* Ronny Richardson and Max North Business Administration Department Southern Polytechnic State University rrichard@spsu.edu and max@spsu.edu ABSTRACT Online course offerings are one of the most effective and efficient delivery methods for contents and skills globally. However, the level of trust in online courses is very low, particularly in respect to the credibility of online assessments. Simply, if the trust in online assessment can be equated to the level of the classroom assessment environment, trust in online courses will increase accordingly. To investigate and strengthen the trust of online assessment, the researchers have incorporated a simple and obvious solution: proctoring the tests portion of the online courses and comparing with non- proctored exams. In this study, 11 proctored exams against 22 identical non- proctored exams. In 19 out of 22 cases, the class average on the non-proctored exam was higher than the class average on the corresponding proctored exam; the difference was statistically significant at the 95 percent trust level in 15 cases. Findings indicate that proctoring seems to be an effective common sense approach and in turn potentially strengthening the trust in online courses and distance education. INTRODUCTION Although trust in online courses and degrees seems to be low, online course offerings are one of the most effective and efficient methods of delivery of content and skills. There are many common senses advantages to the online course approach, including the wide flavor of degree programs and classes offered, flexible study times, and the ability for students to balance between career and education. The online movement is evolving dramatically, leading to more creative philosophies such as ___________________________________________ Copyright 2013 by the Consortium for Computing Sciences in Colleges. Permission to copy * without fee all or part of this material is granted provided that the copies are not made or distributed for direct commercial advantage, the CCSC copyright notice and the title of the publication and its date appear, and notice is given that copying is by permission of the Consortium for Computing Sciences in Colleges. To copy otherwise, or to republish, requires a fee and/or specific permission. 266 CCSC: Central Plains Conference Massive Open Online Courses (MOOC), which promotes sharing information and has created many opportunities for teaching and learning in a variety of disciplines [1]. Among the pioneers of the MOOC is the MIT OpenCourseWare-provided, individual, self-paced learning environment and the recently incorporated, open-source learning management system Moodle. At a more limited level, many universities and colleges are offering either Hybrid or Fully online courses in many departments, such as Business Administration, Accounting, Mathematics, English, Physics, and Chemistry, just to name a few. However, the level of trust in online course outcomes, and especially in assessment (e.g., exams), has been very low [2]. In general, the delivery method of the online courses is commonly acceptable, but assessment lacks credibility, due to the possibility of cheating in many different ways while taking the online tests [3, 4, 5]. Therefore, if the element of the cheating can be reduced to the level of the classroom testing environment, the trust and credibility of the online course assessmentand by extension, online course outcomeswill increase tremendously [6, 7, 8]. To strengthen the trust of online assessment, the authors have devised a simple solution by proctoring the tests portion of online courses. To validate this approach, a series of the courses were selected and an extensive experimental research was conducted. METHODOLOGY To validate our approach, the testing facility in the XYZ University (XYZ) Continuing Education Center arranged proctored testing services for four sections of Online Business Administration courses. Specifically, four courses were selected for proctoring (see Table 1). These courses ranged from freshman undergraduate to advanced graduate courses. For each of these classes, except ACCT 6003, the authors had testing results without proctoring from Spring 2011 and Fall 2010, and there were no significant changes to the courses or exams over this period. A direct comparison, therefore, made sense. ACCT 6003 is only taught in the fall, so we were initially going to compare Fall 2010 to 2011, and we did indeed use that comparison. However, we ended up offering ACCT 6003 in the first and second eight weeks of Fall 2011. Thus, the first eight weeks were proctored and the second eight weeks were not proctored, creating two non-proctored sections for comparison for all four courses. Experimental Courses Exam Structures and Miscellaneous ACCT 2101 Accounting I This course is taken by a number of majors on campus. It has four exams, three chapter exams, and a comprehensive final exam. All exams are true/false and multiple-choice, so there is no possibility of instructor bias in the grading. ECON 1101 Introduction to Economics This course is a core class. It has three exams, two chapter exams, and a final exam. All three are true/false and multiple-choice, and there is no instructor involvement in the grading, so there is no possibility of instructor bias in the grading. In fall 2011, all three exams were proctored. 267 JCSC 28, 5 (May 2013) MGNT 3105 Management and Organizational Behavior This course is taken by a number of majors on campus and is required for a minor in Business, so many students take it. It has four chapter exams and a comprehensive final exam. Extended University felt this was too many exams to proctor, so only the first and third chapter exams and the final exam were proctored. The second and forth chapter exams were not proctored, which gave us the ability to compare proctored and non-proctored exams within the same course. All exams are true/false and multiple-choice, so there is no possibility of instructor bias in the grading. ACCT 6003 Accounting Theory Only taken by students in the Master of Science in Accounting graduate program. This class uses cases and other non-exam material, so there is only one exam in the class. Unlike the other exams in the study, this course did not use an exam that was exclusively true/false and multiple-choice. For this exam, 76 percent was true/false and multiple choice, while 24 percent was instructor-graded short answer and essay questions. Since only 24 percent was instructor-graded, it was felt that the impact of any instructor bias in grading would be minimal. Table 1. Exam structures and miscellaneous information of the Experimental courses Fifty five (55) students attended our facility, and an additional ten (10) students attended external test proctoring sites not directly connected