IEEE TRANSACTIONS ON EDUCATION, VOL. 12, NO. 1, MARCH 1969

Computer-Aidee Control of Learning

HE~LENE BESTOUGEFF, JEAN-PIERRE FARGETTE, AND ROMAIN JACOUD

Abstract-This paper presents the first computer-aided experi-ment on learning control which took place in Europe at the Facultedes Sciences de Paris. The teaching objectives, the level and theneeds of the interaction (dialogue, open answers, feedback to thestudent) are described in the first part of this paper. Then the com-puter configuration is given and the technical solut'ions decided arediscussed.

The third part of the paper deals with the man-machine com-munication problem to which has been given great importance in thedefinition of the experiment. Some examples showing either the waythe course is described and put into the system or the dialogue be-tween the computer and a student are included in the paper.

IT WILL BE argued in this paper that the improve-ments in a student's knowledge ought to beassessed more frequently than is usual nowadays.

An annual exam session and some tests during the yeardo not supply enough information for effective controlof the learning process. This is particularly important inthe universities where the large number of students im-pede communication between teachers and students.We shall deal in turn with the following points:

Teaching objectives, technical solutions decided so far,and the problem of man-machine communications.

TEACHING OBJECTIVESThe evolvement of an educational package implies a

strict definition of the teaching objectives and dependson the importance given to experimentation within theteaching program.

Level of InteractionThe teacher-student interaction normally includes

lectures, tutorial sessions, and laboratory assignments.The experiment described below fits in with these tradi-tional contacts and belongs to the second stage of thelearning process, which occurs after lectures have beenattended and a first contact with the subject has beenmade by the students.

Aims of the InteractionThe aims of the experiment can be seen from two

angles: that of the teacher and that of the student.The experiment ought to enable the student to assess

the extent of his knowledge by analyzing the items whichare essential to his grasp of the course. He ought to begiven a chance to combine the different elements of thesubject matter into a rational abstract of his own.

Manuscript received July 26, 1968.The authors are with the Facuilte des Sciences de Paris, Paris,

France.

The experiment ought to enable the teacher to detectshortcomings in student knowledge which would eitherallow him to hint at or to give complementary informa-tion to the students by reviewing some essential pointsin his next lecture.

The Context of the InteractionIn the conditions previously defined, the experiment

should be conducted by a question-answer system,permitting a dialogue with the student and a degree ofadaptability to the student's progress. To reap the fullbenefit of such a system, the dialogue must allow openquestions and answers and virtually eliminate the "tickthe right answer" process, so that the student will nothave the impression of being able to choose betweenseveral acceptable solutions.An open answer enables the student to structure his

thoughts in a logical way without being influenced bysuggestions for a choice. The system must also be adapt-able to the student's current level of knowledge, hisreactions should be recorded and fed back to the ques-tioning program. It may be a general feedback appli-cable to all students involving a complete or partialreview of the question asked. This, however, is notenough; the questioning program must be capable ofadjusting itself to the particular needs and problems ofeach user. The set of tested points must, therefore, beeasily altered.The course of the program should satisfy all these

conditions. It will consist of a series of alternative setsof questions out of which a selection will be madeaccording to the needs of the student (see Fig. 1). As theprogram proceeds, new questions are brought forwardfollowing a logical sequence. A "reply file" is associatedwith each question and contains 1) the right answer,and 2) some examples of the right and wrong answers.This reply file must be capable of alteration after ananalysis of unforeseen answers which will in some casesbe included in it.

TECHNICAL SOLUTION DECIDED So FARConsidering the objectives described above and the

large number of students involved, this assignmentcould only be carried out by a computer. However, it isimpossible to run such an experiment if the hypothesisexpressed and the algorithms of the processing are notfirst checked on a small scale. Thus it seems that arather small computer is sufficient in that we do notdemand a very short response time.

4

BESTOUGEFF et al.: COMPUTER-AIDED CONTROL OF LEARNING

Fig. 1. Structure of a set of questions.

The Computer Configuration (Fig. 2)

The central processor is an IBM 360/30 with a 16 Kbyte memory (4 K words). Attached to it we have one

disk driver (7 million bytes) on a selector channel. On a

multiplexor channel besides a card reader and a high-speed printer there is a buffer of 1252 characters con-

nected to 20 terminals (1052) through a control unit, eachterminal having its own part of this buffer (59 char-acters). The communications are always initiated by

the computer. But, as the same buffer is used for inputas for output, a higher priority is given to the informa-tion coming from the terminal.To write the system, we have used the "basic operat-

ing system" of IBM with many modifications to fit theproblem.

