The project of Prof. Maciej Zięba from the Department of Artificial Intelligence at our Faculty received funding under the 28th edition of the Opus programme. In total, the National Science Centre awarded 18 million PLN to nine scientists from Wrocław University of Science and Technology. Our university is also a partner in four more projects.

The project by Professor Maciej Zięba, titled "Probabilistic methods for generating reliable, stable, and varied counterfactual explanations," has received funding of 962,800 PLN from the National Science Centre.

Artificial intelligence (AI) increasingly impacts our lives, from product recommendations to supporting medical diagnostics. However, many AI systems operate as "black boxes", making their decisions difficult to understand.

– When the system denies a loan, the applicant often doesn't know why they were refused or what they could change to improve their chances. Such a lack of transparency undermines trust in AI – explains Prof. Maciej Zięba.

Counterfactual explanations offer "what if" scenarios, showing what needs to be changed in the data to achieve the desired outcome. For example, the system might indicate that increasing the income or reducing the debt would help obtain a loan. Such explanations must meet three conditions: be realistic, stable with minor data changes, and offer various alternative solutions.

The research project led by Prof. Maciej Zięba aims to enhance the quality of counterfactual explanations by employing probabilistic methods to consider three crucial aspects: credibility, stability, and diversity.

– Probabilistic models, by analysing the uncertainty and variability of data, enable the creation of realistic and practical suggestions. For example, instead of suggesting unattainable income growth, they indicate realistic modifications based on actual data – says the researcher.

The task of our scientists will be to develop advanced techniques that generate counterfactual explanations with consideration to these three key aspects. The developed methods will find applications in many fields, from healthcare to finance, providing practical tools for building more transparent AI systems.

– Our goal is to increase trust in AI by making its decisions more understandable and practical for users – explains the scientist.

A million-zloty support

Opus is a broadly structured competition in which researchers at any stage of their academic career, regardless of age and level of experience, can apply for funding to basic research projects.

Thanks to the grant, they can build large research teams, carry out projects using major international research facilities, and also collaborate with foreign partners.

The ranking list for the 28th edition of the Opus programme includes 234 projects, with total funding surpassing PLN 448 million. Among this group, there are nine projects from Wrocław University of Science and Technology, with a total 18 million PLN awarded by the National Science Centre. Besides Prof. Maciej Zięba from our Faculty, the funds also went to:

• Prof. Sławomir Pietrowicz (Faculty of Mechanical and Power Engineering)
• Prof. Jarosław Myśliwiec (Faculty of Chemistry)
• Prof. Katarzyna Matczyszyn (Faculty of Chemistry)
• Dr. Renata Grzywa (Faculty of Chemistry)
• Prof. Rafał Kowalczyk (Faculty of Chemistry)
• Prof. Katarzyna Smolińska-Kempisty (Faculty of Chemistry)
• Dr. Eng. Adrian Chajec (Faculty of Civil Engineering)
• Prof. Robert Zaleśny (Faculty of Chemistry)

Additionally, our scientists are part of four more projects as their consortium members.