whitney@sciences.sdsu.edu (Roger Whitney) (08/01/90)
[Reprinted with permission from Academic Computing Magazine, March 1990] MICROCOMPUTERS IN THE MATHEMATICAL SCIENCES: EFFECTS ON COURSES, STUDENTS AND INSTRUCTORS Roger E. Whitney Computer Science Group Mathematical Sciences Department San Diego State University San Diego, CA 92182 whitney@sdsu.edu N. Scott Urquhart Experimental Statistics New Mexico State University Las Cruces, New Mexico 88003 nsu@nmsu.edu This article grows out of an effort to integrate computers into two courses at San Diego State University. Computers were intended to perform low level operations for the students. Our goals were to increase motivation and understanding of concepts, and to allow students' to work realistic problems. We felt this approach would particularly aid the weaker students who we frequently observed struggling with mechanical or low level operations. Although our efforts were directed toward improving our students' performance, we found the computer usage had more important impacts on the courses, the instructors, and the department. The content and intent of the courses changed, becoming more relevant to the students' careers. The change in content made evaluation of the impact of the computer usage on student performance difficult; the change in intent made such measurements rather meaningless. The instructors' and the department's work load increased, which may have significant ramifications in the future. We did observed more enthusiasm and better performance among some students. Our initial impression of performance suggested that weaker students performed worse than they should have, rather than better. We spent considerable time investigating the weaker students' performance. From observing these students with wildly differing computer experience, we recognize the necessity of computer literacy. Although both of us are committed to a computer enriched instructional approach, we decided for totally different reasons not to continue this experiment under present conditions. In what follows we discuss our motivation, how we used computers in our courses, the impact this change had, and offer a way to avoid some of the pitfalls we encountered. Our Motivation Our primary motivation in this project was to improve our students' performance. For some time we had noticed that students entering our courses were inadequately prepared and rather unmotivated. As an example of inadequate preparation, computer science seniors commonly struggle with any mathematical operation dealing with logarithms even though they had taken at least a year of calculus. Our students were rarely motivated to go beyond the minimal requirements. Even by the end of our courses students often displayed a lack of conceptual understanding of the material covered. Consequently they had almost no ability to apply the course material to solve problems outside the narrowly focused cookbook problems found in textbooks. Frequently we noticed students memorizing pages of information to compensate for a lack of understanding of basic concepts. Many students did not seem to notice that the course material was relevant outside the class. We were dismayed by this situation for two reasons: the long term consequences and a professional sense of failure. Students eventually must apply some course material in their jobs. Although higher education has several objectives, an important one is to help students learn skills which will enable them to be a productive member of the work force. The National Research Council summarizes, "We have inherited a mathematics curriculum conforming to the past, blind to the future and bound by a tradition of minimal expectation....on average, U.S. students do not master mathematical fundamentals at a level sufficient to sustain our present technologically based society." Students are not learning the mathematical skills society needs. This cannot continue long without industry, education, and thus the nation suffering. If our students are not learning the skills we think they need, then at least some of the blame belongs to us as educators. After years of being feed facts, students will not magical understand how all these facts relate and become adroit at solving problems. We must teach the skills the students need. However the problem is global to the educational system, hence hard to overcome locally in a few university courses. A Role for Computers In order to address the problems we were observing we turned to the computer for help. Computers can play an important role in addressing inadequate preparation, limited motivation, lack of conceptual understanding, and weak problem solving skills among students. Our hypothesis is that computers can aid students learn by: * reducing the cognitive load on a student and * improving students' ability to complete computations. We felt that the computer would particularly aid the weaker students. Little learning occurs when people are cognitively overloaded. The more information students must process to perform a particular task, the less they will comprehend the basic principles underlying the task. Students with a poor grasp of low level computations face enormous obstacles in learning computationally complex topics: They must simultaneously process the new information and recall how to perform the low level operations. If these low level operations can be performed easily and readily by a machine, whether computer, calculator, or abacus, the cognitive load for these students