AI Driven Clinical Communication

This goal lays the algorithmic foundation for the project by developing a hybrid AI dialogue system that integrates LLMs with symbolic AI reasoning to assist healthcare professionals and patients in retrieving and interpreting information from medical documents. The system (i) combines medical expert knowledge with real-world clinical data thus enhancing current hybrid AI methodologies, (ii) delivers accurate facts by reducing risks of hallucinations and gender- and demographic-specific biases and (iii) generates responses that are understandable by a broad range of stakeholders, from medical experts to patients with limited or no medical knowledge.

Our adaptive approach ensures that language, tone, and level of detail are clear, pleasant and relevant for healthcare professionals and patients, fostering trust and preventing misunderstandings. We also prioritize (i) the privacy preservation of sensitive, clinical data, (ii) the development of trustworthy solutions based on open-source LLMs as well as (iii) the research on smaller, domain-specific models in view of energy consumption (ecological sustainability).

This goal focuses on overcoming economic and social challenges by improving efficiency for healthcare professionals and enhancing public health literacy for patients, not only as separate objectives but also in ways that strengthen communication and interaction between them.

For healthcare professionals, the system will significantly reduce the time spent retrieving patient information by providing deeper insights into a patient’s clinical situation (e.g., through individualized patient summaries), contributing to economic
sustainability by streamlining workflows and improving decision-making.

From a patient perspective, the dialogue system will combat misinformation by offering reliable, easy-to-understand health information. This will help patients make informed health decisions, while ensuring that the information provided is accurate, accessible, and inclusive, regardless of their prior medical knowledge or health literacy levels. The design of the dialogue system will prioritize both functionality and usability. This will make information retrieval as effective, intuitive, and understandable as possible, thereby also contributing to technology acceptance.

A key focus of this project is to integrate the dialogue system into existing healthcare ecosystems, ensuring seamless interoperability with both private and public infrastructure, such as ELGA. This integration must account for infrastructure limitations and computational constraints in healthcare environments, requiring the development of smaller, domain-specific models that can operate efficiently within these systems.

Careful consideration will be given to both technical and regulatory requirements, ensuring the dialogue system adheres to strict data privacy laws and interoperability standards within healthcare settings (trustworthy AI). Additionally, this integration will establish a foundation for future AI advancements, particularly in NLP, fostering long-term scalability and driving innovation in the healthcare system (technological sovereignty). This goal aligns closely with Austria’s e-health strategy, supporting the expansion and more efficient use of healthcare infrastructure for future needs.