Journal of Science Education and Technology, Vol. 16, No. 4, August 2007 ( 2007) DOI: 10.1007/s10956-007-9055-5
Transforming an Introductory Programming Course: From Lectures to Active Learning via Wireless Laptops Miri Barak,1,3 Judson Harward,2 George Kocur,2 and Steven Lerman2
Within the framework of MITÕs course 1.00: Introduction to Computers and Engineering Problem Solving, this paper describes an innovative project entitled: Studio 1.00 that integrates lectures with in-class demonstrations, active learning sessions, and on-task feedback, through the use of wireless laptop computers. This paper also describes a related evaluation study that investigated the effectiveness of different instructional strategies, comparing traditional teaching with two models of the studio format. StudentsÕ learning outcomes, specifically, their final grades and conceptual understanding of computational methods and programming, were examined. Findings indicated that Studio-1.00, in both its extensive- and partial-active learning modes, enhanced studentsÕ learning outcomes in Java programming. Comparing to the traditional courses, more students in the studio courses received ‘‘A’’ as their final grade and less failed. Moreover, students who regularly attended the active learning sessions were able to conceptualize programming principles better than their peers. We have also found two weaknesses in the teaching format of Studio-1.00 that can guide future versions of the course. KEY WORDS: studio-based learning; conceptual understanding; wireless laptops; object-oriented programming; undergraduate education
distribution allows faculty to refer to and comment on professional implementations of classroom topics (Tyma, 1998). Despite its many advantages, Java was not specifically created for beginning programmers, and hardly any thought was given during its design to making basic programs simpler to write (Horstmann, 2002). Several studies have found that learning a new programming paradigm or language is difficult and that studentsÕ understanding is hampered by preconceptions that later may develop into misconceptions (Benander et al., 2004; Milne and Rowe, 2002). Hence, it is imperative to provide students with a learning environment that supports knowledge acquisition and the conceptual understanding of object-oriented programming. Such a learning environment can be provided by the studio format that offers a multimodal environment in which lectures, recitations and laboratories are combined and mutually reinforce one another. The studio learning and teaching format provides a framework
INTRODUCTION Undergraduate students in engineering, science, and mathematics programs are usually expected to acquire programming skills as a part of their education. To date, object-oriented programming is the dominant programming paradigm, and Java is the leading computer language taught in higher education institutions (Benander et al., 2004). JavaÕs extensive class libraries provide students with useful building blocks to encourage the development of complex programs, and implement many of the algorithms and data structures addressed in lectures. The source code provided as part of the standard
1
Technion, Israel Institute of Technology, Haifa, 32000, Israel Massachusetts Institute of Technology, Cambridge, MA, 02139, USA 3 To whom correspondence should be addressed; E-mail:
[email protected] 2
325 1059-0145/07/0800-0325/0 2007 Springer ScienceþBusiness Media, LLC
326
Barak et al.
for real-life learning, problem-based solving, and hands-on practice. In the present study, the studio format was introduced into a leading programming course at the Massachusetts Institute of Technology (MIT): Introduction to Computers and Engineering Problem Solving, cataloged as course 1.00, and therefore entitled: Studio-1.00. This study is a part of a series of ongoing projects at the Center for Educational Computing Initiatives (http://www.ceci.mit.edu) at the Massachusetts Institute of Technology. These studies examine the integration of new pedagogy and technology into higher education courses and their effect on studentsÕ learning (Barak et al., 2006; Dori and Belcher, 2005a; Dori et al., 2003). The current paper reports on a study that investigated the effect of the studio format on undergraduate studentsÕ academic achievements and conceptual changes. It grows out of the constructivist philosophy of learning and teaching that views knowledge as a constructed entity (von Glasersfeld, 1995).
