CSECS 2011, pp. 000 - 000 The 7 Annual International Conference on Computer Science and Education in Computer Science, July 06-10 2011, Sofia, Bulgaria DIMENSIONS OF COURSE DESIGN AND DELIVERY AND THEIR EFFECT ON STUDENT SATISFACTION/PERCEPTION IN ONLINE LEARNING Tanya ZLATEVA, Svetlana WILLETT, Suresh KALATHUR, Robert SCHUDY, Leo BURSTEIN, Lou CHITKUSHEV, Masatake SAITO, Elizabeth M HAINES Abstract: Online learning is a disruptive technology that has significantly transformed the educational landscape in a very short time. It is therefore of imminent importance to understand the major factors that determine the online learning experience. This paper analyzes student perceptions based on course evaluations from 53 computer information courses with a total enrollment of 4,089. A multiple regression analysis of factors along four dimensions—course, instructor, facilitator and technology—identified as significant factors the course material organization, discussions, assignments, and the instructor’s ability to present. This indicates that student satisfaction is independent from the delivery medium. A correlation analysis of the relationship of course size and student satisfaction suggests that courses with more than 100 students pose a greater challenge but that this challenge can be addressed by adding novel multi-media components such as synchronous video-collaboration sessions and ad-hoc whiteboard discussions. Keywords: online learning, student perceptions, face to face learning, course design parameters, media-rich, synchronous, asynchronous, correlation and regression analysis, ACM Classification Keywords: computer science education, computer information education, multi-media, animation, simulation 2 1 Zlateva et al. INTRODUCTION Online learning continues to expand within a fundamental dichotomy: lauded for its dynamism, flexibility, multi-modal technology its academic validity continues to be questioned for the (presumed) disconnect between students and teachers. For the last seven years the growth of online enrollments has consistently far exceeded the growth of the overall college population [Allen, 2010]. In the United States more than 5.6 million students (or 30% of the higher education population) were taking at least one online class in the fall 2009 semester. This is a substantial 21% increase as compared to a less than 2% increase of the higher education population [Allen, 2010]. In contrast to this the perception of the quality of online education improved more modestly: 66% of academic leaders believe it to be the same or superior to face-toface education as compared to 57 % in 2009. Independently on where one stands in the online education debate it is clear that we are witnessing a disruptive innovation in one of the most traditional fields of human endeavor. It is therefore of eminent importance to understand the major factors that determine the online learning experience. Considerable work has been devoted to this problem and a growing number of studies address specific design aspects, courses in different fields, as well as student and faculty perceptions of online learning in general, e.g. [Volery, 2000], [Soong et al., 2000], [Sun et al., 2008]. In a previous study [Zlateva et al., 2010] we introduced a parametric model for online courses that included class size, course content, assessments, and student satisfaction and used it to analyze the online learning experience based on data from 51 online courses delivered in the MS in Computer Information Program at Boston University’s Metropolitan College. This paper follows up on the previous results and takes a more in-depth look at student perceptions as reflected in the Dimension of Course Design and Delivery CSECS 2011, July 7-11 2011, Sofia, Bulgaria 3 student course evaluations. The data set was expanded with three more courses and more importantly with the full student evaluation survey. All courses were part of the Master’s in Computer Information Systems that is offered online in a fully asynchronous mode with optional live webinar and video-conferencing for discussions sessions. The courses are offered in an intensive seven week format instead of the traditional 14 week semester and are implemented in Blackboard Vista with mediarich online content, discussion boards, videos, simulations, selfassessments, virtual laboratories, online exams. The courses are developed and delivered almost exclusively by full-time faculty and capped at 150 students. In addition to the faculty of record for the course a facilitator is assigned to every 15 students with responsibilities to answer questions, lead discussions and grade homework assignments. 