- Email: [email protected]
Online supervision of CBL through LANs
243
Online Supervision of CBL through LANs Bertrand I B R A H I M and Christiane K U H N I University of Geneva, 24, rue du gdndral Dufour, 1211 Genbve 4, Switzerland
Microcomputers are now powerful enough to be widely used as professional tools. The computer science department of the University of Geneva (Switzerland) uses a number of them in a Computer Based Learning (CBL) project, because they allow timing control and graphics. The microcomputers are connected through a Local Area Network (LAN), not only allowing the sharing of a file server but also providing queueing and synchronization primitives. A L A N allows the online collection of information about the learning process. This information can be used to generate statistics about the students and about the CBL programs. These statistics can be used for automatic curriculum maintenance and for designers' feedback. Online collection of information can also be useful for the teacher to watch the learning process while it is taking place, without disturbing the students by watching over their shoulders.
Keywords: CAL; CBL; CAI; LAN; Network; Supervision; Teaching; Instruction; Learning; Aid; Microcomputer.
Bertrand Ibrahim graduated from the University of Geneva in 1977 with a diploma in computer science. He spent the following 5 years doing teaching and research in interactive computer graphics ( P h . D . in 1982) until he turned to computer based learning. Now a research associate, he is active in a number of related fields (videodisks, network, multilinguism) in close cooperation with UC Irvine.
Christiane Kiihni graduated from the University of Geneva in 1983 with a licence in computer science, and in 1985 with a diploma in computer science. She spent two years working on Computer Based Learning.
North-Holland Education & Computing 1 (1985) 243-247
Introduction The first Computer Based Learning projects were developed m a n y years ago on big mainframes, with time-sharing operating systems. Developers noticed that CBL material was more f / O bound than CPU bound, since it had much more to do with writing text on the screen and reading user's answer than making computations. Full time availability and low cost of microcomputers make them interesting targets for CBL. So, when a Computer Based Learning project started at the University of Geneva, the main reasons that guided the choice for microcomputers were: 1) availability of graphic capabilities and 2) full control of tiining. We consider it essential to be able to control the output pacing and even more the input timing to deal with situations such as those in which the student does not know what to do and thus needs some help to continue the learning process. Microcomputers were first used as stand-alone machines intented for environments like homes and public libraries. The only ways of getting some feedback about the CBL material were to let the learners fill out a questionnaire and express their feelings, or to let the program store some information on the floppy disk about what had been done during the learning session. The latter has been mainly used to record ratings for quizzes. The CBL techniques we use conform to those of Professor Alfred Bork [1,2], whose team we have been collaborating with for a few years. The scenarios of the lessons are prepared by teachers or specialists in the field being taught, then a team of coders implement these scenarios as Pascal programs. We use the U C S D p-system(TM) environment to develop programs for m a n y reasons developed in a paper recently presented at
WCCE'85 [31. We have developed a set of CBL modules to teach business students how to use spread sheet programs [4]. The first modules have been tested with a group of students and the feedback we got was through a questionnaire filled in by the students just .after they used the modules. The questions (prepared by the teacher) were mainly re-
0067-9287/86/$3.50 © 1986, Elsevier Science Publishers B.V. (North-Holland)
244
B. Ibrahim, (2. Kuhni / Online Supervision of CBL through LANs
lated to the student's reaction to this new teaching medium. As interesting as it may be, such a questionnaire does not allow us to determine how to improve the modules, nor can it show us objectively how the students react to them. The most important questions of the questionnaire, with some statistics follows: - How much time did you spend on this experience? (138 rain.) - How far have you been through the three lessons? (2.81 lessons) - Did you do the final exercise? (yes, 63%) - Where do you think Spread Sheets can a p p l y most usefully? - Other learning methods are possible. What are the pros and cons of this one? - Do you feel you have learned something about computer science in general? -About your apprenticeship, what are the sentences that correspond at best to your opinion? Sentence
Goes too fast Goes too slow Adequate pacing D o n ' t have enough time to think before next activity starts Somehow simplistic Sequencing hard to follow Order of presentation is logical Is user-friendly Is not user-friendly Have been sometimes lost Received good help when lost Feel free to do anything G o o d to have nobody watching over your shoulder Too verbose Explanations not easy to understand If this is Computer Aided Learning, I am disappointed Too many things on screen, don't know where to look Problems with keyboard This kind of lesson is not bearable longer than 15 minutes This kind of lesson is not bearable longer than 30 minutes This kind of lesson is not bearable longer than I hour Can do errors without feeling guilty
%
(6.3) (12.5) (62.5) (25) (18.8) (12.5) (50) (62.5) (12.5) (68.8) (25) (0) (25) (18.8) (18.8) (6.3) (0) (18.8) (6.3) (6.3) (31.3) (37.5)
If more lessons were available, would you like to try them? Yes
No
Using
(100)
(0)
a LAN
As we have stated above, the technique we use involves working with a whole team of programmers. Maintaining and developing software that will be shared with other people and used by a lot of students raises problems like those concerned with using the last version, getting the updates, making sure old versions are no longer used, etc. These are mainly software engineering problems that, by themselves, might already justify using a LAN. However, the most important reason we found for using a network was the need for an unbiased means of evaluating our software. Indeed, developers as well as teachers need some kind of feedback because of the very interactive nature of the material being developed (openended questions, non-linearity in the course of the lesson, etc.).
