Roger Azevedo (UCF PI), Lujie K Chen (University of Maryland PI)
Total: $134,000
Funder: National Science Foundation
In an era when AI tools such as chatGPT can generate code according to specific instructions, it is important to train future data scientists to hone their higher-order reasoning and problem-solving skills that are not easily replaced by AI. Caselet (bite-size case studies) is a self-paced online practice tool to support the accelerated development of Data Science Problem Solving (DSPS) and self-regulated learning skills at a large scale. DSPS is a cluster of complex skills built upon but beyond the mastery of conceptual knowledge and procedural skills. Though the underlying component skills are frequently covered, the higher-order DSPS are not commonly taught in a typical graduate-level data science curriculum. However, a competent data science problem solver in the real world must develop proficiency in properly formulating real-world problems and efficiently exploring large and complex solution spaces. The road toward mastery is typically slow and requires trainees to be exposed to many hands-on, real-world experiences under the guidance of experts, which are often unstable or even inequitable. With Caselet, learners are presented with a scenario that mimics a real-world problem and will be asked to answer 5-7 well-crafted multiple-choice questions specifically targeting the main areas of DSPS. Students will be given tailored feedback and detailed explanations based on their answers. We have compiled about ten caselets and 70 items, which were piloted among 100 + graduate students at Carnegie Mellon University and the University of Maryland Baltimore County as part of the graduate level data science course assignment. This seed work won the DARPA AI Tool Competition for Adult Learning in the 2022-2023 cycle.