Collaborative Research: Scalable scaffolding of novice programmers' learning and automated analysis of their online activities

点击翻译按钮获取中文摘要

The need for a programming-savvy workforce, the challenges encountered nation-wide in teaching programming skills, and the advent of massively open online courses (MOOCs) all stress the importance of revolutionizing existing technologies meant to support learning via unsupervised practice. The reliance on hand-designed learning experiences in Intelligent Tutoring Systems has created a bottleneck in the enhancement of such technologies. The significance and importance of this project will be the creation of a system that automatically provides programming practice problems via a tutoring system. Parsons puzzles, which are the focus of this work, have already demonstrated improvement in programming skills. This project will continue research on the effectiveness of Parsons puzzles and determine whether an evolutionary algorithm approach to creating Parsons puzzles is effective in producing the same or better learning outcomes.To this end, this project will achieve two complementary goals. First, Interactive Evolutionary Algorithms will be used to autonomously design Parsons puzzles. These relatively new practice problems have already shown great promise in supporting the development of programming skills. By integrating Vygotsky's Zone of Proximal Development theory in the design of our fitness function, the pedagogical soundness of the evolved Parsons puzzles will be further enhanced. Second, recent breakthroughs in Co-evolutionary Algorithms will be leveraged to data-mine the underlying learners-problems interaction space. This will permit autonomous extraction of insights about the most informational interactions. Such information has the potential to provide new perspectives on the significance of both evolved and hand-designed Parsons puzzles, along with the order in which to guide students through these puzzles to ensure proper scaffolding. Quantitative and objective measures of differences in mean difficulty and number of attempts to solution will be used in comparing hand-designed Parsons puzzles and evolutionary algorithm created Pasons puzzles.