should be reduced. Secondly, machines can aid the students in successfully completing computations. Learning is hindered when students cannot separate their arithmetic errors from those of strategy or approach. When a machine is used appropriately to perform computations, the results of the computations are correct. Students can then focus attention on the far more important task of selecting the appropriate computations for the situation. Student motivation improves when they use computers to perform computations. This allows them to solve realistic problems in a timely fashion, thereby reducing repetitious and boring assignments. Until recently it has not been practical to assign large-sized problems. For example beginning statistics students have not been able to handle data sets containing hundreds of observations and several variables using manual computations. However, we gave beginning statistics students data sets of this size which they analyzed all semester. Although the same ideas and principles could be illustrated with a small data set of ten to twenty observations, the students were highly motivated to analyze their own data set. Once computers are routinely used to perform a computation, the need for students to perform that computation by hand diminishes; in turn, the number of "turn-the-crank" problems assigned can be reduced. Finally, the computer can help students learn concepts. If a picture is worth a thousand words, then dynamic simulation may be worth a thousand pictures. Computers can provide dynamic illustrations and even give students control over simulations, greatly aiding understanding of processes or concepts being learned. Students also can perform computer based experiments to validate concepts and theorems covered in class. This computer use shifts the emphasis from "Are these computations performed correctly?" to "Are these the proper computations to perform for this situation?". In turn this shifts attention from mechanics toward understanding the underlying principals. Reducing the amount of time needed to perform low level operations increases time which can be invested in understanding underlying concepts. The skills needed to successfully accomplish problem solving defy simple description; successful problem solvers use diverse approaches. Nevertheless, solutions usually are based on an understanding of important concepts, the representation of the problem in terms of these concepts, and the execution of steps leading from the representation to the solution. Exercises of progressive difficulty are essential to hone these skills. Computers can help students learn by accomplishing a solution once the problem has been represented in an appropriate manner. Much practice is required to learn to represent problems in useful ways; computer- based instructional materials can be organized to provide many opportunities and paths to practice. The work reported here didn't focus heavily on this use of computers. The Two Courses The two courses involved in this report are an introductory statistics course and a course on analysis of computer algorithms. The results reported here cover one semester. However the software developed locally for the algorithm course was used intermittently for several years during its development. The goal of computer use in each was to: * reduce cognitive load on the students, * allow the students and the instructor to concentrate on concepts, and * allow the student to work realistic problems. We will describe how this was done in each course in a later section. No attempt was made to use the computer as a replacement for the instructor, textbook, or lecture. Macintosh microcomputers were used in this study, because we felt students could master this device in a short time and their price was not prohibitive. The goal was to reduce cognitive load, not increase it. The computer lab used by the students contained twenty-one Macintosh computers with hard drives. Each of the computers could access either of two dot matrix printers via an AppleTalk network. The room housing the equipment was open eighty-five hours per week, but was staffed only twenty-five hours per week, ten hours by an undergraduate teaching assistant and fifteen by faculty. The computer lab was located across campus from the instructors' offices, a walk of ten to fifteen minutes. When problems arose during unstaffed time, students either walked across campus looking for help or waited until office hours. Unfortunately, this made it hard for students to correct errors in use of the software. We also used a Macintosh connected to a projector, both secured on a cart. This Mac- on-a-cart was used in the classroom to demonstrate the use of software and as an instructional tool. Introductory Statistics This course introduces students to the main concepts and tools of statistics at the sophomore/junior level. It covers descriptive statistics, sampling and probability, the binomial and normal distributions, estimation, tests of hypotheses and confidence intervals, inference to one and two populations, followed by a quick survey of simple linear regression and the one-way analysis of variance. The course had no prerequisite for computer knowledge because this study was the first attempt at introducing computers into this course. The mathematics prerequisite is approximately two years of high school algebra, enforced by departmental examination. The instructor, Scott Urquhart, was at San Diego State University on sabbatical leave from New Mexico