Extracting Student Motivation Factors in Education with Contextual Inquiry

Transcription

1 Exracing Suden Moivaion Facors in Educaion wih Conexual Inquiry Waaru Takahashi Graduae School of Informaion Science and Engineering, Risumeikan Universiy, Shiga, Japan Fumiko Harada, Hiromisu Shimakawa, and Takahiro Koyama College of Informaion Science and Engineering Risumeikan Universiy, Shiga, Japan {harada, Absrac In educaion, here are many sudens in a class. Each suden has is own characerisics in moivaion. However, in a class, a single uniform curriculum is applied o all he sudens. I is difficul o each every suden according o heir characerisics. I is necessary o undersand suden moivaion. I needs a grea labor o undersand he every suden moivaion. In his paper, we propose a mehod o undersand he suden moivaion facors. In his mehod, we ge repors from sudens based on he conexual inquiry. The mehod exracs words from he repor. Using he TF/IDF mehod, i calculaes he weigh of every word from he viewpoin of 4 major moivaions. The mehod regisers he se of a word and weigh values ino a dicionary. The mehod calculaes he moivaion facors of he arge suden, summing up he weigh of words appearing in he repor. High ineres for every moivaion facor is examined from 74 suden repors. Even hough here is accuracy difference wih he lengh of he repor, he mehod has classified he repors almos in he same way as manual judgmen. The proposed mehod can reduce he effors of eachers who have o undersand moivaion of various sudens. Index Terms educaion, MSLQ, conexual inquiry, TF/IDF, moivaion I. INTRODUCTION In educaion, sudens have various characerisics, such as a reques, voliion, and capabiliy. Each suden has his own moivaion for learning. Moivaion facors vary wih individual sudens. However, all sudens are augh along one curriculum regardless of suden moivaion facors in many lecures. As a resul, sudens are prevened from heir improvemen of abiliy because all sudens are no moivaed. In order no o decrease he suden moivaion, we should reflec needs and opinions of sudens o eaching conens. We need o device a eaching mehod according o hem. In order o include needs and opinions of sudens o he curriculum, convenional mehods ofen use quesionnaires. However, in quesionnaires, any informaion oher han answers o prepared quesions canno be obained because all Manuscrip received Nov 30, 2012; revised Jan 28, quesions are prepared in advance. For problems ou of educaion assumpion, we canno judge suden needs and opinions of sudens exacly. As oher mehods, educaions make sudens describe heir problems in repors wrien in free exs. However, in his mehod, eachers should read all suden repors, o undersand suden moivaion facors. I makes he load of eachers high. I is necessary o devise a mehod which exracs suden moivaion facors wihou increasing he load of eachers. In his paper, we propose a mehod which exracs suden moivaion facors based on he conexual inquiry [1]. The conexual inquiry is superior o quesionnaires and free-ex descripions, because i looses no conex. The proposed mehod analyzes scenarios obained from he conexual inquiry, o calculae suden ineress for each moivaion facor. We can improve he curriculum using suden moivaion facor exraced wih his mehod. II. DIFFERENCE OF MOTIVATION FACTORS IN EDUCATION A. Imporance of Teaching According o Moivaion Facors Currenly, many researches on moivaion have been done [2]-[4]. Therefore, i is believed ha moivaion is assumed o be an imporan facor in educaion. I needs o undersand moivaion facors of each suden in order o devise an efficien eaching mehod which draws suden moivaion. Curren eaching mehod canno encourage all sudens. I needs o devise a eaching mehod which can encourage many sudens in one curriculum. The realizaion of he mehod needs o examine realize suden characerisics, ha is, we have o know when sudens are moivaed. B. Mpivaion Facors MSLQ [5] defines moivaion facors specifying which suden ges moivaed. Here, four ypical moivaion facors are pu ino consideraion. The inrinsic goal orienaion facor is a moivaion facor which comes from he inside of a suden. I involves is ambiion and 2013 Engineering and Technology Publishing doi: /joace

