Enhancing the Accountability and Safety of AI through a Participatory Knowledge-Based Approach

This commentary addresses some of the challenges of artificial intelligence (AI) in healthcare, focusing on data bias, transparency, and the validation of real-world data (RWD). We propose the integration of knowledge-based approaches, particularly ontologies, as a solution for validating the vast amounts of data used in training AI models. Ontologies provide automated verification capabilities to identify errors and biases within datasets. More accurate and trustworthy AI systems can be created by combining machine learning with knowledge-based approaches and incorporating citizen science models in the development of ontologies. This integrated approach ensures that AI technologies will benefit society but also addresses concerns about accountability and public engagement.