2026 ASEE Annual Conference & Exposition

The emergence of large language models (LLMs) such as ChatGPT has fundamentally changed the landscape of programming education. While these models provide unprecedented accessibility and assistance, they simultaneously threaten the integrity of traditional programming assessments by enabling automated code generation and plagiarism. To address these challenges, CodeShuffler was introduced as an open-source framework for shuffling and implementing paper-based coding exam questions. It shuffles lines of code, creates an image of shuffled instructions, and generates multiple options, including the one with the correct coding sequence. Students are then provided with the shuffled instructions on paper with the requirements to arrange them in the right order on the given paper coding canvas and select the correct answer from the given multiple choices. The approach not only helps educators to fairly assess students’ abilities to write structured code but also upholds academic integrity while reducing grading burden on instructors.

Building on the success of the initial prototype, we present CodeShuffler 2.0, a comprehensive iteration on the system aimed at improving usability, scalability, and utility in broader academic settings. CodeShuffler 2.0 introduces several significant enhancements over the original release. First, a partial credit scoring algorithm has been integrated to reward partially correct answers and better reflect the nuanced nature of programming. This algorithm captures the structurally dependent nature of programs and automatically manufactures potential partial answers based on the provided code snippet.

Secondly, the updated version introduces automated image resizing and cropping to dynamically adjust dimensions according to the complexity and length of coding instructions. This improvement resolves a key limitation of the previous version, which relied on fixed-size outputs unsuitable for large programming problems.

Thirdly, to broaden CodeShuffler’s accessibility and utility, a systematic mechanism has been introduced to generate multiple versions of the same exam from a single set of questions. This feature enables instructors to upload their MCQ-based exams, after which the tool automatically shuffles both the questions and their corresponding answer choices to produce distinct versions. Such functionality empowers instructors to distribute multiple exam versions efficiently, thereby minimizing the likelihood of academic dishonesty during paper-based assessments.

Finally, the tool evolves from a command-line utility into a comprehensive application featuring an intuitive graphical user interface (GUI). This interface streamlines the entire exam-generation workflow by enabling instructors to upload exams, configure shuffle parameters, insert headers and footers, and export fully formatted exam images - all within a single integrated environment. Collectively, these enhancements transform CodeShuffler from a programming-specific prototype into a versatile and instructor-friendly tool adaptable for use across diverse courses and assessment formats.