2 DATA AND METHODOLOGY The raw data was drawn from 54 online classes delivered over seven semesters from Summer 2008 to Summer 2010. A large number of parameters describing course structure, assessments, class size, and student perception was collected. In this paper we discuss the major factors for student satisfaction and follow up on our initial findings [Zlateva et al., 2010] that pointed to a negative impact of class size and student satisfaction. The data used for the present discussion include along the class size responses to an online survey of 30 questions that are rated on a five-level Likert scale with 1 (negative/strongly disagree) and 5 (positive/strongly agree). The latter fall into four groups that assess perceptions of the course, instructor, facilitators, and technical support. The exact wording of the questions is given in Table 1. The response rate for the survey ranged from 27.78% to 59.62% with a mean of 52.38% and standard deviation 8.89%. A confidence interval of 95%, (including mean response rates from 34.96 to 69.81) was computed and The Title of the Section 4 Zlateva et al. led to the exclusion of one class. The remaining aggregated data from 53 classes with an overall enrollment of 4,089 provided the basis for the regression analysis. The survey questions directly relate to course content and design parameters such as intellectual challenge, structure, discussion, multiple modalities; instructor qualifications and ability (subject mastery, presentation skills, grading, openness to question); facilitator contributions (clarity and timeliness of response, encouraging of discussions, added value); and course technology (navigation, user manuals, accessibility, student services). Determining the degree to which the individual parameters contribute to the overall course perception is at the heart of an effective course improvement and development. Toward this goal we undertook an analysis of the relationships within the questionnaire as well as between the individual survey questions and the class size. The correlation matrix was computed as an exploratory first step for identifying potential dependencies. It displayed two salient characteristics: a considerable complexity and range of correlation levels (from the slightly negative to over 0.9 correlation coefficients) and consistent negative correlation of class size with all survey questions. To better understand the underlying relationships we analyzed the significance of the individual survey questions through multiple regression analysis (section 3) and the relationship of class size with overall course satisfaction in different time period (section 4). Dimension of Course Design and Delivery Table 1: Student Evaluation Survey Course CC01 – I found the class intellectually challenging CC02 – Course materials were well organized and clearly presented CC03 - Discussion topics enhanced the learning experience CC04 - Assignments furthered understanding of course content CC05 - Textbook/cases/course materials furthered understanding of course content CC06a - Animations/Simulations enhanced understanding of key concepts CC06b - Videos enhanced understanding of key concepts CC06c - Webinars/Web-Meetings (e.g. GoTo meetings) enhanced understanding of key concepts CC06d - Video-Conferencing enhanced understanding of key concepts CC07 - I would recommend this course to others CC08 - The overall course experience was: Instructor CE09 - The instructor’s mastery of the course materials was: CE9a – The instructor’s ability to present course material is: CE10 - The instructor’s grading criteria are fair and clear CE11 – The instructor was supportive and responsive to my questions CE12 – Assignments were returned in a timely manner by the instructor CE13 – I would rate the instructor overall as: Facilitator FE14 – Facilitator feedback was informative and clear FE15 – Assignments were returned in a timely manner by the facilitator FE16 – The facilitator responded to my questions in a timely manner FE17 – The facilitator added value to my learning experience FE18 – The facilitator’s ability to encourage questions/discussion s is: FE19 – The facilitator’s overall rating is: Course Technology CT20 – Navigation allowed easy access to information. CT21 – Instructions as to how to use media technologies (audio, video, CD-ROMs, etc.) were CT22 – Access and response time to courseware system was: CT23 – Technology support was: CT24 – I was able to resolve course problems with the help of Technical Support in a timely manner CT25 – The Student Services Representative/Manager was: CT26 – Overall, I would rate the course technology and support as: 3 MULTIPLE REGRESSION ANALYSIS AND DISCUSSION OF RESULTS The correlation matrix revealed a number of strong correlations between