2 curiosiy. Sudens wih moivaion facor would ry o achieve as much as possible. On he oher hand, he exrinsic goal orienaion facor is an exernal moivaion facor. For example, sudens wan o ge good grade and wan o win fame from oher sudens. The ask value is a moivaion facor where learning asks become he source of moivaion. For example, sudens hink learning is imporan for his fuure works, or sudens an ineresed in asks in a arge field of learning. MSLQ also menions ha sudens should acquire self-managemen skills. The help seeking skills is one of hem. The help seeking skill makes sudens ask for help o oher people when hey encouner difficulies. Owing o his skill, sudens can overcome he difficulies. C. Conexual Inquirys I needs o fully undersand opinions and requess agains lecure sudens have o undersand suden moivaion facors. There is quesionnaire and free descripion as a way o obain informaion from sudens. Bu, in quesionnaire, any informaion oher han answers o prepared quesions oher han canno be obained because of he fixed quesion. In free descripion free opinions can be obained, bu he opinions are rough. In hese mehods, necessary informaion can be obained. The conexual inquiry obains free opinions in deail. The conexual inquiry is a way o undersand poenial problems and needs of users from heir behavior and sysem usage hrough inerviews. In inerviews, a user explains his usual acions. If here is anyhing an inerviewer does no undersand, he inerviewer asks he user why he user ake such acion. Thereby, he inerviewer can undersand a way of hinking inheren in he user. If we can obain he opinions and requess of sudens using he conexual inquiry, we can undersand he suden moivaion facor by he analysis of opinions and requess. In educaion, here are wo imporan poins when obaining informaion from sudens. Firs, i needs o obain many suden daa. Teachers need o analyze he resuls of all suden inerviews. Therefore, he loads of eachers increase. Second, i needs o obain valid informaion o improve lecures. I needs o reduce informaion eachers analyze. In addiion, i is imporan o devise a eaching mehod according o moivaion facors. III. EXTRACTING MOTIVATION FACTOR FROM ANALISYS OF REPORTS A. Overview of he Mehod This paper shows a mehod o deermine suden moivaion facors and o exrac sudens who have high ineres in a specific moivaion facor from analysis of repors. This mehod helps a eacher know which sudens have high ineres in a specific moivaion facor. I reduces he load of he eacher. Fig. 1 shows he ouline of he mehod. Firs, sudens inerview each oher according o he conexual inquiry. Inerviewers submi resuls as repors. The mehod looks up words sudens use in he repors wih he morphological analysis. I calculaed he TF/IDF [6] imporance of every word for each of moivaion facors. Every combinaion of a word and is imporance is regisered ino a dicionary. Figure 1 Ouline of he mehod To analyze new suden repors, we prepare a dicionary from pas repors. When we ge a new repor, we exrac every word enrolled in he dicionary. A he same ime, we sum up he imporance of he word for every moivaion facor. For a specific moivaion facor, we make a hisogram whose verical axis and horizonal axis represen he number of sudens and he accumulaed imporance for he specific moivaion facor. We regard sudens locaed in he low level in he hisogram do no have ineress in he moivaion facor, while sudens in he high level have enough ineress. We can undersand suden moivaion facor wih he analysis. B. Collecing of Repors To obain poenial problems and needs from user behavior and sysem usage, he mehod les sudens inerview wih each oher, afer lecures on he basis of he conexual inquiry. Each inerviewer submis he repor which is he resul of he inerview. We can obain many suden daa afer we have he sudens inerview each oher. I is hand for sudens o ge programming abiliy. Many sudens do no undersand significance of programming. Therefore, suden who menions disconen and grumbling increases if his inerview is carried ou during he programming course. Emoion of disconen and grumbling prevens requiremens for programming from appearing in inerviews. Therefore, suden moivaion facor is no exraced well. On he oher hand, sudens who have passed lo of ime since hey finished he programming lecure may have forgoen requiremens. We decide he arge of inerviews o sudens who have jus finished he programming lecure o carry ou he inerview raional. C. Weigh of Word by TF/IDF Sudens have some moivaion facors. I is considered ha suden who has srong ineress in a specific moivaion facor frequenly use words which concern he moivaion facor. In his sudy, many repors which sudens show ineress in a specific moivaion facor are 209