Getting Feedback
There are two kinds of feedback one might want to have: - Online supervision by the teacher; - Automatic collection of statistics for later analysis. Both of these can use the same information that could be generated by the modules in use. The kinds of information one might need to have concern: 1) The path: what are the different tasks the student has done, in what chronological order has he done them, and what are the most usual paths followed? 2) The timing: how long did it take the student to answer the various questions, did a specific answer often time out? 3) The difficulties: what are the proportions of good answers to different questions; what kinds of wrong answers were given; did we get any answers at all? 4) Unused sequences: were certain sections of the modules hever reached, or seldom used?
B. Ibrahim, C. Kuhni / Online Supervision of CBL through LA Ns
Online Supervision
The idea of supervising the learning process online came from the comparison with language laboratories where teachers can hear what the student is saying, judge how well he is doing and eventually intervene if necessary to give him some help. However, if it is easy in our case to look at what the student is doing, it is much more difficult to let the teacher intervene through the network. Direct intervention would imply deeper modification of the CBL modules. Where should the teacher's message appear on the screen? And for how long? Should the student be able to reply? These questions are only a part of those that need to be answered before going further in this direction. Having the same underlying software to handle the screen layout, the keyboard input and the timing for all the CBL modules would allow an implementation that does not need to be done again for every new module.
Basic Tools
For both online supervision and offline statistics the same underlying tools are used. The first tool needed was provided with the network software. It is a queuing mechanism that allows using a hard disk partition as a big F I F O accessible from any machine on the net. A mutual exclusion mechanism is also provided to ensure safe access by all machines to the same FIFO. The other tools are: - packages allowing the CBL modules to put informations (called " m a r k s " ) in the F I F O - processing facilities to read these informations ("marks") assynchronously. The structure of these " m a r k s " is two-fold: a student identification and an activity descriptor. The student identification is determined once and for all at the beginning of the session and is then used to "prefix" all the marks sent to the FIFO. The activity descriptor is a variable length string indicating in a compact format what section of the CBL module is being executed. It indicates the subject being taught, the lesson, the task (a distinction is made if the task is started for the first time in the session or if it has already been done before), the question asked and whether the answer was correct (if any) or wrong, with the possibility
245
of categorizing the type of wrong answer. Collecting this information implies that already existing CBL material has to be modified to generate them. In fact, if the programs have been correctly structured to reflect the structure of the scenario, it is easy to add the adequate procedure calls at the correct place. For those modules that have not yet been coded, including the " m a r k " generation in the code is negligible compared to the rest of the coding effort. As a matter of fact, a great deal of the effort associated with " m a r k " generation occurs in the design process, to decide where and what kind of information has to be generated.
