On Learning How to Program via an Interactive eBook with Adaptive Parsons Problems

Abstract

Novice programmers need well-designed instruction and assessment informed by research and critical perspectives to conquer the historical challenges associated with completing introductory computer programming courses successfully. These issues include high dropout and failure rates, the struggle to acquire and retain basic programming knowledge, and bias and stereotype threat due to social markers (race, gender, dis/ability, sexuality, etc.).

Unfortunately, traditional programming practice, such as code writing, can be arduous, time-intensive, and frustrating. Adaptive Parsons problems, which require learners to place mixed-up code blocks in the correct order and indentation, are designed to support learners’ individual differences in knowledge acquisition, reduce extraneous cognitive load, and improve affect while learning how to program. These problems modify the difficulty of the current or next problem based on a learner's prior performance and help-seeking behavior. Adaptive Parsons problems are a more interactive way to learn while using worked-examples of stereotypical solutions to programming problems; learners acquire strategies for arranging and creating these solutions. Hence, they can help novice programmers build up the kind of mental library of solutions experts have at their disposal when writing code from scratch to solve any number of critical problems related to computing.

This multi-manuscript presents studies aimed at the exploring the problem-solving efficiency of Parsons problems that optimize cognitive load as a substitute for traditional computer programming practice. Mixed methods are used to understand how learners think, behave, and feel when learning how to program via an interactive eBook with adaptive Parsons problems and equivalent write-code problems. First, I conducted field experiments to evaluate the design of these problems for active learning during lecture. Second, I redesigned these problems and tested hypotheses about cognition and learning to understand cognitive, behavioral, and affective learning outcomes impacted by these design changes. And third, I explored access and equity issues for neurodiverse learners.

First, results showed undergraduates are significantly more efficient at solving a Parsons problem versus an equivalent write-code problem, but not when the solution to the Parsons problem was uncommon (not the most common student written solution). And, while most students (80.6%) reported finding Parsons problems useful for learning, some students with prior programming experience expressed strong negative reactions to them. This led to the development of a feature to toggle between a Parsons problem and an equivalent write-code problem. Second, I confirmed the following hypotheses about Parsons problem solutions. Novice programmers were significantly more efficient at solving a Parsons problem created with the most common student written solution versus writing the equivalent code. And, when first presented with a Parsons problem that had an uncommon solution, learners tended to use that solution to solve an equivalent write-code problem. Third, I note four observations about accessibility for learners with seizure disorders, ADHD, mental health disabilities, and memory impairment. Cross-synthesis analysis evidenced that participants benefited from readings in the eBook that chunked information into smaller chapters and sections.

This research has implications for creators of computer programming practice problems who are tackling historic issues related to underrepresented populations in computing by engaging in a critical analysis of how to provide adaptive scaffolding for all learners. Novice programmers with and without disabilities, who require extended time to retain information, benefit from increasing the efficiency and quality of knowledge acquisition, positive attitudes about assessments, and the accessibility of interactive programming practice environments.