with XYZ. The number of tests varied by class. For the students who took tests at our facility we were involved in scheduling testing times within the guidelines prescribed by the instructors, verifying the identification of the test takers, providing secure testing computers and monitoring the test takers for the duration of the test. These services were provided as part of an experiment at no charge to the students of the Business Administration department. For the students who took tests at external sites, we were involved in helping the students identify acceptable testing locations in their area, communicating the proctoring requirements to these locations, and reimbursing the students for fees paid to the external proctoring sites. The Continuing Education Center absorbed the cost of the test reimbursements. To make comparisons between courses meaningful, all exam scores were scaled to 100 points before computing the mean and standard deviations. (Unscaled points ranged from 50 to 1,000.) The median was also computed to check for outliers but comparing the means to the medians did not raise any concerns regarding outliers. RESULTS Collectively and distinctly, this experiment provided us the results for eleven proctored exams. In ten of those eleven exams, the score on the proctored exam was lower than on both of the two non-proctored exams to which it was compared, often by very significant amounts. The single exception was the final exam in ECON 1101, where the proctored final exam results were virtually identical to the non-proctored results. The largest difference between proctored and non-proctored exams was in ACCT 2101, and the smallest difference was in the graduate ACCT 6003 course. Detailed results of the selected courses are given on a course-by-course basis in Table 2. In order to provide a quick glance at the results of this experiment, we are providing the following Table 3 which compares the non-proctored exams to proctored exam showing any statistically significant difference at the 95% trust level. 268 CCSC: Central Plains Conference Results of Non-Proctored Exams Compared to Proctored Exams Spring 2011 Significant at 95% Fall 2010 Significant at 95% ACCT 2101 Exam #1 Higher Yes Higher Yes Exam #2 Higher No Higher No Exam #3 Higher Yes Higher Yes Final Exam Higher No Higher Yes ECON 1101 Exam #1 Higher No Higher Yes Mid-Term Lower Yes Higher Yes Final Exam No Difference No No Difference No MGNT 3105 Exam #1 Higher Yes Higher Yes Exam #3 Higher Yes Higher Yes 269 JCSC 28, 5 (May 2013) Final Exam Higher Yes Higher Yes ACCT 6003 Exam #1 Higher No Higher Yes Table 3. Results at a glance of Non-Proctored Exams compared to Proctored Exams To further assist the readers with a quick graphical view of the results, we are providing yet another high level summarized line graph depicting the non-proctored and proctored exams for all four courses in this experiment. See Graph 1. CONCLUSIONS Nineteen out of 22 exams had higher average scores when they were not proctored than when they were proctored. This is much higher than the eleven expected if the results were randomly distributed. This is a strong indicator of significant cheating. While we would expect to see eleven non-proctored exams have higher average scores than the proctored exams if the data were randomly distributed, we would not expect them to be statistically significantly higher. However, fifteen out of the nineteen differences were statistically significantly higher at the 95% trust level. This too is a strong indicator of significant cheating. Additionally, it was noted that: (i) When some of the exams in a course were proctored and other exams in the same course were not proctored, grades on the non- Graph 1. High level summarized line graph showing both non-proctored and proctored exams 270 CCSC: Central Plains Conference proctored exams were also lower than the same non-proctored exams in other sections of the course. (ii) The difference between proctored and non-proctored exams was much lower at the graduate level than at the undergraduate level. Since the non-proctored exams in the proctored section of MGNT 3105 had average scores in line with the averages on the proctored exams and lower than the averages on the comparable exams in the non-proctored sections of MGNT 3105, proctoring some of the exams appears to be a strong deterrent to cheating even when not all of the exams are proctored. DISCUSSION AND RECOMMENDATIONS From our perspective, the test proctoring experiment was conducted without any significant problems. There was a small learning curve for both instructors and students who had not been involved with proctoring, but nothing that warrants any changes to our already established procedures. In the future, we would suggest that a meeting be held with all faculty new to on-line proctored testing prior to the beginning of the semester. In order to accommodate future demand for proctoring it will be important for the testing facility to be informed before the beginning of the semester, how many classes, students and exams will require proctoring, as well as the dates for the exams. Instructors may need to be flexible about the dates for testing, since it is likely that testing windows across classes will tend to coincide. It may be necessary for administrators to enforce flexibility from their instructors to accommodate the finite resources of the testing facility. REFERENCES [1] Martin, F.G., Will massive open online courses change how we teach? Communications of the ACM, 55, 8 (August 2012), 26-28. [2] Prince, D.J., Fulton, R.A., Comparisons of proctored versus non-proctored testing strategies in graduate distance education curriculum, Journal of College Teaching & Learning, 6, 7 (November 2009), 51-62. [3] Rowe, N., Cheating in online student assessment: Beyond plagiarism, Online Journal of Distance Learning Administration, Retrieved on June 1, 2012 on www.westga.edu, 2004. [4] Bosch, T., Why Would Someone Cheat on a Free Online Class That Doesnt Count Toward Anything? Posted Monday, Aug. 20, 2012, at 2:48 PM ET http://www.slate.com/blogs/future_tense/2012/08/20/coursera_plagiarism_why_ would_students_cheat_ina_free_online_class_that_doesn_t_over_academic_cre dit_.html [5] Rogers, C. F., Faculty perceptions about e-cheating during online testing, Consortium for Computing Sciences in Colleges, 22, 2 (December 2006), 206- 212. [6] Greenberg, R., Online Testing, Techniques: Making Education & Career Connections, 73(3), 26-28, 1998. 271 JCSC 28, 5 (May 2013) [7] Baker, R., Papp, R., Academic Integrity Violation in the Digital Realm, Proceedings from the Southern Association for Information Systems 2003 Annual Conference, 193-202, 2003. [8] Carlson, R., Assessing Your Students: Testing in the Online Course, Syllabus, 12(7), 16-18, 2000. 272