Skip to main content

Project

WP1: Project Management​

WP1 mainly provides the coordination of scientific and administrative activities. Project management, budget allocation, and reporting activities will all be coordinated as part of this work package. This work package will be led by UEvora, who has extensive experience in project management.

WP2: Requirement Gathering and Use Case Co-Design​

WP2 is focused on requirement gathering and use-case co-design. Consequently, the key task is to gather requirements and expectations from healthcare stakeholders for the design and evaluation of trustworthy AI solutions and establish high-level requirements for technical research activities of the project. We will then define relevant and convincing use cases in view of promoting the capabilities of developed solutions.

WP3: Explainable AI Models and Evaluation Metrics​

WP3 is geared towards developing explainable AI models and evaluation metrics. Harmonic AI will develop explainable AI models with respect to use cases. We will establish formal quantitative evaluation metrics for the quality of AI explainability in the healthcare domain. Lastly, the WP will propose a design framework for explainable AI in digital health.

WP4: Bias Mitigation Techniques and AI Fairness Metrics​

WP4 aims to develop bias mitigation techniques, AI fairness metrics, and operational guidelines. The WP will design bias mitigation techniques to improve AI fairness with respect to use cases. WP will also explore fair AI operational guidelines for healthcare organisations. Lastly, we will establish unified quantitative AI fairness evaluation metrics.

WP5: Hybrid Learning Paradigm for Secure and Privacy-Preserving AI​

WP5 will propose a three-tier privacy-preserving hybrid learning paradigm for prioritised predictions. It will integrate state-of-the-art privacy-enhancing technologies into the proposed hybrid learning paradigm. The WP will develop security mechanisms to defend against attacks on the cyber-physical system, AI models, and datasets. Lastly, the WP will establish a quantitative metric to measure global privacy loss.

WP6: Human-Machine Collaborative Multi-Objective Model Selection​

WP6 will integrate the proposed explainability, fairness, and privacy innovations into a unified AI pipeline. We will develop a human-guided multi-objective model selection algorithm. Further, we will develop a human-machine co-design framework for domain-specific, requirement-oriented trustworthy AI in digital health.

WP7: Implementation, Testing, and Stakeholder-Centered Evaluation in Use Cases​

WP7 will implement the proposed explainable, fair, and private AI solutions in use cases. The WP will evaluate the proposed human-machine collaborative multi-objective design framework for trustworthy AI. Lastly, WP will evaluate the proposed guidelines for AI practitioners and healthcare stakeholders.