Keywords: AI planning, plan recognition, hypothesis generation, symbolic computation, knowledge engineering
We outline some of our successes and stumbling blocks in attempting to bring together data analysis and expert knowledge in almost a decade of research at IBM. Most of our work is focused around using planning technology to automate the things data scientists worry about: looking at data in many possible ways, recognizing patterns and processes, and providing foresight. The good news: we can scale up the ability to do data analysis - we will show examples of our latest system for real-world enterprise risk management; the bad news: there is no silver bullet, and we may need problem instance-specific methods and assumptions.