survey questions, some predictable, (e.g. there is a 0.91 correlation between the course overall rating and the degree to which the course is recommended to others) some not so obvious (e.g. it is not readily clear why the instructor’s ability to present strongly correlates with a coefficient of 0.72 with his/her support and responsiveness). Given the complexity of the learning experience, its multiple aspects and the large number of parameters needed for an accurate representation it is critical to identify the most significant factors that shape the overall course perception. Towards this goal we performed multiple regression analysis for the “overall course experience” (CC08) as the dependent variable. The independent variables were drawn from the remaining parameters with the exception of parameters in close to linear dependence (r > 0.9) with the predictor, such as “I would recommend this course to others” (CC07), and parameters that address the overall experience and do not yield information about specific aspects of course design, such as “I would rate the instructor overall as”(CE13), “The facilitators’ overall rating is”(FE19), “ Overall, I would rate technology and support”(CT26) . Additional parameters were excluded to ensure that the predictor set will be free of strong pairwise correlations, i.e. correlation coefficients between predictors will be less than 0.7. In order to exclude highly correlated pair(s) regression computation was performed for all combinations of uncorrelated predictors. The combination with the largest R-Square was retained for the final regression computation. The multiple regression was performed in two stages: First, the statistically significant parameters within each category were determined. These category parameters were then combined and considered predictors in an integrated model including the course, instructor, facilitator, and CSECS 2011, July 7-11 2011, Sofia, Bulgaria 7 educational technology aspect. Regression analysis of the integrated model identified the strongest predictors for the overall course experience. 3.1 Course Dimensions The course category has eleven parameters and after excluding the dependent variable “overall course experience” (CC08) and the almost linearly related “recommend course to others” (CC07) we are left with a pool of nine candidates for the predictor set. Regression including all nine variables results in an R-Square of 0.8656, i.e. the predictors account for 86.56% of the variance of the dependent variable CC08 . However, two of the nine predictors—“webinar/web-meetings”(CC06c) and “video-conferencing” (CC06d)—are highly correlated (r=0.91 ). Table 1: Course dimensions. (statistically significant for p < 0.05; 95% confidence level) Variable Pr > |t| (p value) Intercept 0.0026 CC01 intellectually challenging 0.6046 CC02 materials organized clearly <.0001 CC03 discussion enhanced learning 0.0033 CC04 assignments furthered understanding 0.0019 CC05 materials furthered understanding 0.2406 CC06a animations/simulations enhanced 0.0609 CC06b videos enhanced 0.3302 CC06d video-conferencing enhanced 0.5253 Regression analysis for the two uncorrelated predictor combination, without CC06c and without CC06d, yields an R-Square values of 0.8655 The Title of the Section 8 Zlateva et al. and 0.8648 respectively; thus we exclude CC06c and keep CC06d for further analysis. The p-values for the parameters obtained from the regression are shown in Table 1 and reveal that there are only three statistically significant predictors. 3.2 Instructor Dimensions The instructor category contains five parameters and when all are used as predictors R-square is 0.8163. Strong correlations exist between CE09 and CE09a, and between CE11 and CE09a. The combination excluding CE09 and CO11 has the highest R-square (0.8040) among all parameter combinations with no strong pairwise correlation and is retained for further analysis. Table 2 shows the p-values of the parameters. There are two statistically significant predictors—the instructor’s ability to present (CS09a) and the fairness and clarity of the grading criteria (CE10). Table 2: Instructor dimensions. (statistically significant for p <0.05; 95% confidence level) Variable Pr > |t| (p value) Intercept 0.0028 CE09a ability to present <.0001 CE10 grading criteria 0.0013 CE12 assignments returned timely 0.2730 3.3 Facilitator Dimensions The survey questions in the facilitator category show substantial dependencies (between FE14-16, FE14-17, FE14-18, FE15-16, FE1517, and FE16-17). Regression over all variables yields an R-square of Dimension of Course Design and Delivery CSECS 2011, July 7-11 2011, Sofia, Bulgaria 9 0.3556. The parameter combination retaining FE15 and