3 gahered. Words are exraced from hese repors. Every moivaion facor is associaed wih words using he weigh. If sudens have ineres in only a specific moivaion facor and oher sudens have no ineres in he moivaion facor, he suden frequenly use words which relae o he moivaion facor. In addiion, he words are no used oher suden repors. Therefore, we use TF/IDF o deermine he word weigh. The value of TF/IDF is calculaed wih produc of he word frequency of occurrence and he degree he word appears only in he specific documen. If oal number of words is N and he word w occurs n imes in he documen, he frequency of he word occurrence f is represened by he formula 1. f n / N Suppose he oal number of repors which are obained is R and he word w occur in repors. When he degree idf o which he emergence word w in a specific repor is calculaed by he formula 2. idf log 2 ( R / r) If word w occurs in some repor, idf becomes large. For word w, he value of TF/IDF is represened by he formula 3. f f idf Therefore f becomes small when word w occurs in a repor and repor where w rarely occurs. The value of TF/IDF each word has is defined as he weigh agains he moivaion facor. Le he weigh of he word for a moivaion facor be f. We can obain n pairs of moivaion facor. i ( w, f, f,..., f 1 2 n This pair represens weigh of all moivaion facors agains each word. This pair is made for each word. A se of he pair is referred o as a dicionary which is used in his mehod. If he weigh of word for a moivaion facor is large, sudens who use his word have large ineres in he moivaion facor. D. Exracing Ineres in Moivaion Facor If we can judge a suden who has srong ineres in a specific moivaion facor, a eacher can focus on repors he suden wrie when he eacher devises his eaching mehod relaed o he moivaion facor. In his secion, we explain a way o exrac sudens who have srong ineres in each moivaion facor. Fig. 2 shows he flow o exrac he degree of he ineres on each moivaion facor. For each word appearing in a suden repors, he weigh regisered in he dicionary is summed up for each moivaion facor. The repor of various lenghs may come ou. Long repor has many words, which increases he oal value. Finally, he oal value is divided by he number of words which is used in a repor when we ) (1) (2) (3) calculae he oal value, so ha we should evaluae he fair degree of an ineres regardless of he repor lengh. Namely, he degree of an ineres is he average weigh of he emerging word regisered o he dicionary. If he oal number of words in he repor is n, he degree of an ineres o a moivaion facor is calculaed wih he following formula f i / n. The degree of an ineres in each moivaion facor is scored by he above procedure. Nex, we make hisogram for each moivaion facor o exrac upper sudens on hisogram. I is conceivable ha hese sudens frequenly use words which relae o he moivaion facor in he repor. Therefore, we can judge he suden have ineres o he moivaion facor. Figure 2 Exracing moivaion facor The performance of he sysem is affeced by a way o exrac upper suden. Generally, many sudens have ineres o some moivaion facors. I is no good ha a eacher devises his eaching mehod only o refer o he repor. Therefore, i needs o exclude sudens who have a low degree of he ineres. In addiion, if here are sudens who have srong ineress only in a specific moivaion facor, he average is large. For his reason, we should be avoided o deermine a hreshold wih he average. Considering he above, in his sudy, we decide he hreshold as he quarile poin of he foureen. The sudens of op 75 percen are regarded as ones having srong ineress in he moivaion facor, while hose of he boom 25 percen are regarded as ones having no ineress. IV. EXAMPLE OF EXTRAQCTING DEGREE OF INTEREST TO MOTIVATION FACTOR A. Obain Repors by Inerview We examine o exrac suden moivaion facors on he programming exercise by his mehod. An experimen has been conduced for 74 sudens. The sudens are augh o conexual inquiry in advance. The average number of words in he repor in his experimen in 574 words, he variance of he number of words in he repor is 163,472. The variey of lengh of repors in conceived from his dispersion value. In he experimen, sudens who have finished he programming exercise are imposed a repor o sae, when hey ge moivaed in programming. We oblige he sudens o make a pair, o 210

