Generating personalized student synthetic datasets for large-scale educational assessment

Large scale, remote education comes with many challenges, some of them quite different from in-person education. One of them is the ability to create proper assessment activities that are also safe against information leakage that may lead to fraudulent behaviors (i.e., copying the correct responses from classmates) or failure to achieve the proposed educational objectives (i.e., providing responses without participating in the activity). In this talk, we introduce a pilot activity for a “Data Analysis 101” course in the Bachelor’s Degree in Psychology at the National Distance Education University (Spain) with around 7,600 students.
A simple Shiny app generates a personalized dataset for each student, which they can download and use to complete an exercise in Jamovi (an open-source statistical software). Students browse to the app using a GET request link created in the OpenLMS course, using their university email address as a parameter. The app uses this address to initialize the random seed before generating the student’s dataset. Their responses are automatically scored offline, and another Shiny app provides them with feedback with the correct responses. Our app applies Shiny on the field, showcasing how a very simple application can satisfy a real-life, large-scale educational necessity without much sophistication . At submission time, the activity assessment and results are yet to be analyzed.