of managers, developers and other stakeholders.

The process of embedding AI therefore already begins when an organisation formulates a vision and mission that genuinely come to life in practice and in which its most important values are translated into concrete action.