Higher Level Tools
The overall process can be illustrated (Fig. 1) with an S A D T (Structured Analysis and Design Technique [5]) data diagram. Each rectangular box represents a set of data involved in the process, and the arrows in between represent the transformation processes. We have already seen before that the CBL programs (Box 1) will produce " m a r k s " (Box 3) during execution. The content of the " m a r k s " will also depend on student input (Box 2). On the other hand, before any execution of the CBL modules, the coders will have to build a sort of database (Boxes 5 & 6) that reflects the structure of the modules (Box 4). Generally, the coders will already have a document specifying how the subject is divided into lessons, the lessons into tasks and what the steps are that comprise the different tasks. During the coding phase, an interactive program is used to capture this structure and create two kinds of files. The first one is a dictionary that contains a human-readable string indicating each part of the overall structure, allowing later production of readable statistics. The other kind of file is called "information files". There is one information file for each lesson on a given subject. Each information file describes the skeleton of a lesson and is composed of a three-level tree Structure where each node contains a counter. The root indicates how m a n y times the lesson has been started; the first level nodes indicate how many times each task of this lesson has been started; the second level nodes indicate how m a n y times each question has been asked and the third level nodes
246
B, lbrahim, C. Kuhni / Online Supervision of CBL through LANs
indicate how m a n y times each kind of answer has been given. All these files have to be created before the CBL material is used by learners. At that stage, all the counters in the information files are set to zero. It is only after the CBL modules have been used that the " m a r k s " are processed and the counters incremented accordingly. In this implementation, the dictionary as well as the information files are anonymous in the sense that the student identification is not used nor stored. This is because our first motivation was to get feedback that would help us review the design. The infrastructure we have built allows us, however, to do much more. It is possible to maintain statistics about the students: what lessons did they finish successfully; where did they last stop; how did they answer to specific questions. It is also possible to watch what is going on while it is happening. Regardless of the online intervention (that can be implemented later), the online supervision can be implemented in at least two ways: - by developing a primitive to examine non-destructively the F I F O from its end; - b y sending the " m a r k s " simultaneously to another F I F O that would be cleared, for instance, every minute. The teacher would then have only to run a program on one of the machines connected to the network to know who is working and what they are doing. This program can show the information in a human-readable format by using the same dictionary a n d information files as the statistics program.
Problems There are some solvable aspects that can be considered as problems related to these new facilities. None of them raise real difficulties but they are nonetheless enumerated here to make the reader aware of them. 1. After some deep design review there are chances to change the structure of the lessons. It is important to adapt the dictionary and the information files accordingly. Should all the counters used for the statistics then be reset? 2. The CBL modules are often run for demonstrations. It is important that these demos do not alter the statistics. This is easy to take into account: if the student's name is D E M O no " m a r k " is generated. 3. In case of heavy use (many hundreds of students), we might have to run the scrutinizing program at least once a day. Failure to do this might result in data loss or system hang up. 4. Another difficulty is related to our multilingual environment. Since the student can choose the language in which he will run the CBL modules, should we mix the statistics related to different languages, given the fact that some difficulties might come from the translation itself? All these questions are related more to design issues than to implementation. We are not far enough in our experimentation to have found answers to them yet.
Coding programCBL~L./Execute the CBLprogramsandproduce"marks"
-
1 Selecq User ~ / 1 ' subject I 'nput 2 ' ~
Script design
Marks L 31 / ~ Extractthe "marks"fromthe FIFO ( ~ Analyse ~. I ~ ~
]
.
ofStructUremodu[e 4
Dicti°nnarYdF Create information file and dictionnary
Fig. 1. SADT model for flow of data in online supervision.
conc,usion designab°ut
B. lbrahim, C. Kuhni / Online Supervision of CBL through LANs
Implementation In the present state of the project, all the basic tools (i.e. "piping" of marks and reading them) are fully coded and already used in some CAL material. Also implemented is the utility used to create, update and display the database containing the description of the CAL modules and the statistics about their usage. The implementation of the basic tools and the statistics utility corresponds to an approximate coding effort of six man-months. For the moment, we have not implemented the online supervision utility since we do not know if teachers will be interested in using it. However, all the components needed are available and we estimate the coding effort to be one or two manmonths.
Conclusions The primitives we have developed are obviously very useful to the designers in helping them review the design of the lessons. The utility of online
247
supervision is less obvious. One of the main advantages of CBL is that the student feels free to go at his own pace and is not disturbed by anyone looking over his shoulder. In this context, how would the students react to online supervision? On the other hand, online supervision could alleviate some of the frustrations encountered in dealing with an entirely automated process. Human help could be requested or alarms could light up if the monitoring showed that the student was experiencing real difficulties.
References [1] A. Bork, Learning with Computers. Digital Press, 1981. [2] A. Bork, Personnal Computers for Education, Harper & Row, Publishers, 1985. [3] S.D. Franklin, B.C. Levrat, "Portability of Computer Based Learning Materials Across Machines and Across Natural Languages", WCCE 85 Proceedings, p. 291-295. [4] B. Ibrahim et al., "Interactive Modules to Teach Spread Sheet Analysis", WCCE 85 Proceedings, p. 181-186. [5] D.T. Ross, K.E. Sehoman Jr., "Structured Analysis for Requirements Definition", IEEE Transaction on Software Engineering, Janua~ 1977.