1997). Indeed, active learning is consistent with the constructivist theory that maintains that knowledge cannot simply be transmitted from teachers to learners; rather, learners must be engaged in constructing their own knowledge (Bruner, 1990; von Glasersfeld, 1995). Active learning environments encourage students to be engaged in solving problems, sharing ideas, giving feedback, and teaching each other (Johnson et al., 1998; Marbach-Ad and Sokolove, 2002). Concurrently, Studio-1.00 aimed at enhancing studentsÕ involvement in learning by encouraging them to explore different solutions to a certain problem, write their own segments of code, and, test their solutions by compiling and running their program. From this perspective, educational studies have suggested that integrating active learning strategies, as part of the formal learning sessions, can advance studentsÕ learning and their conceptual understanding (Carey, 1985; West and Pines, 1985). This supposition was investigated in the current study.
THEORETICAL BACKGROUND
The Studio Format for Teaching and Learning
Conceptual Understanding
The core of the studio format is in the shift from passive, lecture-based, teacher-centered instruction to active and interactive student-centered learning. Its origins are found in schools of architecture, where it became a model of human-problem engagement for teaching students how to actively engage in resolving problems and designing programs. It pre-dated the use of computer technology. However, as the concept of ‘‘Studio classes’’ has evolved, it is now conceptualized as a mixture of student exercises, instructor coaching, and laboratory work, which generally takes advantage of modern technology to deliver instructional materials (Cummings et al., 1999). The studio format has many different interpretations (Dori and Belcher, 2005a; Turbak and Berg, 2002), but its essence lies in the increased interactions between the students and themselves and their instructors, as they work through course projects and/or classroom exercises. Recent research on studio-based learning indicated four major educational gains. First, it increased studentsÕ performance and interest in the subject matter (Barak et al., 2007; Foulds et al., 2003). Second, it enabled students to freely interact with each other, with the instructors, and the computer software (Barak et al., 2006; Glinkowski et al., 1997). Third, it increased studentsÕ satisfaction (Wilson, 1994). Fourth, when visualization was integrated, the
Conceptual understanding is the ability to deduce, perceive, and comprehend a concept. Byrnes defines conceptual understanding as having knowledge of principles and what they mean (Byrnes, 1996). Conceptual understanding is considered to be one of high level thinking skills in cognitive psychology (Novak, 1988; Piaget, 1970). Providing a proper learning environment and experiential basis for students to develop conceptual understanding is one of the most important goals of effective instruction (Smith et al., 1993). Correspondingly, conceptual understanding is a critical outcome of the educational process and knowledge construction (Carey, 1985; She, 2004). This does not, by any means, suggest that teachers need to teach concepts and students need to memorize definitions. On the contrary, in order to promote conceptual understanding, students should learn in context and experience the concept or phenomenon learned (Bruner, 1990). Various methods to promote studentsÕ conceptual understanding in science education were investigated in the past few years. Incorporating active learning methods, such as drawing and analyzing diagrams, writing summaries, and solving problem, were found to promote studentsÕ conceptual understanding (Gobert and Clement, 1999; Huffman,
Transforming an Introductory Programming Course studio format improved studentsÕ conceptual understanding (Dori and Belcher, 2005a). As it evolved throughout the recent years, the studio format, in its contemporary version, entails the incorporation of personal computers and one-on-one interactions with the instructors, thus limiting the number of participants and/or requiring the course to be held in a specially equipped computer classroom. For instance, in the TEAL project at MIT, a special learning environment was designed to include large round tables (Dori and Belcher, 2005b), and in the Dynamic Systems studio (Glinkowski et al., 1997) the number of students were limited to fit into a computer laboratory. Conversely, Studio-1.00 was designed to implement the characteristics of the studio format, that is: activity-based pedagogies, collaborative learning, and context-rich problems, in a standard lecture hall using the laptops as a tool for facilitating mobility, flexibility and accessibility of the learning materials. PROJECT DESCRIPTION AND UNIQUENESS Studio-1.00 The Studio 1.00 project aimed at the development of an innovative learning environment and a new curriculum for supporting the teaching of MITÕs subject 1.00: Introduction to Computation and Engineering Problem Solving. For the past 35 years, this introductory course has been an important component of the engineering curriculum at MIT. It is considered to be a core course for learning the basics of programming; hence, it is offered every term, with enrollments between 100 and 200 students each semester. The course is designed for students who need to make use of computers for work in their respective disciplines. It is not intended for computer science majors although every semester a few of them take it to expand their experience in software engineering. Like most engineering courses in higher education, Introduction to Computation and Engineering Problem Solving was traditionally taught as a lecture course accompanied by weekly tutorial sessions. Though the course was well received, it did not offer many opportunities for contact or collaboration between instructors (faculty and teaching assistants) and students. With the rise of interactive, object-oriented computing, the instructors of the introductory programming course faced the challenge of reinventing the course to teach computation in a way that reflects todayÕs software development paradigms.