4 inerview wih each oher. Each suden submis a resul of he inerview as a repor. B. Making a Dicionary by Analyzing Repors Suden repors are divided a a raio of 7:3. The former is used for analysis, while he laer is used for creaing a dicionary. Suden repors facor for creaing a dicionary are analyzed based on MSLQ o exrac ineress in moivaion facors. The repors are grouped according o each moivaion facor. In his experimen, if suden repor includes a descripion relaed o a moivaion facor which defined in MSLQ, i is judged ha he suden has he moivaion facor. Thereby, four groups are made. A repos can belong some groups. Nex, in each group, we exrac words from repors. The value of TF/IDF is calculaed for each exraced word. The value of TF/IDF is regarded as he weigh of he word in he each moivaion facor. Pairs of a word and is weigh consiue a dicionary. In his experimen, 1881 words are exraced o a dicionary. C. Exracing Suden Moivaion Facor by Using Dicionary Using he dicionary, we have calculaed he degree of suden ineres in each moivaion facor. On he oher hand, readers judge he suden moivaion facor conens of repors manually. We have compared he resul of deermining o quanify he degree of suden ineres from he weigh of a word for he moivaion facor wih he resul of judging presence or absence of ineres o moivaion facor by readers of he repor. The degree of he ineres o he four moivaion facors are calculaed from he analysis of repors using he dicionary. In analysis, words are exraced from suden repors. The weigh of each moivaion facor of words which are regisered in dicionary is summed up. Nex, he sum is divided by he number of imes. The calculaed value is regarded as he degree of he ineres o each moivaion facor. We collec he degree of he ineres o a suden moivaion facor and make a hisogram. We compare he resul of exracing op of 75 percen sudens on hisogram wih he resul of classified sudens obained he reading of suden repors. We examine how much percenage of sudens who are classified by he reader is included in he op of 75 percen on he hisogram. V. EVALUATIONS A. Experimen Resul We evaluae four moivaion facors. The inrinsic goal orienaion one, he exrinsic goal orienaion one, he ask value one, he help seeking one, which are defined in MSLQ. Fig. 3 shows a hisogram which obained in experimen for he inrinsic goal orienaion one. The horizonal axis is he degree of he ineres o he moivaion facor and he verical axis is he number of people. Hisogram shows he disribuion of he degree of he suden ineres o each moivaion facor. Sudens judged as hey have srong ineres o he inrinsic goal orienaion moivaion facor by readers of he repor are eigh. Seven of hem locae in he op of 75 percen sudens on he hisogram. The hisogram shows ha mos of sudens are locaed in he range from 1.4 o 1.6. They have srong ineres in he exrinsic goal orienaion one by readers of he repor are seven. Five of hem locae in he op of 75 percen sudens on hisogram. For his moivaion facor, mos of sudens are locaed in he range from 1.6 o 1.7 in he hisogram. On he oher hand, sudens who do no locae in he op of 75 percen sudens are locaed from 0.9 o 1.0 in he hisogram. Sudens judged as hey have srong ineres in he ask value one by readers of he repor are seven. Five of hem locae in he op of 75 percen sudens on he hisogram. Two of hese nine sudens have srong ineres. Bu, oher seven sudens disribue near he hreshold which is deermined in his mehod. Sudens judged as hey have srong ineres in he help seeking one by readers of he repor are welve. Eleven sudens of hese welve sudens are judged as hey have srong ineres in his mehod. We