State University. He has many years of experience using diverse computers for statistical analyses and using computers for helping students learn. He had designed, implemented, and used an interactive mainframe computer package for supporting instruction in a second semester regression course for students who (possibly) knew nothing about computer usage. He also teaches advanced statistical analysis in sophisticated computing environments. In the instructor's past experience, he found that students frequently confused the subject (statistics) with the tool (computers). Given this background, the plan for this experimental course was to concentrate on computer usage for about five weeks, using statistical software as a continuing illustration. Thereafter attention was focused on the statistical issues of sampling and inference. The intent was to introduce computer usage through descriptive statistics, then to encourage students to use the computing resource to explore how the statistical techniques work. It was hoped that this would allow less attention to details of computing and more concentration on concepts. Each student was given his or her own data set consisting of three parts: * a population (of 100 wages), * 20 samples selected out of 200 random samples, each of size ten, which were taken from this population; * the means and variances of the wages for each of the 200 random samples. Each student had a unique population and thus unique samples. Most of the thirteen homework assignments were given in terms of students' own data. The unique data sets meant unique homework solutions, hence students had to do their own work; substantial copying of homework didn't occur as it often does in statistics courses at this level. (The issue was learning, not grades.) Further comparisons of results between students and summaries in class emphasized variations of the sort statistics is intended to cope with. A separate computer program written by the instructor and used here as well as at his home institution produced individualized answer sheets. The software used was DATA DESK, Student Version (1.0), written by Paul and Agelia Velleman. Several features in the Professional Version of this software had been removed to produce a simpler and inexpensive Student Version. Functions of the DATA DESK software have three relatively distinct areas: data entry and manipulation, graphics, and statistical analysis. Its data-oriented functions support data entry in a spreadsheet- like format with many user configurable attributes. Macintosh clipboard actions are supported and are consistent with other Macintosh applications. Further data-oriented features include a data import facility and a substantial facility for performing transformations and/or restructuring of a data set. Variables also can be created from entered patterns or according to random specifications. The graphics-oriented functions of the DATA DESK capitalize on the Macintosh's graphics capabilities. The major features support box plots in several forms, histograms, and scatter plots. These features allow resizing of displays, but not overlaying of several plots on the same axis system nor simple labeling for axes and titles. (The intent is for graphic results to be passed out of the DATA DESK for subsequent editing in a word processor.) The analysis features of the DATA DESK support most elementary parametric statistical analyses; by prior application of a ranking transformation several nonparametric analyses also are supported. Currently available techniques extend through one-way analyses of variance and multiple linear regression. More recent versions of the DATA DESK support more complex statistical analyses. Several features clearly are designed to encourage students to explore statistical testing and confidence intervals. Each student was required to purchase a copy of the DATA DESK. The instructor then built an operational disk containing the operating system, needed parts of the software, and that student's personal data sets. Students in this part of the experiment were in two lecture sections. The first met for fifty minutes, afternoons, three times a week. It began with thirty-one students of whom twenty-eight completed the course. The second section met during evening hours, for seventy-five minutes, twice a week, and started with twelve students of whom nine completed it. These classes contained mainly students typical of a commuter university: undergraduates, working part time, living several miles from campus. About half of the students came from the health sciences and related biological areas; the remainder were nearly evenly divided between the mathematical and social sciences. Although the course was numbered to be at the sophomore level, about two-thirds of each section was past that level. The two sections together included seven students who had completed at least a bachelors degree; one was an MD who had practiced. These were diverse classes, but performance ranged from good to poor in every distinguishable group. However, performance was modestly associated with both major and level: math science students scored about 0.3 grade points higher than the rest of the class (from C+ to B-), and the postgraduate students scored another 0.3 higher (from C+ to B). In-class instruction consisted of presentation of material using a chalk board, demonstrations using the Mac-on-a-cart, planned discussions of the objectives and results of the homework assignments, and spontaneous discussion of