Online Supervision of CBL through LANs Bertrand I B R A H I M and Christiane K U H N I University of Geneva, 24, rue du gdndral Dufour, 1211 Genbve 4, Switzerland
Microcomputers are now powerful enough to be widely used as professional tools. The computer science department of the University of Geneva (Switzerland) uses a number of them in a Computer Based Learning (CBL) project, because they allow timing control and graphics. The microcomputers are connected through a Local Area Network (LAN), not only allowing the sharing of a file server but also providing queueing and synchronization primitives. A L A N allows the online collection of information about the learning process. This information can be used to generate statistics about the students and about the CBL programs. These statistics can be used for automatic curriculum maintenance and for designers' feedback. Online collection of information can also be useful for the teacher to watch the learning process while it is taking place, without disturbing the students by watching over their shoulders.
Keywords: CAL; CBL; CAI; LAN; Network; Supervision; Teaching; Instruction; Learning; Aid; Microcomputer.
Bertrand Ibrahim graduated from the University of Geneva in 1977 with a diploma in computer science. He spent the following 5 years doing teaching and research in interactive computer graphics ( P h . D . in 1982) until he turned to computer based learning. Now a research associate, he is active in a number of related fields (videodisks, network, multilinguism) in close cooperation with UC Irvine.
Christiane Kiihni graduated from the University of Geneva in 1983 with a licence in computer science, and in 1985 with a diploma in computer science. She spent two years working on Computer Based Learning.
North-Holland Education & Computing 1 (1985) 243-247
Introduction The first Computer Based Learning projects were developed m a n y years ago on big mainframes, with time-sharing operating systems. Developers noticed that CBL material was more f / O bound than CPU bound, since it had much more to do with writing text on the screen and reading user's answer than making computations. Full time availability and low cost of microcomputers make them interesting targets for CBL. So, when a Computer Based Learning project started at the University of Geneva, the main reasons that guided the choice for microcomputers were: 1) availability of graphic capabilities and 2) full control of tiining. We consider it essential to be able to control the output pacing and even more the input timing to deal with situations such as those in which the student does not know what to do and thus needs some help to continue the learning process. Microcomputers were first used as stand-alone machines intented for environments like homes and public libraries. The only ways of getting some feedback about the CBL material were to let the learners fill out a questionnaire and express their feelings, or to let the program store some information on the floppy disk about what had been done during the learning session. The latter has been mainly used to record ratings for quizzes. The CBL techniques we use conform to those of Professor Alfred Bork [1,2], whose team we have been collaborating with for a few years. The scenarios of the lessons are prepared by teachers or specialists in the field being taught, then a team of coders implement these scenarios as Pascal programs. We use the U C S D p-system(TM) environment to develop programs for m a n y reasons developed in a paper recently presented at
WCCE'85 [31. We have developed a set of CBL modules to teach business students how to use spread sheet programs [4]. The first modules have been tested with a group of students and the feedback we got was through a questionnaire filled in by the students just .after they used the modules. The questions (prepared by the teacher) were mainly re-
0067-9287/86/$3.50 © 1986, Elsevier Science Publishers B.V. (North-Holland)
244
B. Ibrahim, (2. Kuhni / Online Supervision of CBL through LANs
lated to the student's reaction to this new teaching medium. As interesting as it may be, such a questionnaire does not allow us to determine how to improve the modules, nor can it show us objectively how the students react to them. The most important questions of the questionnaire, with some statistics follows: - How much time did you spend on this experience? (138 rain.) - How far have you been through the three lessons? (2.81 lessons) - Did you do the final exercise? (yes, 63%) - Where do you think Spread Sheets can a p p l y most usefully? - Other learning methods are possible. What are the pros and cons of this one? - Do you feel you have learned something about computer science in general? -About your apprenticeship, what are the sentences that correspond at best to your opinion? Sentence
Goes too fast Goes too slow Adequate pacing D o n ' t have enough time to think before next activity starts Somehow simplistic Sequencing hard to follow Order of presentation is logical Is user-friendly Is not user-friendly Have been sometimes lost Received good help when lost Feel free to do anything G o o d to have nobody watching over your shoulder Too verbose Explanations not easy to understand If this is Computer Aided Learning, I am disappointed Too many things on screen, don't know where to look Problems with keyboard This kind of lesson is not bearable longer than 15 minutes This kind of lesson is not bearable longer than 30 minutes This kind of lesson is not bearable longer than I hour Can do errors without feeling guilty
%
(6.3) (12.5) (62.5) (25) (18.8) (12.5) (50) (62.5) (12.5) (68.8) (25) (0) (25) (18.8) (18.8) (6.3) (0) (18.8) (6.3) (6.3) (31.3) (37.5)
If more lessons were available, would you like to try them? Yes
No
Using
(100)
(0)
a LAN
As we have stated above, the technique we use involves working with a whole team of programmers. Maintaining and developing software that will be shared with other people and used by a lot of students raises problems like those concerned with using the last version, getting the updates, making sure old versions are no longer used, etc. These are mainly software engineering problems that, by themselves, might already justify using a LAN. However, the most important reason we found for using a network was the need for an unbiased means of evaluating our software. Indeed, developers as well as teachers need some kind of feedback because of the very interactive nature of the material being developed (openended questions, non-linearity in the course of the lesson, etc.).
