AIXPERT

An agentic, multi-layer, GenAI-powered backbone to make an AI system explainable, accountable, and transparent.

AIXPERT envisions a future where AI systems are explainable, transparent, accountable, and robust, fostering trust in human-machine interactions. The AIXPERT solutions will measure the efficiency and accuracy of AI systems and will also provide contextual explanations, offering insights into the data and models used to generate outcomes. The integration of Knowledge Graphs (KGs) with AI agents will link the data and the actions taken by AI agents. Reinforcement Learning from Human Feedback (RLHF) will enable AI to learn and adapt based on real-time user interactions. The dynamic creation and configuration of AI agents will allow users to customise and deploy agents, primarily powered by Large Language Models (LLMs). AIXPERT will leverage human-centred design principles, incorporating participatory design and usability studies.

Consortium and funding:

The project is led by ATHENA Research and Innovation Centre, Greece.
There are 17 partners from 10 countries (9 European countries plus Canada), in a multi-disciplinary consortium which brings together partners with expertise from AI technology, social sciences, humanities, and industry, as needed for successfully designing, developing, evaluating and pushing forward the envisioned solutions. The overall budget is about 7.9 million euros. In this project the RuG team will lead the development of an AI auditing and evaluation framework, the development of social equity-centred and ethically aligned models, and will contribute to the Integration of Knowledge Graphs and Large Language Models (LLMs) for enhanced decision-making, including their application in Manufacturing use cases, and the overall cross-domain (the project includes application cases in the domains of healthcare, manufacturing, educational robotics, creative arts and music, as well as human resources operations and services) validation and societal impact assessment of AI systems. The project duration is three years.

Envisaged impact:

The AIXPERT project aims to develop a comprehensive, human-centric AI framework that enhances explainability, transparency, accountability, and autonomy in AI systems, particularly in the context of generative AI models and their integration in agentic workflows. This solution provides essential tools for multi-agent workflows, multimodal data integration, and adaptive learning, addressing complex challenges related to AI trust, user engagement, and ethical compliance. Thus, it will contribute towards a human-centred and ethical development of digital and industrial technologies, leading towards a more resilient, inclusive, and democratic society.

Check the project factsheet.