This community places the inherent and unavoidable complexity of the real world at the heart of designing, training, testing, and deploying models for data analysis. Reality-centric Data Science aims to operate effectively, reliably, and accountably in the real world.
We propose a reality-centric research agenda consisting of three pillars to unify different areas of current research and build the necessary novel data science (DS) tools and models to deliver the world-changing potential of Artificial Intelligence (AI) while addressing the following questions:
- Inputs – How can we model the world? How can we operate with real-world data? What data should we acquire? Because “data is food for AI,” our actions will support the transition from model-driven AI (collect whatever data you can and develop a model that handles noisy data; keep the data fixed and iteratively improve the code/model) to data-driven AI (where data consistency is paramount; use various tools to enhance data quality to enable multiple models to perform well; keep the code fixed and iteratively improve the data).
- Outputs – How can we adapt to changing circumstances post-deployment? How can we respond to dynamic measures of success and performance? How can we respect human constraints?
- Ecosystem – How and when should we interact with humans or other systems? How can we support interoperability with other components or systems?
The CC has a dedicated website: https://www.cs.ubbcluj.ro/~lauras/research-2/research-projects/rcds/