Getting Feedback
There are two kinds of feedback one might want to have: - Online supervision by the teacher; - Automatic collection of statistics for later analysis. Both of these can use the same information that could be generated by the modules in use. The kinds of information one might need to have concern: 1) The path: what are the different tasks the student has done, in what chronological order has he done them, and what are the most usual paths followed? 2) The timing: how long did it take the student to answer the various questions, did a specific answer often time out? 3) The difficulties: what are the proportions of good answers to different questions; what kinds of wrong answers were given; did we get any answers at all? 4) Unused sequences: were certain sections of the modules hever reached, or seldom used?
B. Ibrahim, C. Kuhni / Online Supervision of CBL through LA Ns
Online Supervision
The idea of supervising the learning process online came from the comparison with language laboratories where teachers can hear what the student is saying, judge how well he is doing and eventually intervene if necessary to give him some help. However, if it is easy in our case to look at what the student is doing, it is much more difficult to let the teacher intervene through the network. Direct intervention would imply deeper modification of the CBL modules. Where should the teacher's message appear on the screen? And for how long? Should the student be able to reply? These questions are only a part of those that need to be answered before going further in this direction. Having the same underlying software to handle the screen layout, the keyboard input and the timing for all the CBL modules would allow an implementation that does not need to be done again for every new module.
Basic Tools
For both online supervision and offline statistics the same underlying tools are used. The first tool needed was provided with the network software. It is a queuing mechanism that allows using a hard disk partition as a big F I F O accessible from any machine on the net. A mutual exclusion mechanism is also provided to ensure safe access by all machines to the same FIFO. The other tools are: - packages allowing the CBL modules to put informations (called " m a r k s " ) in the F I F O - processing facilities to read these informations ("marks") assynchronously. The structure of these " m a r k s " is two-fold: a student identification and an activity descriptor. The student identification is determined once and for all at the beginning of the session and is then used to "prefix" all the marks sent to the FIFO. The activity descriptor is a variable length string indicating in a compact format what section of the CBL module is being executed. It indicates the subject being taught, the lesson, the task (a distinction is made if the task is started for the first time in the session or if it has already been done before), the question asked and whether the answer was correct (if any) or wrong, with the possibility
245
of categorizing the type of wrong answer. Collecting this information implies that already existing CBL material has to be modified to generate them. In fact, if the programs have been correctly structured to reflect the structure of the scenario, it is easy to add the adequate procedure calls at the correct place. For those modules that have not yet been coded, including the " m a r k " generation in the code is negligible compared to the rest of the coding effort. As a matter of fact, a great deal of the effort associated with " m a r k " generation occurs in the design process, to decide where and what kind of information has to be generated.