327 During the academic year of 2001–2002, the instructors decided to change the course content from the C++ language to Java, as well as apply active learning techniques. The traditional version of the course consisted of three weekly lectures and one weekly recitation section, each 60 min long. The students were expected to use a computer cluster on their own free time for solving the exercises given in the tutorials and work on their problem sets. The assessment of studentsÕ achievements included two quizzes, ten problem sets and one final examination. The new course format: Studio-1.00, consisted of three 90 min sessions per week, and 1 h per week of small group tutorials. It involved enhanced active learning techniques, including interactive programming, visualization, and the exploration of software development methods, all facilitated by the use of mobile laptop computers. The studio format differed from the traditional format in: (a) New learning materials were developed including real-life problems, visualization, and, handson assignments. (b) More time was allocated for each session, so students will have time to think about and solve problems in the classroom where instructors and teaching assistants were available to provide assistance on an as-needed basis. (c) The sessions were located in regular lecture halls and relied extensively on network connectivity. (d) The students used their laptop computers for working on their course assignments. They did not need to go to computer clusters. (e) More instructors were present at the active learning sessions. Two lecturers and an average of five teaching assistants (TAs) guided the students each session. Although the instructional format changed from traditional to studio format, the grading processes stayed the same. Introduction to Computation and Engineering Problem Solving course is taught twice a year, in the fall and spring semesters. In each semester, different students enrolled, and the course was taught by different pairs of lecturers. The Studio-1.00 project applied two active learning (AL) techniques: PartialAL and Extensive-AL. The Partial-AL model was applied by the instructors teaching the fall semester, introducing new concepts within a single lecture followed by a broad active learning session; The
328
Barak et al.
Extensive-AL model was applied during the spring semester, introducing new concepts through a sequence of short lecture-style presentations, each followed by a short active learning session. The differences between the three instructional strategies are summarized in Table I. Students enrolled in the revised introductory programming course were assigned with a loaner laptop for the entire semester. The laptops were equipped with wireless cards and with a Java Integrated Development Environment (IDE), which enables the creation, development, testing and debugging of Java applets and applications. The laptops fulfilled four purposes: (a) providing students with an easy and convenient hands-on computing experience in a large lecture hall setting, (b) enabling access to the course website or other Java resources on the Web, to read lecture summaries and download parts of Java code, (c) enabling immediate implementation of new programming concepts or procedures taught in class, and, (d) providing students with immediate feedback, from both the IDE program and the instructors. METHODS Research Objective, Population and Methodology Literature in educational research have suggested that the integration of active learning strategies, as part of the curriculum, may advance studentsÕ learning achievements and their conceptual understanding of the learning material (Johnson et al., 1998; West and Pines, 1985). This hypothesis raised the following research questions: 1. Are there differences in studentsÕ learning outcomes between the traditional (lecture-based) and the two studio models: partial and extensive active learning? 2. Are there differences in studentsÕ understanding of key concepts between those who regularly
attend the active learning sessions and those who seldom do? 3. How effective are different teaching and learning strategies applied in the studio sessions in helping students understand the material taught in the course? Data from a total of 502 students spanning five semesters were collected. Two semesters were taught traditionally, before the induced change; in two semesters the studio partial-AL model was applied; and in one semester, the studio extensive-AL model was applied. Table II presents the distribution of students by academic year and major course in the traditional teaching courses and the partial- and extensive-AL courses. As presented in Table II, Introduction to Computation and Engineering Problem Solving course consists of students from diverse academic backgrounds and different majors. The fall semesters included few or no freshmen, and had an average of 15% graduate students; in the spring semester there were about 15% freshmen and proportionately fewer graduates. In all semesters, most of the students, 55–60%, were engineering majors, and around 30% majored in management and economics. The students participating in this study had a similar background since no statistically significant differences between semesters were found in the distribution of the studentsÕ majoring course or academic index. The ‘‘mixed methods research’’ model (Johnston and Onwuegbuzie, 2004) was employed in the current study, using both quantitative and qualitative methodologies in the analysis and interpretation of data. In order to assess differences in learning outcomes, we analyzed and compared the final grades of cohorts studying before and after the change in teaching format, comparing measured differences between various pairs of three research groups: traditional teaching, partial-active learning, and extensive-active learning.