How the Course Is Described and Put into the System

The first point to deal with is an easy means to enableteachers to make their sets of questions ready for theauxiliary storage of the computer. Some difficultiesarise from the fact that any part of the text might bemodified and that a lot of special characters have beenchosen for the terminals; this guided us to punch cardsfrom the terminals themselves. This method makes eachset of questions independent, easy to modify and to

reproduce, and at the same time provides a clear list ofthe whole texts. The use of punched cards means that ifany accident occurs on a disk pack the questions can beeasily restarted. It also avoids the writing of a program

which would allow direct modifications on the disks.The questions are recorded in two phases. They are

first punched from the terminals and then written on

disks from punched cards (previously verified and modi-fied if necessary). For the first operation, a very simplelanguage describing questions and answers has beendeveloped. The text of a question appears preceded by

its identificator and followed by the identificator of theassociated answer (reply file):

Q12 " text " *R12

The reply file is the gathering of any number of com-binations formed by an answer, the associated commentand one or two question identificators:

R12 answer, comment *Q13

answer, comment *Q03-Qo5

answer, comment *Q19

comment *Q20

So, to any foreseen answer (correct or not) a commentcorresponds which is sent by the computer to the stu-dent before the next question designed by its identifi-cator. In the above example, the case of an unforeseenanswer brings us to the last line where the last commenton the list and the question number 20 appear.

The Dialogue with the StudentsLet us describe the permanent exchange of informa-

tion between the computer and the twenty students.The student sees the questions automatically typed out.He then answers using the keyboard to compose hismessage and send it to the computer with a special key.However, the organization of all processing informationappears more clearly when we consider the operationsfrom the computer's side.The input-output has been programmed in the inter-

rogative form so that we can start either an addressingsequence or a polling sequence. A "polling" brings intothe core memory every message sent by the studentssince the last polling, and sets the corresponding buffersfree. An "addressing" sends to the buffers a group ofmessages previously prepared in the core memory; theyare automatically typed out at the rate of 15 charactersper second. The size of the buffer allocated to eachterminal limits to 59 characters the length of the mes-sage transmitted to one printer. Therefore, long textshave to be cut into several messages. The control of thelinkage between them is assumed by means of a "stu-dent's array" stored in the core memory and updated atevery step of the dialogue. From a general point of viewthe output area of core memory is filled again as mes-sages are addressed to the terminals. A polling sequenceoccurs between one addressing and the next one (atleast to indicate that some terminals are free).The programs are written in assembly language IBM

360. The 4 K word core memory has been divided. A 3 Kword area contains resident programs; the exchangingof all other programs is performed in the 1 K word arealeft in the core memory, where they are processed (forinstance, answer's analysis, keyword program, compilerfor the desk calculator and so on . . . ).

5

IEEE TRANSACTIONS ON EDUCATION, MARCH 1969

20TERM

NALS

Fig. 2. Computer configurations.

VUE D'ENSEMBLE DES PRINCIPALES NOTIONS MATHEMATIQUESNOUS ALLONS REVOIR LES ELEMENTS FONDAMENTAUX DU CALCUL VECTORIEL, LES PRINCIPAUX OPERATEURS DIFFERENTIELSAINSI QUE QUELQUES INTEGRALES VECTORIELLES USUELLES LA SEQUENCE NORMALE COMPORTE 13 QUESTIONS-1-DANS UN REPERE CARTESIEN ORTHONORME OX1X2X3, ON DONNE TROIS VECTEURS AM(, 2, 1), B(1, 2, 1), C(2, 1, 0)

CALCULER LE PRODUIT SCALAIRE a=A-C-a =+2BIEN -

2-CALCULER LA VALEUR DE LA COMPOSANTE Di SELON L'AXE OX1 DU VECTEUR D=A!XBD1=l

INCORRECT-ON APPELLE Ao,A2,As LES COMPOSANTES DE A, ET Bi,B2,Bs LES COMPOSANTES DE B DONNER

L'EXPRESSION ANALYTIQUE DE LA COMPOSANTE D1 SELON L'AXE OX1 DU PRODUIT VECTORIEL D =A XBD= -B2A8+A2BaEXACT-2-CALCULER LA VALEUR DE LA COMPOSANTE D1 SELON L'AXE OX1 DU VECTEUR D=A XB

D1 =0

BIEN-

4-L'ETUDE DE L'ELECTRON LIBRE EST UN CHAPITRE FONDAMENTAL DE LA PHYSIQUE QUI PERMET LAPPROCHEDES SYSTEMES REELS (PHYSIQUE DU SOLIDE). C'EST POURQUOI SON MAGNETON PORTE UN NOM =LE MAGNETON DE BOHR

DONNER LA VALEUR NUMERIQUE DU MAGNETON DE BOHR go DANS LE SYSTEME M-K-S-ALE RESULTAT SERA DONNE SOUS LA FORME,o=0, .PUIS(-..) SANS ETRE SUIVI D'UNITESpo=0, 53 PUIS(-23)

INCORRECT.VOICI LES VALEURS NUMERIQUES PERMETTANT D'EFFECTUER LE CALCUL

e=1, 60 PUIS(-19)M=9, 10 PUIS(-31)H=6, 62 PUIS(-34)

DONNER LA VALEUR NUMERIQUE DE ,uo*CAL LIR(e, M, H)' po=e*H/(4*r*M)' ECR(,o)'TAPEZ 3 DONNEE(S) SVP

1-6 E-19, 9.1 E-31, 6 62 E-34

gto= 0262467 E-23CALCUL TERMINE

go=0, 93 PUIS(-23)BIEN -

Fig. 3.

