for HLTH-2025-01-CARE-01 call, partnership offer

Our AI research group is looking for a consortium to participate in the project focusing on advancing trustworthy and ethical AI solutions. Their contribution will center on Ontology-Based Bias Mitigation in Generative AI Solutions for Multimodal Datasets which is an essential component in WP3. Given the rapid developments in AI technologies, ensuring fairness, transparency, and accountability has become a crucial challenge, particularly in the context of multimodal data (as in the project scope: text, images, audio, and video related to health).

To achieve a high level of user satisfaction in the health assistance process, semantic based features have to be taken into account to detect and prevent bias. An ontology (structured collection of domain specific elements, their attributes and the relationships among them) can be constructed to support the process through multiple channels. Vector space based (using similarity measures) and probabilistic (based on the pdf) methods will be utilized (and compared) for possible bias classification where a domain ontology is necessary to compute the similarity and the probability.

With expertise in ontology-driven methodologies, I aim to develop a structured knowledge representation that can enhance bias detection, bias mitigation, and explainability in generative AI models. By utilizing formal ontologies, I will work towards designing systems that systematically identify and rectify bias at multiple levels: data preprocessing, model training, and output generation. In this way, we will not only aim to improve AI fairness but also the resulting virtual assistant will align with key EU regulatory frameworks such as the AI Act and Ethical AI Guidelines.

Considering the well-structured concept note, we also wish to emphasize our expertise in “design thinking” based co-design which can be utilized in multiple work packages (WP1-2-4). In 2019-2020, our university adopted a “project based learning” approach where our department (Electrical-Electronics Engineering) was the pilot site. Therefore, we have developed extensive experience in design methodologies.

The AI research group at Kadir Has University focuses on the application of computational intelligence algorithms for the development of smart solutions. KHAS-AI is a multidisciplinary research group with members from engineering (Electrical, Computer, Industrial), bioinformatics and data science fields. The group develops state of the art solutions using machine learning (statistical methods, neural networks, deep learning, CNN and LSTM networks, GenAI), Big Data processing, embedded IoT systems, network science, and optimization methods for various problem domains such as indoor/outdoor localization, spatio-temporal environmental anomaly detection, segmentation, risk prediction, epidemic modeling, emergency response and disaster relief management. We believe that collaborations with our group can contribute substantially to the development of novel data driven methods for effective smart systems with special emphasis on the problems of the society. The projects carried out by KHAS-AI members are funded by the Scientific and Technological Research Council of Turkey (TUBITAK) and by various industrial partners.