Table I. Traditional Teaching versus Two Models of the Studio Format
Programming language Lectures per week Recitations per week Assessment of studentsÕ achievements Teaching method
Traditional teaching
Studio format: Partial AL
Studio format: Extensive AL
C++ 3*60 min 1*60 min around 20 students Two quizzes, 10 problem sets and one final examination Lectures–teacher centered auditory presentations
Java 3*90 min 1*60 min around 10 students Two quizzes, 10 problem sets and one final examination Active learning—a single lecture followed by a broad active learning session
Java 3*90 min 1*60 min around 10 students Two quizzes, 10 problem sets and one final examination Active learning—a sequence of short lectures, each followed by a short active learning session
Transforming an Introductory Programming Course
329
Table II. The StudentsÕ Distribution (in percentage) by Academic Year and Major Course Traditional teaching
Year Freshmen Sophomore Junior Senior Graduate Major Engineering Science1 and Mathematics Management and Economics Other2 1 2
Studio format: Partial AL
Fall 1999 N = 60
Fall 2000 N = 91
Fall 2003 N = 98
Fall 2004 N = 83
Studio format: Extensive AL Spring 2004 N = 170
3 27 20 40 10
4 13 33 25 25
– 30 29 24 17
1 38 17 36 8
15 43 21 14 7
54 8 35 3
56 12 28 4
59 7 31 3
59 10 26 5
54 18 23 5
Physics, Chemistry, Biology, Earth Atmospheric & Planetary Sciences, Brain and Cognitive Sciences. Architecture, Urban Studies and Planning, Humanities, Linguistics and Philosophy.
In order to investigate studentsÕ conceptual understanding, conceptual questions were added to the quiz and the final examination. The ‘‘conceptual questions’’ were open-ended in the sense that they did not require students to solve a problem or write lines of code, but rather to explain a certain programming process or concept, and provide examples or strategies for supporting their answers. The studentsÕ responses to these questions were analyzed quantitatively as well as qualitatively by two educational experts who were not involved in teaching the course. Finally, in order to establish the effectiveness of different teaching and learning strategies applied in the studio sessions, an online survey was administered (Appendix I). The survey was developed and validated by the course instructors and an educational researcher. It included the following closedended questions answered on a 1-to-5 Likert scale: • How effective were the following teaching and learning strategies in helping you understand Java programming? • During the studio sessions, how often did you: (a) Attempt solving a problem in class, (b) Succeed in solving a problem in class?
The internal consistency, measured by CronbachÕs a, for each question was: 0.75 and 0.82, respectively. FINDINGS
Interestingly, but not surprisingly according to the constructivist theory (Bruner, 1990; von Glasersfeld, 1995), the highest grade average was received by students who studied in the Extensive-AL course. Students who participated in the Partial-AL courses were only a half a point away; and the lowest average, by 6 points, was received by students who participated in the traditional format. Moreover, the final grades of traditionally taught students had the highest standard deviation (i.e., grade distribution), and relatively low minimum and maximum. These results suggest that the studio format in both its Extensiveand Partial-AL models induced studentsÕ learning outcomes. One way Analysis of Variance (ANOVA) showed a statistically significant difference between the different instructional formats (F = 13.30, p < 0.001). HochbergÕs GT2 post hoc test (for nonequal sample sizes) showed statistically significant mean differences between traditional instruction and Partial-AL (Mean difference = 6.75, p < 0.001) and Extensive-AL (Mean difference = 7.26, p < 0.001). No statistically significant differences were indicated between the grades of the two studio formats. The comparison of studentsÕ final grades as represented in letters indicated two compelling results Table III. Means and Standard Deviations of StudentsÕ Final Grades by Instructional Formats
Traditional Teaching versus the Two Studio Models
Instructional format
A comparison between the three instructional formats: traditional teaching, partial- and extensiveAL are presented in Table III.