THE PROBLEM OF MAN-MACHINE COMMUNICATIONS of routines have been written which first check theApart from technical problems, it is very important

to create a natural way of communication between thecomputer and the students.

In view of the fact that the students were alreadyacquainted with the subject matter, the multiple-choice response was considered not suitable as theycould answer by guessing, and so the constructed freeresponse was chosen. As the system is applied tophysics, two types of response are allowed: a formula,and a sentence. For the first case, we have defined a

general syntax for the different formulas found in thecourse and expressed them in normal backus form. A set

syntax of the formula answered, and give a diagnosis ifit is not correct, then put it in a canonic form repre-sented by a succession of terms with their signs. Thesecond member of the relation is equal to zero.The expected answer has been processed in the same

way before its recording and so can be compared to thestudent's response. For example, all these answers havethe same meaning for the computer.

rot H - i = + E49tE

rot H =e(i/E+ OtE)(1/E) (rot H - i) =tE

6

IEEE TRANSACTIONS ON EDUCATION, VOL. 12, NO. 1, MARCH 1969

To analyze the second type of response (sentence) it wasconsidered that a key-word technique could be used withsuccess as far as only a few answers of this type areexpected. To improve this basic technique, apart fromthe key words themselves, the negative forms aresearched and processed. For example, if we expect asentence such as "le champ est nul" with champ and nulas key words a response as "champ non nul" would beconsidered as wrong.

It appears that some questions involve numericalcalculations, and so it becomes necessary to provide thestudent with a kind of desk calculator which allows himto calculate any expression by writing a simple sequenceof arithmetical statements. (see Fig. 3).

All the possibilities described so far are orientated forthe students, but the important thing in this kind ofexperiment is of course the feedback. So, through thewhole dialogue, we must keep a record of the path thestudent has taken and write down his results (time ofresponse, number of correct answers, unexpected replies,etc ... ). The automatic sorting of all these data and the

editing of the results should allow the teachers to im-prove the questions and to better follow the generalunderstanding of the subject matter they teach.

CONCLUSIONThe system described above is in its experimental

stage at the moment, but many groups of students areusing it regularly.

In spite of a very limited core memory (4 K words)and a considerable number of terminals (20), the waitingtime for a new question on the average does not exceedten seconds (which is just about the maximum accept-able from such a system). Of course, it is too early tohave any results other than technical ones on the extentof the experiment.

It seems that this kind of use of computers in uni-versities has been welcomed by the students. This reac-tion can be considered all the more positive since inFrance a test of knowledge and a systematic control arenot readily admitted.

Distributed Electrical Systems: A New

Introduction to Electromagnetics

FRANKLIN D. MOORE, MEMBER, IEEE

Abstract-A new introductory sequence in engineering electro-magnetics is described. Its novel features include an axiomatic de-velopment of electromagnetic theory, the application of transformanalysis to the solution of the important partial differential equa-tions, and the use of engineering terminology in developing a dis-tributed systems viewpoint. The sequence establishes a close rela-tionship among electromagnetics, circuits, and system theory, andhas been used successfully at the University of Iowa for the pastthree years.

INTRODUCTION

r nHE presentation of engineering electromagneticsT has undergone no general development since the

subject was first put into the undergraduate cur-riculum after World War II. Despite the imaginativeand increasingly sophisticated books in solid-state elec-tronics, circuit and svstem theory, and materials sciencewhich have characterized the general progress in elec-trical engineering education, the introductory fieldsbooks of the sixties (with minor exceptions) are in-distinguishable from their predecessors of the forties.I would like to describe a significantly different type ofpresentation which is under development at the Uni-versity of Iowa. To keep the discussion within reasonablebounds, I will concentrate on the novel features of theprogram. First let us establish the context within whichwe will operate.

Manuscript received July 25, 1968.The author is with the Department of Electrical Engineering,

the University of Iowa, Iowa City, Iowa 52240.

THE AXIOMATIC APPROACH TOLUMPED SYSTEM THEORY

Three primary features characterize the presentation:an axiomatic, deductive approach to electrical science,transform calculus solution of the fundamental equa-tions, and engineering concepts and terminologythroughout. The deductive approach proceeds in thefollowing way.

Step 1: State the basic laws of electrical science.Step 2: Use these laws to derive the equation govern-

ing the physical system of interest.Step 3: Solve the equation.Step 4: Interpret the solution in physical terms.

Steps 1 and 3 are primarily mathematical, while Steps2 and 4 are primarily engineering.To establish a consistent set of terminology, and to

show how the distributed system concept helps unifythe student's understanding of circuits and fields, let usbriefly summarize the approach used in introductorycircuit and system theory. We postulate the existence ofideal resistors, capacitors, and inductors, after definingpotential and current (as best we can), and assert thatthey are completely characterized by their terminalequations. The conservation of energy and charge implythe Kirchhoff equations,

3V = 0 j ii = o,

7