relevant questions raised by students. The Mac-on-a-cart was taken to class except for exam review days (3), exams (2) and the final exam. It was used sometime during the class period most days, for as little as five minutes or as much as forty minutes. Some initial information was gathered about the students' knowledge of computing. About one-third of the students had a little experience with a Macintosh; none regarded him- or herself as proficient. About half of the students had experience with another micro, mainly PCs. Only about one-fifth of the students had experience with a system larger than a microcomputer. Nine of the students reported they had no experience with any kind of computer. Grading initially caused understandable anxiety. An unknown instructor was teaching a difficult course with a dramatically different instructional approach. The announced grading plan had points for homework, quizzes, participation, and exams, but no fixed boundaries for specific grades. The instructor assured the classes that the class average grade would be between 2.25 and 2.50 (C = 2.00) for all students who actively participated for the whole semester; the final average was 2.35. Thus, comparison of grades of students in these classes with other classes has no validity. Shortly after mid-semester, an anonymous survey of student attitudes was conducted; a second was conducted near the end of the semester. Some of our observations about attitudes in this course come from these surveys; others come from personal observation and numerous casual conversations with the students. Algorithms Course The computer science course involved in this study was a senior level computer algorithms course. It covers both important computer algorithms and basic principles and techniques of algorithm analysis. Typically such a course covers algorithms concerned with sorting, graphs, string matching, geometry, and some numerical evaluations. The analysis of an algorithm determines the resources, usually time and space, the algorithm uses. Such an analysis uses mathematical techniques and timing studies of the algorithm. The mathematical analysis shows the amount of a resource required by the algorithm as a function of the size of the input. Often this analysis only identifies the order of magnitude of this function. For example an algorithm with input size N may run in a + b*N or fewer time units, where a and b are constants. In many cases constants, such as a and b, may be extremely difficult to determine analytically. Timing tests on machines can give an idea of their values and how the algorithm really performs. The algorithms course is meant to provide future programmers with insights and tools to make informed decisions affecting the speed and space requirements of programs. The algorithms course is difficult for students. Algorithms are a dynamic process which can be difficult to understand. Our students tend to have some of the mathematical deficiencies mentioned earlier. Since the course requires understanding and performing mathematical analysis, any deficiency of mathematical skills hinders students in the course. Because a high level of mathematics can be required to analyze algorithms, students sometimes can neither perform nor understand the analysis of an algorithm. This can be devastating to the students. Finally, students find it difficult to relate the mathematical analysis to performance of algorithms and programs. The software used was MacBalsa, written by Marc Brown, and Algorithms Lab (AL), written locally by Roger Whitney. MacBalsa is an algorithm animation program. This program graphically demonstrates the operation of algorithms. Each algorithm has many different views. Algorithms can be viewed singly or in groups. The user can step though an algorithm or run it as a movie. This program constantly wins kudos from students. They claim it aids understanding of the algorithms far more than lectures or the text. The text for the course, Algorithms by Sedgewick, includes graphics from the workstation version of MacBalsa. Preliminary versions of AL were written for a Vax 780 minicomputer by John Donald and Roger Whitney. The software was moved to the Macintosh using HyperCard as an interface. AL has three tools to aid the study of algorithms: a timer, plotter, and a least squares fitter. Students select all actions and commands by clicking a mouse on a button or menu entry. The timer allows students to measure the elapsed running time of an algorithm for data sets of sizes they choose. The results (size of a data set, associated running time for the algorithm) are displayed in a table. Once an algorithm is timed, the two other tools support investigation of the resulting numerical data. Although students can examine the source code for all the algorithms in AL, they can neither modify existing algorithms nor add their own. Providing such a facility is a major goal for future versions, although earlier versions of AL demonstrated the importance of keeping the programming details of such a process to a minimum. The software in this course was used to: * illustrate algorithms, * verify the mathematical analysis of algorithms, * determine analysis not possible via mathematical means, and * allow students to select proper algorithms for given situations. When an algorithm was covered, MacBalsa was used in class to demonstrate the algorithm. Students were asked to investigate its performance using AL. If the mathematical analysis indicated that algorithm A is faster than