FE18 has an R-square of 0.3468 which is the largest of all combinations free of strong correlation. The resulting parameter estimates and p-value (Table 3) indicate that only one parameter, the facilitator’s ability to encourage questions and discussions (FE18), is statistically significant. Table 3: Facilitator dimensions. (statistically significant for p <0.05; 95% confidence level) Variable Pr > |t| (p value) Intercept FE14 Facilitator feedback informative and clear 0.0309 FE15 Assignments returned in a timely manner 0.2670 FE18 Facilitators encourage questions/discussion 0.0047 3.4 0.6794 Educational Technology Dimensions Educational technology parameters were very weakly correlated and therefore were all included in the regression that yielded an R-square of 0.3162 and as a single statistically significant parameter “navigation allows easy access to information” (Table 4) . Variable Table 4: Educational technology dimensions. (statistically significant for p <0.05; 95% confidence level) Pr > |t| (p value) Intercept 0.1229 CT20 Navigation easy to access CT21 -Instructions for technology use are clear 0.0537 CT22 Access and response time to courseware system 0.1746 CT23 Technology support CT24 Technical support timely 0.1432 CT25 Student services representative/manager: 0.3415 The Title of the Section 0.4273 0.5790 10 3.5 Zlateva et al. Significant Factors for Overall Course Satisfaction The regression model that integrates the statistically significant factors of the categories was characterized by an R-square of 0.8988 and yielded four statistically significant predictors—clear organization of the material (CC02), discussion (CC03), assignments (CC04), and the instructor’s ability to present (Table 5). Table 5: Significant factors for overall course experience (statistically significant for p <0.05; 95% confidence level) Variable Pr > |t| (p value) Intercept 0.0014 CC02 materials organized clearly 0.0182 CC03 discussion enhanced learning 0.0539 CC04 assignments furthered understanding 0.0065 CE09a instructor’s ability to present 0.0004 CE10 grading criteria 0.0765 FE18 facilitators encourage questions/discussions 0.7481 CT20 navigation easy access 0.2484 In a learning environment distributed through cyber space and so crucially dependent on technology for access, communication and assessment it is reasonable to expect that perception are substantially shaped by the nature and quality of the online medium. The surprising result in our analysis is that none of the statistically significant factors relates to a technology dimension. This is in contrast to prior studies that identified technology as critical for success (e.g. [Volery, 2000], [Soong et al., 2000]) Dimension of Course Design and Delivery CSECS 2011, July 7-11 2011, Sofia, Bulgaria 11 Instead the determining factors—class structure, discussion, presentation ability--are mainstays of learning theory since its very beginnings. In theories of distance education, more notably Moore’s theory of transactional distance [Moore, 1993], they are considered the basic dimensions that define the learning experience. 4 CLASS SIZE AND STUDENT SATISFACTION Earlier analysis based on data from 2008 to 2009 identified a problem with satisfaction in higher enrollment courses. For example, not one of the six courses with more than 100 enrollments in 2008 and 2009 had attained a student satisfaction of more than 4.0, while more than a dozen courses with enrollments below 100 had attained satisfaction above 4.0. During these years the correlation coefficient between overall student satisfaction and enrollment was -0.4561 [Zlateva et al., 2010]. We revisited these findings in the expanded data set and took a closer look at the relationship between class size and student satisfaction in consecutive years. We found that the overall trend was only slightly negative with a correlation coefficient of -0.156 (Figure 1). In addition several classes had achieved higher than 4.0 average rating. More interestingly the scatter plots for the three consecutive years showed the correlation changing from -0.434 to -0.368 in the first two years to a slightly positive 0.174 in 2010-2011. (Figure 2-4). We hypothesize that this is due to the addition of novel multi-media interactive components (lecture and problem solution recordings, live classroom, additional animation) and expanding the self-assessment questions and adding term projects that are sequenced with the content. The Title of the Section 12 Zlateva et al. 5 4.5 4 3.5 3 2.5 0 50 100 150 200 Figure 1: Course Overall vs. Enrollments SU 2008-SP 2011 correlation coefficient = -0.15655, 75 classes with 5,951 overall enrollment 5 4.5 4 3.5 3 2.5 0 50 100 150 200 Figure 2: Course Overall