Date of Award

2020

Document Type

Open Access Dissertation

Degree Name

Doctor of Philosophy in Computational Science and Engineering (PhD)

Administrative Home Department

Department of Computer Science

Advisor 1

Charles Wallace

Committee Member 1

Linda Ott

Committee Member 2

Robert Pastel

Committee Member 3

Adam Feltz

Abstract

Students in introductory computer science courses, are learning to program. Indeed, most students perceive that learning to code is the central topic explored in the courses. Students spend an enormous amount of time struggling to learn the syntax and understand semantics of a particular language. Instructors spend a similar amount of time reading student code and explaining the meaning of the cryptic error messages displayed by compilers. Messages provided by compilers are intended to give feedback on the adherence of one’s code to the language specification and conventions. Unfortunately, these message are geared towards experts who have a clear understanding of the language syntax and semantics and a deep model of what comprises a program and how a program is developed. These students are novices who lack fundamental understanding of the structure of a program and have no basic mental model of how a program works. Novices make different kinds of mistakes than experts. Instructors need to spend a lot of time simply assisting novices in using compilers and understanding their output. In addition to mastering the syntax and semantics of their first programming language, novices are exposed to the question of what constitutes good design. Instructors can identify virtuous design choices and articulate areas of improvement. But contact time with students is limited, and waiting for in-person feedback or replies to personal messages can be a critical delay. Novices, still struggling to use the compiler, have not yet developed the sophisticated analytical processes employed by experts and this is reflected in their design choices and the kinds of mistakes they make. When a novice approaches an instructor with a question, the instructor must often provide a balanced critique that assists the student with understanding both the structure and the design aspects of their own code. My research has focused on whether we can identify examples of early programming antipatterns that have arisen from our teaching experience, and describe different ways of detecting them automatically. Novice students may produce code that is close to a correct solution but contains syntactic errors; code critiquers attempt to salvage the promising portions of the students submission and suggest repairs in ways more meaningful than typical compiler error messages. Alternatively, a student misunderstanding may result in well-formed code that passes unit tests yet contains clear design flaws; through additional analysis, code critiquers can detect and flag these flaws. Finally, certain types of antipatterns can be anticipated and flagged by the instructor, based on the context of the course and the programming activity; code critiquers allow for customizable critique triggers and messages. This dissertation presents several key contributions to our understanding of novice misconceptions and their representation, diagnosis and repair using antipatterns. My research focuses on identifying antipatterns and detecting them in novice code, then using this information to provide the student with a meaningful critique of their work. I have developed WebTA, a tool to critique student programs in introductory computer science courses. WebTA is used to teach students test-driven agile development methods through small cycles of teaching, coding integrated with testing, and immediate feedback.Through the use of WebTA in introductory computer science courses since 2014, I have amassed a significant corpus of novice programmer submission data. Lastly, I have compiled a library of antipatterns found in novice code.

Download

Share

COinS