algorithm B, the students were asked to verify the analysis by experimentation with AL. This helped students translate factual information given in the text and lecture into personal experience. When the mathematical analysis could not determine which algorithm is faster, students were asked to determine this experimentally. In a conventional algorithms course students implement the algorithms and then time them. Such implementation can take days or weeks to complete, instead of the few minutes required when using AL. Local experience indicates that when students implement a set of algorithms they do not investigate their performance, and fewer algorithms can be examined in a semester. Finally, students were asked to select the best algorithm in several typical sorting situations. They were instructed to support their selections with suitable evidence. The situations were designed to approximate sorting problems confronting working programmers; the answers could not be determined exclusively by either mathematical analysis or timing algorithms. Many students in this course experienced difficulty performing a mathematical analysis of an algorithm, apparently often as a consequence of their deficiencies in mathematics. Given our thoughts on applying computers to learning, why didn't we provide tools to aid such students in performing the mathematical analysis of an algorithm? The task of producing such a system was overwhelming. Any system that we designed to aid students solve a reasonable set of problems was beyond our resources to construct. Two sections of the algorithms course, taught by different instructors, used AL. About seventy students in total took the course. All were computer science majors for whom it is a required course; nearly all were seniors. They had extensive experience with mainframe and minicomputers. Some had experience with several different operating systems. Many of these students had work as computer professionals. However the Macintosh computer was new to most of them. The ability of the students varied dramatically; some take the course several times before passing. Results As stated earlier the computer usage was designed to: * reduce cognitive load on the students, * allow the students and the instructor to concentrate on concepts, and * allow the student to work on realistic problems. We had hoped the computer usage would improve student motivation, increase students' understanding of concepts, and aid the weaker students. We meet with some success, primarily among the better prepared students. Perhaps the most interesting results of the computer usage are: the lack of improvement by the weaker students, the cost of the experiment to the instructors and the department, and the changes that occurred in the courses. The weaker students, primarily those in the statistics course, did not show improvement. On the surface this seems counter to our expectation that a reduction in cognitive load should improve learning: reducing cognitive load should aid the weaker students as much as the strong students. Closer examination reveals the situation to be rather complex. In comparing the statistics students against the computer science students we conclude that the computer imposed a large cognitive load on the weaker students. We feel a computer literacy course would alleviate the problems the weaker students faced with the computer. Our experiment extracted a high cost from the instructors and the department. Equipment had to be obtained, housed and maintained. This required a concerted effort before the experiment was run and requires a continuing effort into the future. The full consequences of these cost are not yet known. The course changed not only in content but also in intent. Clearly some material had to be included into the courses cover the computer usage. The courses did focus more on conceptual understanding of the material and thus became more useful to the students. Below we report how well we achieved our goals, as well as the project's impact on the courses, the instructors, and the department. Motivation Motivation and enthusiasm defy simple measurement. We can only report our observations: In both courses the students' attitude toward the use of the computers was very positive and generally appreciative. The statistics students appreciated being required to learn about microcomputers: Most were pleased with the opportunity to learn about this new and useful tool. One student commented that in the real world he would use the computer to perform statistics, hence using the computer in class was the only reasonable way to learn the subject. Others reported using the software on assignments for other courses. One of our goals was to make the material more relevant to the student. Having students utilize material in another course, because of its usefulness, indicates some success in that goal. As a result of this course several undeclared students decided to become statistics majors. Nevertheless, the students were far from unanimous in their evaluation of the value of computer use in this course. They were asked to consider a future student like themselves and to express their recommendation as to how this course should be taught. The results were: strongly favor traditional (noncomputer), 9; weakly favor traditional, 5; neutral, 6; weakly favor computer-based approach, 9; and strongly favor computer- based approach, 5. A slight plurality (19 to 14) said the course helped them to become more interested in computers. They were