Another look at the behaviors of novice programmers

Abstract

This paper reports on the progress of an NSF funded research project investigating the development practices of students in introductory programming courses. In previous work, we describe our extension of the BlueJ IDE to capture events associated with program development. Here we report on data collected during the Fall 2007 and Spring 2008 semesters on CS 1 students. In particular, we show that our data analysis independently confirms the results obtained in separate studies by Jadud [3, 2]. In addition we use our empirical evidence to discern some higher level "patterns" of beginning student programming behaviors including potential cheating and the impact on success of students starting projects late.

Flexible, reusable tools for studying novice programmers

“…Recently, a clean-room implementation of the data collection framework described here was published at SIGCSE 2009 [2]. The results of Fenwick et al's replication support the findings of Jadud and Tabanao et al Along with our successful three-year use of these tools, we take it as external validation that the kind of data we are collecting has utility in the study of novice programmers.…”

Towards improving programming habits to create better computer science course outcomes

“…In addition, it has also been shown that students who submit late on the first assignment are more likely to continue submitting late and do more poorly in the class [8]. Fenwick et al have also found trends suggesting that students who program in more 'sessions' (doing more incremental work) are more likely to achieve higher grades [9]. We aim to replicate and expand on this work, while also more closely examining the effects of hidden tests and release tokens.…”

“…For example, several researchers have examined compilation behavior, in order to determine how students react to errors and which errors are most common [10,9]. Correlations have also been found between students' relationships with errors and their midterm scores which can be used to identify at-risk students [17].…”

“…There have been excellent recent efforts at collecting data to help us better understand novice compilation behavior [10,17,18], and novice programming habits [6,13,9]. This empirical work has allowed us to begin addressing key questions such as "How much time do students spend on programming assignments?…”

Web-scale data gathering with BlueJ

“…Jadud [5] and Fenwick et al [4] have recently used bespoke extensions to the BlueJ Java IDE to capture student interactions at an intermediate granularity, capturing the input to and output from the Java compiler at every invocation. Jadud captured 42,000 events from 186 students over two years, and Fenwick et al captured 55,000 from 110 students in a single year.…”