Exploring the Difficulty of Faded Parsons Problems for Programming Education
Flynn Fromont, Hiruna Jayamanne, Paul C. Denny · 2023
Opportunities for students at all age levels to learn programming are growing and this has led to increased enrolments both online and in the classroom. As the range of prior experiences within a cohort becomes more diverse, it poses a challenge for instructors to produce learning resources at an appropriate level of difficulty for individuals. Techniques to easily modify the difficulty of existing resources can be valuable in addressing this challenge, but not all problem types lend themselves to such customization. Parsons problems are a popular programming exercise that present learners with the lines of code for a solution in a scrambled order. Faded Parsons problems are a recent variation, where some lines of code have missing portions that students are required to fill in. Selecting which parts of the solution code to remove, or fade, is typically a manual process for an instructor and little is known about how different fading strategies impact problem difficulty. In this paper, we present a novel open-source tool for delivering Faded Parsons problems, and we explore the relative difficulty of three distinct fading strategies as part of an evaluation in a first-year programming course with 900 students. We find that fading conditional statements greatly increases the difficulty of Faded Parsons problems, and we observe that some students appreciate the ability to tackle more challenging problems. Results of this work will inform future automated approaches for adapting the difficulty of such problems.