Relação entre padrões de código e métricas de esforço de estudantes em exercícios de programação introdutória: estudo de caso em um juiz online

Abstract

Programming patterns (or plans) are recurring logical structures used to solve small-scale problems in introductory programming contexts, such as counting loops and nested conditionals. This research investigated the association and effect size between the expected programming patterns and difficulty metrics. The objective was to map which structures demand the most difficulty, measured by 12 variables extracted from an online judges logs, including resolution time, logic errors, typing volume, and code edits. A non-parametric analysis of 446 questions extracted from summative assessments revealed that logic errors are distributed homogeneously among different control flow patterns. However, the "nested conditionals" and "accumulation loops" patterns double the typing volume, rewrite frequency, and resolution time compared to the purely sequential flow. The pedagogical implications of the results suggest using these metrics for balancing questions in assessments and adjusting the didactic sequence.