Grant: Canada Excellence Research Chairs

Data science is being done over massive data lakes containing datasets from different sources and datasets that have been updated and cleaned repeatedly (and often incompletely). This creates a mess of slightly different versions and datasets that may have incomplete or ambiguous semantics. An important challenge in doing trustworthy data science is to enable scientists to find relevant datasets, recover and understand the meaning of different datasets and their versions, and integrate data in a way that fully (and faithfully) recovers all possible connections between facts. Data Intelligence is the theory and practice of how to automate and scale the intelligent, reliable, and trustworthy use of big data for data science and AI. To achieve data intelligence, the CERC will 1) develop new mathematical foundations and methods for understanding when data preparation or curation solutions are correct; 2) develop preparation and curation methods that are correct, explainable, and reproducible; 3) develop frameworks for data curation that help data scientists document, share, and reproduce complex data preparation and curation processes. Data intelligence is concerned with the data preparation and data engineering portion of data science. Nonetheless, machine learning, language models, and knowledge bases (ontologies) are important tools in recovering the semantics (meaning) of data. An important outcome of this CERC will be the theory and practice of how to employ these and other foundational tools (including network science and data management) to create principled data curation solutions. This research is critical to ensuring the insights we gain from data are valid and unbiased. Data intelligence will help to ensure that we can use data wisely to understand possible inequity in the data rather than blindly building models that learn to repeat past inequities. An important component of this CERC is to use the research agenda to motivate and expand the set of data science trainees from marginalized groups by exposing them to the power of telling stories through the careful and wise use of data.