invesigae wheher sudens judged as hey have srong ineres in each moivaion facor by readers of he repors conain op of 75 percen sudens. We quanify he percenage of sudens classified by readers is included in he op of 75 percen on he hisogram. Table I shows he resul. We obained recall of more han 70 percen for moivaion facor oher han he ask value. I is possible o exrac sudens who have srong ineres o moivaion facor for he inrinsic goal orienaion, he exrinsic goal orienaion, and help seeking moivaion facor. Therefore, i is possible o reduce suden repors which a eacher should read by selecing repors of sudens who have srong ineres in each moivaion facor. As a resul, we can reduce load of eacher. TABLE I. RECALL OF EACH MOTIVATION FACTOR inrinsic exrinsic ask value help seeking Recall 87.50% 71.43% 55.56% 91.70% Figure 3 Degree of ineres in inrinsic B. Discussion We can exrac sudens who have srong ineres in each moivaion facor in his mehod. Bu, we canno judge wheher he ineres in posiive or negaive. Therefore eacher should read suden repors and judge posiive and negaive of degree of ineres. However, we undersand ha his mehod limis he repors which he eacher should read when he judges posiive and negaive 211

5 of suden ineres o moivaion facor. From his i is possible o suppor ha eacher chooses repors which include valid informaion o improve lecure. If suden repor is shor or long exremely, he judgmen by readers of repors does no mach he resul of his mehod. In his mehod, he degree of he ineres in a moivaion facor is calculaed as he average of he weigh per occurrence of he word. Exremely long repors have many words which do no relae o he moivaion facor. The average is low because of using hese words. Sudens who wrie long repor is judged so as no o have ineres o moivaion facor which he have ineres essenially. This problem is solved by eliminaing word which do no relae o moivaion facor. Exremely shor repors do no have a meaningful descripion. I is impossible o calculae he degree of he ineres o a moivaion facor. As anoher problem, we could no exrac exacly suden moivaion facor if a suden use words differen from ones used in repors o make a dicionary. This problem is solved if we make he dicionary from many repors. REFERENCES [1] H. Beyer and K. Holzbla, Conexual Design, Morgan Kaufmann, [2] K. M. Y. Law, V. C. S. Lee, and Y. T. Yu, Learning moivaion in e-learning faciliaed compuer programming courses, Compuers & Educaion, vol. 55, no. 1, pp. 218, [3] M. J. Lee and A. J. Ko, Personifying programming ool feedback improves novice programmers' learning, in Proc. Sevenh Inernaional Workshop on Compuing Educaion Research, Providence, Rhode Island, USA, 2010, pp [4] S. W. Marins, A. J. Mendes, and A. D. Figueiredo, Diversifying aciviies o improve suden performance in programming courses, in Proc. 11h Inernaional Conference on Compuer Sysems and Technologies and Workshop for PhD Sudens in Compuing on Inernaional Conference on Compuer Sysems and Technologies, Sofia, Bulgaria, 2010, pp [5] P. R. Pinrich, D. A. F. Smih, T. Garcia, and W. J. McKeachie, A manual for he use of he Moivaed Sraegies for Learning Quesionnaire (MSLQ), Technical Repor No. 91-B-004, The Universiy of Michigan, [6] J. W. Han and M. Kamber, Daa Mining, Morgan Kaufmann, Waaru Takahashi received B.E from Risumeikan Universiy in He advanced Graduae School of Risumeikan Universiy. He engages in he research on daa engineering. He is member of IPSJ. Takahiro Koyama received B.E from Risumeikan Universiy in He advanced Graduae School of Risumeikan Universiy. He engages in he research on daa engineering. He is member of IPSJ. Fumiko Harada received B.E. and M.E, and Ph.D degrees from Osaka Universiy in 2003, 2004, and 2007, respecively. She joined Risumeikan Universiy as an assisan professor in Risumeikan Universiy in 2007, and is currenly a lecurer. She engages in he research on real-ime sysems and daa engineering. She is a member of IEEE. Hiromisu Shimakawa received Ph.D degree from Kyoo Univ. in He joined Risumeikan Univ. in Currenly, he is a professor in Risumeikan Univ. His research ineress include daa engineering, usabiliy, and inegraion of psychology wih IT. He is a member of IEEE and ACM. 212