nearly equally split (15 to 17) in whether the course helped them become more interested in statistics, but eight of them indicated the course made them plan to take further statistics course(s), and another nine answered with "MAYBE." Clearly the computer-based approach appealed to and motivated some students. The computer science students recognized and appreciated the amount of resources directed at them, both the hardware and the effort to produce the software. This modest effort to improve a course was far more than they normally observe. When surveyed at the end of the course eighty percent of seventy-three students preferred using the software to using standard methods for timing algorithms. Ninety-two percent of the students felt that they learned "a lot" or "some" by timing the algorithms. The remaining eight percent felt that they learned very little or nothing. The sorting assignment (where students selected the best sort for each situation) received high praise in the course survey. Learning The instructor of the statistics course has taught statistics for twenty-four years. His experience with students in this type of course was used to evaluate the performance of these students. The better students made more progress than normal. For these students the computer was a useful aid in learning statistics. Hence, for the better students our hypothesized computer usage was valid. However, the weaker students did not perform as well as weak students in similar courses not using computers. We observed that using the computer was an obstacle for a number of students. These students had a hard time operating the computer and separating the software usage from the statistical concepts. During the course of the semester, various students would finally master the machine and make rapid gains in statistics. The students (nine of forty- three) who reported no computer experience at the beginning of the course divided into two distinct groups, ones who overcame this obstacle (three of the nine), and those for whom the computer remained an obstacle throughout the course. We investigated how these students performed in other courses by examining their entire transcripts at the end of the following semester. We found that student performance in this course was highly correlated with past performance. There were exceptions, some noteworthy but others attributable to causes unrelated to the course. For example one student had just started a new business so she spent little time on the course and performed worse than her past record indicated she should have. Several of the students receiving grades of D or F in this course retook it the following semester; most received almost the same grade the second time, even though it was taught by a different instructor and without computer use. We also discovered a high correlation between computer experience and the strength of the student. The stronger students had higher GPAs and more computer experience. This makes it impossible to judge whether the computer hampered the weaker students, since they were the ones with the least computer experience. A plot of student's GPA at the end of the following semester against rank in the computer-based statistics course displayed several outliers deserving comment. Of the four students who did substantially better than their GPA history, three were among the math science students who declared a concentration in statistics by the end of the semester. Several social science students who struggled with the computer usage faired less well than their GPA history. This contrasts with the effects in the computer science course. As far as we can determine none of the algorithms students were seriously impeded by the computer. The lab helped the students learn about data collection and data analysis, a topic not covered in previous years. The students also performed better at determining which algorithm to use in a given situation. For the first time, students could answer such questions from direct experience, rather than only from a book or lecture as in the past. Both instructors of this course felt that students were better prepared to use the course material in their professional lives. The students did not improve in the their ability to perform mathematical analysis of algorithms. However, this is not surprising as the changes in the course were not directed toward increasing students' ability to perform mathematical analysis. The course changes made comparison of grades in previous classes rather meaningless. In summary we feel the use of computers helped the students learn the important concepts in the courses, with the better students benefitting the most. As seen in the statistics course, the weaker students did not benefit from the use of the computer and probably were the ones who least liked it. Even though the instructor felt they performed worse, their grades were comparable with their past performance. Given the experience with the algorithms course, we feel that if the weak statistics students had prior computer experience they probably would have benefitted more from the computer use. Effect on Course An interesting side effect of the computer usage was a tendency to change the courses. The algorithms course changed in several ways. The first and unexpected way was the need to discuss the collection and analysis of data, since the computer science students had no background in this area. They had to be taught what type of data to collect, what tests to use on the