vs. Enrollments SU 2008-SP 2009 correlation coefficient = --0.4342, 22 classes with 1,677 overall enrollment Dimension of Course Design and Delivery CSECS 2011, July 7-11 2011, Sofia, Bulgaria 13 5 4.5 4 3.5 3 2.5 0 50 100 150 200 Figure 3: Course Overall vs. Enrollments SU 2009-SP 2010 correlation coefficient = -0.3681, 24 classes with 1,920 overall enrollment 5 4.5 4 3.5 3 2.5 0 50 100 150 200 Figure 4: Course Overall vs. Enrollments SU 2010-SP 2011 correlation coefficient =-0.1745, 29 classes with 2,354 overall enrollment 5 CONCLUSION Our analysis indicates that the main course design and delivery parameters that determine student satisfaction are independent from the delivery medium. Students perceive online courses online based on The Title of the Section 14 Zlateva et al. the same course attributes as students perceive face-to-face courses. They key determiners of satisfaction are the quality and organization of the content, clear grading policy, the instructor’s ability to present the material well, assignments that furthered understanding, and discussions that enhanced learning. A follow up analysis of the relationship between class size and student satisfaction showed that the initially negative correlation has become slightly positive in the last year. We believe this is due by the increasing maturity of the courses and greater experience of the faculty as well as the additional resources allocated to the course. More specifically we added facilitators without group, responsible for conducting synchronous sessions with the students and address more difficult aspects of the material and/or demonstrate problem solutions. The synchronous sessions were conducted in a multi-media video-collaboration environment and included formal lecture presentations, question and answer sessions, ad hoc whiteboard discussions, chat, or/and other multimedia content. The sessions were recorded and made available for review. Further analysis is needed to better understand how and to what extent each of these aspects contributes to improve student perceptions. 6 BIBLIOGRAPHY [Allen, 2010] I.E. Allen, J. Seaman. Class Differences: Online Education in the United States 2010. 8th Annual Survey of the Sloan Consortium http://www.sloan-c.org/publications/survey/pdf/learningondemand.pdf [Moore, 1993] M. Moore. Theory of transactional distance. In "Theoretical Principles of Distance Education, Keegan, D. (ed.). Routledge, pp. 22-38. [Soong et al., 2000] M. H. B. Soong, H.C. Chan, B.C. Chua, K. F. Loh. Critical success factors for on-line course resources. Computers & Education, Volume 36, Issue 2, February 2001, Pages 101-120. [Sun et al., 2008] P-C. Sun, R. J. Tsai, G Finger, Y-Y. Chen, D. Yeh. What drives a successful e-Learning? An empirical investigation of the critical Dimension of Course Design and Delivery CSECS 2011, July 7-11 2011, Sofia, Bulgaria 15 factors influencing learner satisfaction. Computers & Education 50 (2008) 1183–1202. [Volery, 2000] T. Volery, D. Lord. Critical success factors in online education, International Journal of Educational Management, Vol. 14 Iss: 5, pp.216 223 [Zlateva et al., 2010] T. Zlateva, M. Saito, S. Kalathur, R. Schudy, A. Temkin, L. Chitkushev. A Unified Approach for Designing, Developing, and Evaluating Online Curricula. 6th International Workshop on Computer Science and Education in Computer Science, Fulda-Munich, Germany, June 2010 7 AUTHORS' INFORMATION Tanya ZLATEVA, Ph.D., Assoc. Professor, Boston University, Boston, MA, USA, zlateva @bu.edu Major Fields of Scientific Research: cyber security, computer science education, visual recognition Svetlana WILLETT, BS, Master’s candidate Major Fields of Scientific Research: statistics Suresh KALATHUR, Asst. Professor, Boston University, Boston, MA, USA, Major Fields of Scientific Research: data mining, programming languages, web technologies, computer science education, visual recognition Robert SCHUDY, Assoc. Professor, Boston University, Boston, MA, USA, Major Fields of Scientific Research: data bases, computer science education Leo BURSTEIN, MS, Senior Architect, Boston University, Boston, MA, US, Major Fields of Scientific Research: educational technologies Lou CHITKUSHEV, Assoc. Professor, Boston University, Boston, MA, USA, Major Fields of Scientific Research: networking, medical informatics, cyber security, computer science education, Masatake SAITO, Visting Asst. Professor, Boston University, Boston, MA, USA, Major Fields of Scientific Research: computer science education, online teaching technologies Elizabeth M HAINES, MS, Instructor, Boston University, Boston, MA, USA, Major Fields of Scientific Research: computer science education, online teaching technologies The Title of the Section