Traditional 151 79.26 15.67 Studio: Partial-AL 181 85.47 12.85 Studio: Extensive-AL 170 86.00 9.92
N
Mean SD
Minimum Maximum 22.94 21.78 50.16
97.56 99.88 100.00
330
Barak et al.
(Figure 1). First, although students in the traditionally taught course received a slightly higher percentage of A+ grades, the percentages of students receiving A and A) in the studio designed courses were meaningfully higher. Second, the percentage of the students who received less than 56 points on their final grade (i.e., students who failed or dropped out) was significantly lower in the studio courses than in the traditional courses.
The Effect of Studio-based Learning on StudentsÕ Conceptual Understanding In order to investigate the differences in studentsÕ conceptual understanding between those who regularly attended the active learning sessions and those who seldom do, we examined their responses to conceptual questions on one of the quizzes and the final examination. Unlike most of the questions on the tests that required students either to write segments of code or select the correct statement in multiple choice questions, the conceptual questions required students to explain a phenomenon and provide examples or strategies to support their answers. An analysis of variance (ANOVA) test was conducted to examine the difference between two inter-class groups (Those who attended more than 80% of the sessions versus those who attended less) on their responses to conceptual questions. Since conceptual questions were integrated into the quiz and final examination of only one semester, and because we could use data only from students who signed a consent form, our sample was reduced to 73 students.
45
Traditional N=151 Partial AL N=181 Extensive AL N=170
40
Percentage of students
35 30
The comparison of the scores of the two interclass groups on their responses to conceptual questions indicated a statistically significant difference. It was found that students who regularly attended the active learning sessions were able to conceptualize programming principles better than their peers, as presented in Table IV. Table IV shows that students who participated in the studio-based classes answered the conceptual questions more correctly than their peers. In addition, it was found that on average (considering both the quiz and the final examination), 60% of the students who regularly attended the studio sessions received the maximum score, and only 5% of them did not answer the question or answered incorrectly. Conversely, only 40% of the students who attended the studio sessions inconsistently received the maximum score, and about 15% of them did not answer the question or answered it incorrectly. Having these results, we decided to investigate the studentsÕ answers to the conceptual questions in an attempt to find the cause for the differences in their scores. Content analysis of the studentsÕ responses raised two aspects that characterize the difference between research groups in the way they composed their answers: ‘‘complexity’’ and ‘‘multiplicity’’. Differences in ‘‘complexity’’ between the research groups are presented in Figures 2 and 3, which show studentsÕ answers to the following conceptual question presented in the quiz: What is round-off error? Please give one example that demonstrates its significance.
Figure 2 is an answer given by a student that regularly participated in the studio sessions; whereas Figure 3 is an answer written by his classmate that attended class inconsistently. The answer presented in Figure 2 does not provide a clear ‘‘out of a book’’ definition of round-off error. Instead, the student provides his interpretation of the concept, an example of the process, and an explanation of why the process is important.
25
Table IV. Means, Standard deviations, and ANOVA Tests of StudentsÕ Conceptual Question Scores, on the Quiz and Final Examination, Sorted by Attendance
20 15 10
Conceptual question
5 0 A+ A (94- A- (91- B+ (100- 92) 88) (8795) 85)
B (84- B- (78- C+ 79) 75) (7470)
C (6962)
Fig. 1. Students letter grades by instructional formats.
C(6157)
F (5621)
Inter-class group N Mean SD attendance
Quiz (0–5 points) Regular Inconsistent Final exam Regular (0–10 points) Inconsistent
40 33 40 33
4.25 3.15 8.65 7.10
F
p
1.35 8.16