data, and what the results meant. For the first time, we were able to have the students investigate which algorithms would be appropriate in given situations. The course drifted from explaining algorithms and measuring an algorithm's mathematical performance to which algorithm performs better, why it performs better and when should it be used. More time was spent on how to select an algorithm, which is part of the purpose of the analysis. The students responded enthusiastically to this shift; they viewed it as useful training. The changes in the algorithms course resulted from questioning the relevance of a conventional course to students. The longer the project continued, the stronger the questioning became. The field of algorithm analysis is dominated by academic researchers who investigate the complexity of algorithms and write research papers and books from their results. Our students become programmers and system analysts. The researcher and the programmer perform vastly different tasks. It is not clear how much of the classical subject matter is relevant to the programmer. Even the relevant parts usually are presented in manner which disguises their value. This study led to a redefinition of the course for the algorithms instructors. The intent for the statistics course was to make it more conceptual than are similar traditional courses. The plan was for routine computation to be relegated to the computer so students could concentrate on understanding and interpreting results. The stronger students embraced this approach and made substantial progress; the weaker students were much slower to adapt to the computer usage and resisted the move from "How do you do it?" to "What does it mean?" Computer Literacy The two courses clearly differed relative to how successfully computers were used as a learning tool. We feel that the biggest difference between the courses lay in the difference in computer literacy between the students in the two courses. Students in the algorithms course had worked on several computers and several operating systems although almost none of them had prior experience with a Macintosh. By contrast, about twenty percent of the students in the statistics course had experience with a Macintosh, but twenty-five percent had no computer experience and none of them had the diversity of experience typical of students in the algorithms course. Students in the algorithms course adapted to the Macintosh in a week or so, even though they received less than twenty minutes of explicit instruction on its use and about the same amount of instruction on use of the software. By contrast, the statistics course presented about three hours, spread over five weeks, on using a Macintosh and allied matters such as backup and printing. Some aspect of the statistics software was demonstrated almost every class meeting. Features of the software, not the content of statistics, probably received five hours of instruction spread through the course, but concentrated more heavily during the early part of the semester. Some students in the statistics course mastered the Macintosh and the statistics software in a few weeks. These probably were mostly the fourteen percent who strongly favored the computer oriented approach. On the other hand, a number of students remained frustrated with the computers and/or software throughout the course. Many of these students probably were among the twenty-six percent who indicated that computer use was of little value or hindered their learning. Computer literacy appears to be the fundamental difference between students in the two courses which contributed to the difference in the success and acceptance of computers for learning. When students with experience in several environments had trouble, they recognized that fact and tried something different. Students with less experience had substantial difficulty isolating and resolving problems. They were very prone to repeat the same error several times without recognizing the mistake they made, and then blamed the computer for not doing what they wanted. Our experience suggests some relevant topics for a course in computer literacy. Students need to understand the role of an operating system and how to communicate with it, whether it is mouse driven, menu driven, or command driven. A hierarchical file system provides a powerful way to organize files, but it embodies a degree of sophistication which needs to be explicitly understood before it can be used. The Macintosh paradigm of file folders and icons slightly simplifies the use of a hierarchical file system for a user who is familiar with a typical manual filing system. This paradigm breaks down for a user who never has dealt with a complex filing system because it explicitly indicates only two levels; few manual systems have folders within folders within folders. An experienced computer user takes many actions almost by reflex, i.e., without conscious act: starting and ending use of a piece of software; making choices or indicating actions required for the software to continue; naming files; using a network, even if only for printing. Novice users need to be taught such things. More conscious actions include making backups to recover from the inevitable mistakes we, software, and machines make. Ethical matters related to copyright and respect for the privacy of others' work also deserve attention. Effects on Instructors and Department The use of the computers in these courses increased the amount of time required to teach them. The