Higher Level Tools
The overall process can be illustrated (Fig. 1) with an S A D T (Structured Analysis and Design Technique [5]) data diagram. Each rectangular box represents a set of data involved in the process, and the arrows in between represent the transformation processes. We have already seen before that the CBL programs (Box 1) will produce " m a r k s " (Box 3) during execution. The content of the " m a r k s " will also depend on student input (Box 2). On the other hand, before any execution of the CBL modules, the coders will have to build a sort of database (Boxes 5 & 6) that reflects the structure of the modules (Box 4). Generally, the coders will already have a document specifying how the subject is divided into lessons, the lessons into tasks and what the steps are that comprise the different tasks. During the coding phase, an interactive program is used to capture this structure and create two kinds of files. The first one is a dictionary that contains a human-readable string indicating each part of the overall structure, allowing later production of readable statistics. The other kind of file is called "information files". There is one information file for each lesson on a given subject. Each information file describes the skeleton of a lesson and is composed of a three-level tree Structure where each node contains a counter. The root indicates how m a n y times the lesson has been started; the first level nodes indicate how many times each task of this lesson has been started; the second level nodes indicate how m a n y times each question has been asked and the third level nodes
246
B, lbrahim, C. Kuhni / Online Supervision of CBL through LANs
indicate how m a n y times each kind of answer has been given. All these files have to be created before the CBL material is used by learners. At that stage, all the counters in the information files are set to zero. It is only after the CBL modules have been used that the " m a r k s " are processed and the counters incremented accordingly. In this implementation, the dictionary as well as the information files are anonymous in the sense that the student identification is not used nor stored. This is because our first motivation was to get feedback that would help us review the design. The infrastructure we have built allows us, however, to do much more. It is possible to maintain statistics about the students: what lessons did they finish successfully; where did they last stop; how did they answer to specific questions. It is also possible to watch what is going on while it is happening. Regardless of the online intervention (that can be implemented later), the online supervision can be implemented in at least two ways: - by developing a primitive to examine non-destructively the F I F O from its end; - b y sending the " m a r k s " simultaneously to another F I F O that would be cleared, for instance, every minute. The teacher would then have only to run a program on one of the machines connected to the network to know who is working and what they are doing. This program can show the information in a human-readable format by using the same dictionary a n d information files as the statistics program.
Problems There are some solvable aspects that can be considered as problems related to these new facilities. None of them raise real difficulties but they are nonetheless enumerated here to make the reader aware of them. 1. After some deep design review there are chances to change the structure of the lessons. It is important to adapt the dictionary and the information files accordingly. Should all the counters used for the statistics then be reset? 2. The CBL modules are often run for demonstrations. It is important that these demos do not alter the statistics. This is easy to take into account: if the student's name is D E M O no " m a r k " is generated. 3. In case of heavy use (many hundreds of students), we might have to run the scrutinizing program at least once a day. Failure to do this might result in data loss or system hang up. 4. Another difficulty is related to our multilingual environment. Since the student can choose the language in which he will run the CBL modules, should we mix the statistics related to different languages, given the fact that some difficulties might come from the translation itself? All these questions are related more to design issues than to implementation. We are not far enough in our experimentation to have found answers to them yet.
Coding programCBL~L./Execute the CBLprogramsandproduce"marks"
-
1 Selecq User ~ / 1 ' subject I 'nput 2 ' ~
Script design
Marks L 31 / ~ Extractthe "marks"fromthe FIFO ( ~ Analyse ~. I ~ ~
]
.
ofStructUremodu[e 4
Dicti°nnarYdF Create information file and dictionnary
Fig. 1. SADT model for flow of data in online supervision.
conc,usion designab°ut
B. lbrahim, C. Kuhni / Online Supervision of CBL through LANs
Implementation In the present state of the project, all the basic tools (i.e. "piping" of marks and reading them) are fully coded and already used in some CAL material. Also implemented is the utility used to create, update and display the database containing the description of the CAL modules and the statistics about their usage. The implementation of the basic tools and the statistics utility corresponds to an approximate coding effort of six man-months. For the moment, we have not implemented the online supervision utility since we do not know if teachers will be interested in using it. However, all the components needed are available and we estimate the coding effort to be one or two manmonths.
Conclusions The primitives we have developed are obviously very useful to the designers in helping them review the design of the lessons. The utility of online
247
supervision is less obvious. One of the main advantages of CBL is that the student feels free to go at his own pace and is not disturbed by anyone looking over his shoulder. In this context, how would the students react to online supervision? On the other hand, online supervision could alleviate some of the frustrations encountered in dealing with an entirely automated process. Human help could be requested or alarms could light up if the monitoring showed that the student was experiencing real difficulties.
References [1] A. Bork, Learning with Computers. Digital Press, 1981. [2] A. Bork, Personnal Computers for Education, Harper & Row, Publishers, 1985. [3] S.D. Franklin, B.C. Levrat, "Portability of Computer Based Learning Materials Across Machines and Across Natural Languages", WCCE 85 Proceedings, p. 291-295. [4] B. Ibrahim et al., "Interactive Modules to Teach Spread Sheet Analysis", WCCE 85 Proceedings, p. 181-186. [5] D.T. Ross, K.E. Sehoman Jr., "Structured Analysis for Requirements Definition", IEEE Transaction on Software Engineering, Janua~ 1977.