European AI ethics in practice: the AIOLIA project at THWS

The EU’s AIOLIA project aims to shape artificial intelligence responsibly. At CAIRO at THWS, a research team led by Professor Dr. Ivan Yamshchikov is examining real-world AI systems to derive practical ethical guidelines – ranging from personal assistants to salary advice provided by language models.

Published on 25 June 2026

Artificial intelligence (AI) has long since become part of everyday life, from chatbots and personalised recommendations to digital assistants. The EU-funded project ‘AI Ethics Operationalised – Ideation, Learning, Implementation, Assessment’ – AIOLIA for short – now aims to clarify how these systems can be designed responsibly. AIOLIA seeks to develop ethical guidelines that not only sound good, but are also practically applicable for companies and developers. The Center for Artificial Intelligence (CAIRO) at the Technical University of Applied Sciences Würzburg-Schweinfurt (THWS) plays a central role in this. The team led by Professor Dr. Ivan Yamshchikov is primarily working on one key problem within AIOLIA: how can abstract values such as fairness, transparency and autonomy be translated into application-specific guidelines?

From the gender pay gap to AI-powered salary advice

This tension is particularly evident in a research finding that has attracted international attention. The CAIRO team tested large commercial language models – including those used as personal assistants – using a workplace scenario: a fictional CV, a specific job vacancy, and the question of an appropriate salary expectation. The result: on average, several models recommended lower salaries when they assumed the person making the enquiry was a woman. When the same enquiry was made by a man, the recommended salary was higher.

What is particularly concerning is that users do not even have to actively specify their gender. The systems infer a person’s gender from language, context, or previous interactions, and can thus inadvertently reproduce existing social inequalities. Yamshchikov points out that such effects demonstrate just how deeply AI systems are embedded in social dynamics – and that powerful technology can, at the same time, have undesirable side effects. Newer model versions, such as GPT-5 (from OpenAI), no longer exhibit these problems, as the providers have made improvements.

Areas of tension in everyday life

Dr. Angelica Henestrosa, a postdoctoral researcher at CAIRO, originally comes from the field of psychological research. She has investigated how people perceive AI-generated content and how credible they consider it to be. Even at that stage, it had become clear that ethical issues are omnipresent in this field of research and would continue to grow in importance, particularly with the further spread of large language models.

As part of the AIOLIA project, Dr. Henestrosa is working on the ‘AI as a personal assistant’ use case. This involves systems designed to support families in their everyday lives – for example, with organising appointments or setting personalised reminders. What is technically feasible also raises sensitive questions: Who is allowed to access which data? How much autonomy should children have in relation to such a system? And where does the line between support and control lie?

An industry partner had originally planned to develop an assistant for entire families, designed to communicate with children and young people as well. However, a fundamental conflict of objectives between protecting the privacy of minors and meeting parents’ safety needs soon became apparent. No clear-cut solution could be found. The company ultimately decided to change direction and is now focusing on assistance systems exclusively for adults.

From an abstract principle to a technical decision

A key problem with many ethical guidelines is their abstract nature, explains Professor Dr. Yamshchikov. Whilst terms such as transparency and non-discrimination are found in regulations such as the EU AI Act (see info box), their practical implementation often remains unclear to developers.

AIOLIA therefore adopts a bottom-up approach: rather than discussing ethics in a vacuum, it analyses specific use cases from industry. Through interviews and workshops with companies and start-ups, real-life decision-making situations are examined: What data is collected? How are risks assessed? Where do unclear guidelines hinder innovation? On this basis, case studies are developed which are also intended to help policymakers shape regulation in a more practical way. From the Würzburg team’s perspective, this is a key value of the project: rules should not be divorced from reality, but should be based on actual application scenarios.

Interdisciplinarity as a prerequisite

In the project, researchers from computer science, psychology, philosophy, and law, along with a number of industry partners, are working together on a question that no single discipline can answer on its own. In this context, Dr. Henestrosa speaks of ‘interdisciplinarity in practice’: the focus is not on specialist fields, but on specific problems.

At the same time, AIOLIA highlights just how varied cultural and regulatory frameworks are worldwide. In some countries, the focus is on AI applications for older people; in others, on medical systems or autonomous mobility. The aim is to formulate a common set of core values without ignoring national or cultural differences.

Education and awareness-raising

However, AIOLIA does not stop at reports for Brussels. Based on the case studies, the consortium is developing teaching materials intended for use in courses across Europe, for example for students, professionals, or people undertaking further training. These materials are also set to be incorporated into courses at THWS.

Henestrosa believes it is crucial that people learn to understand AI systems as early as possible. Not necessarily as a separate school subject, but embedded within existing subjects – precisely where applications actually feature in the learners’ everyday lives. As with social media, the aim is to understand both the risks and the opportunities, and to deal with them in a considered manner.

Professor Dr. Yamshchikov also emphasises the role of science communication: through podcasts, social media formats, and publicly accessible examples, the CAIRO team aims to help ensure that public debates on AI are shaped not by fear, but by informed decisions. In a democratic society in particular, it is important to make technological developments understandable, so that regulations are based not on vague fears but on concrete knowledge.

A project at the cutting edge

For those involved, AIOLIA is more than just another EU project. Dr. Angelica Henestrosa sees it as an opportunity to systematically integrate ethical values into technical systems and regulatory processes. Professor Dr. Ivan Yamshchikov emphasises that universities of applied sciences, in particular, can play a key role by linking research, practice and policy. The project highlights that ethical issues in the development of artificial intelligence are not an afterthought, but must be considered from the outset.