Nowadays, information comes from an ever-growing multitude of digital sources, challenging citizens and institutions to weigh each type of information on relevance and reliability. In The Netherlands, libraries, archives and museums have faced these challenges for centuries. They do even more so since these institutions started mass digitizing their multimodal collections, building datasets with terabytes full of information of books, video, music, maps, photographs and documents. These datasets are indispensible sources for reliable fact checking and data validation by citizens of the digital society, journalists and other societal stakeholders. In this way, cultural heritage collections contribute critically to societal resilience.

Modern AI techniques such as 'deep learning' provide the promising prospect to interlink and enrich collections, deepening our understanding of the Netherlands' polyvocal past, present, and future. However, developing these Artificial Intelligence techniques is challenging, given the size and heterogeneous nature of the big data heritage collections, the context-dependency of interpretations and the inherent dynamics of ever-growing collections. Therefore, in the HAICu project, AI and Digital Humanities researchers, heritage professionals and engaged citizens aim for scientific breakthroughs in AI to open up, link and analyze in context large scale and heterogeneous multimodal digital heritage collections to facilitate user-assisted generation of fact-based narratives. Our target groups are all types of users and institutions whose functioning relies on relevant and reliable information. In co-creation with these groups, the HAICu project will develop new methods and user-engineered tools to construct meaningful contexts out of the available big data, assisting users to weigh both historical as well as current multimodal information.