increase in work load had different sources. First, the involved faculty found themselves operating a small computer center. This included raising funds, purchasing equipment, installing software, debugging network problems, hiring and training student lab assistants, deciding lab operation policies and serving as lab proctors. Such work need not be done by the instructor of the courses, but in this case it was. The other source of the increased work came from integrating the software into the course. Course lectures and assignments needed to be changed to incorporate the use of the computers. The department also faces an increase in work load. The first author was the primary force behind the microcomputers project. Funds to purchase, about $150,000 to date, and operate the computer lab were raised by him outside of the department's budget. However, the department must now undertake the operation and maintenance of the lab. This represents an increased work load for an already overworked faculty and staff and increased expenditures in an already meager budget. This equipment was obtained at a time when the department greatly increased its computer equipment holdings without funding and staffing increases. It is not clear how the department's technical staff can maintain their current work load. The department also faces the problem of incorporating the computers into its curriculum. The major effort has been by the two authors, one of whom was a visitor. A few other faculty use the lab in the manner we described. One statistics instructor jumped on the bandwagon and is enthusiastically incorporating computers into his second level courses. Indications are that the statistics curriculum will undergo major changes, partly due to the encouragement of the statistics coordinator and partly due to our initial demonstrations. The changes in the statistics curriculum may be moving more toward national trends than breaking new ground. It is too early to determine if other parts of the department will use the machines as tools for learning. Would We Do It Again? The answer in both cases is a qualified no, but the qualifications are totally different for the two courses. In the algorithms course, the qualifications center around the untenured status of the senior author. In the statistics course, the qualifications focus on some students' lack of computer experience. The statistics instructor would not use computers in the statistics course again under present conditions. The students' problems with computers and software use must be addressed first, not as part of a one- semester introduction to statistics. One solution might be to train the staff sufficiently for the lab. This would require resources that seem far beyond our department or similar ones. Another solution would be a computer literacy course required of all freshman students, perhaps in conjunction with introductory English. As John Wheeler of UC Berkeley indicates: "Students learn by doing in a meaningful context". A computer literacy course needs motivation beyond learn how to use a computer; which is why we feel it should be offered in conjunction with another course. This is our preferred solution; it would allow all subsequent courses, including statistics courses, to concentrate on its subject matter using available tools to enhance learning. The second author is pursuing this approach at his home institution. Summary The mathematical science educational community is facing a challenge in adapting to the demands of our rapidly changing world. We used computers as tools for students in two courses in an attempt to meet this challenge. We found that students improved their grasp of the subject matter, and student motivation increased, although not universally. The courses became far more relevant. We also found the resources needed to successfully integrate the computers into the curriculum were far greater than those available to us. Nevertheless an effective prerequisite course on computer literacy would alleviate many of the difficulties we had in the statistics course. In looking back on the effort, we cannot see how institutions can afford the resources needed to integrate computers successfully into the curriculum. Looking into the future, we cannot see how the universities can survive without integrating computers into the curriculum. Note: Partial Support for this work was provided by the National Science Foundation's College of Science Instrumentation Program through grant #CSI-8750569 and by a donation from Apple Computer, Inc. We would like to thank John Donald for his insightful comments on drafts of this paper. CONTACTS For information about Data Desk write to: Odesta Corporation 4084 Commercial Avenue Northbrook, IL 60062 (312) 498-5615 The student version of Data Desk is available at: Kinko's Academic Courseware Exchange 255 West Stanley Avenue P.O. Box 8000 (800) 292-6640 (in CA) (800) 235-6919 For more information about MacBalsa write to: Marc H. Brown DEC Systems Research Center 130 Lytton Avenue Palo Alto, CA 94301 (415) 853-2152 mhb@src.dec.com (e-mail) For more information about AL write to: Roger Whitney Math/Computer Science San Diego State University San Diego, CA. 92182 (619) 594-6191 whitney@sdsu.edu (e-mail) AL is being prepared for national distribution. REFERENCES National Research Council. Everybody Counts: A Report to the Nation on the Future of Mathematics Education. Washington, D.C.: National Academic Press (1989). John Wheeler, Private communication, July 1990. - - -