CUSTOMIZED INSTRUCTIONAL PEDAGOGY IN LEARNING PROGRAMMING PROPOSED MODEL

Similar documents
Difference in Characteristics of Self-Directed Learning Readiness in Students Participating in Learning Communities

Going Below the Surface Level of a System This lesson plan is an overview of possible uses of the

APPLYING THE MIXED RASCH MODEL TO THE FRACTION CONCEPT OF PUPILS

National Assessment in Sweden. A multi-dimensional (ad)venture

Implementation of a planar coil of wires as a sinusgalvanometer. Analysis of the coil magnetic field

Optimize Neural Network Controller Design Using Genetic Algorithm

PRELIMINARY STUDY ON DISPLACEMENT-BASED DESIGN FOR SEISMIC RETROFIT OF EXISTING BUILDINGS USING TUNED MASS DAMPER

AN ANALYSIS OF TELEPHONE MESSAGES: MINIMIZING UNPRODUCTIVE REPLAY TIME

Form. Tick the boxes below to indicate your change(s) of circumstance and complete the relevant sections of this form

REGRESSION ASSOCIATION VS. PREDICTION

Blind Estimation of Block Interleaver Parameters using Statistical Characteristics

DISCUSSION ON THE TIMEFRAME FOR THE ACHIEVEMENT OF PE14.

Volume 3, No.2, March - April 2014 International Journal of Advanced Trends in Computer Science and Engineering

EXPERIMENT 4 DETERMINATION OF ACCELERATION DUE TO GRAVITY AND NEWTON S SECOND LAW

African Journal of Education and Practice ISSN (Online) Vol.2, Issue 2 No.1, pp 1-15, 2017

TWO REFERENCE japollo LUNAR PARKING - ORBITS / T. P. TIMER. (NASA CR OR rmx OR AD NUMBER) OCTOBER 1965 GODDARD SPACE FLIGHT CENTER

AGE DETERMINATION FROM RADIOLOGICAL STUDY OF EPIPHYSIAL APPEARANCE AND FUSION AROUND ELBOW JOINT *Dr. S.S. Bhise, **Dr. S. D.

EXPERIMENTAL DRYING OF TOBACCO LEAVES

Reliability Demonstration Test Plan

Probability, Genetics, and Games

MATH 1300: Finite Mathematics EXAM 1 15 February 2017

Impact of literacy status on Participation of Tribal Women in Panchayati Raj A case study of Nilgiri ITDA Block of Balasore district in Odisha.

FEM Analysis of Welded Spherical Joints Stiffness Fan WANG a, Qin-Kai CHEN b, Qun WANG b, Ke-Wei ZHU b, Xing WANG a

Chapter 12 Student Lecture Notes 12-1

Evaluation of Accuracy of U.S. DOT Rail-Highway Grade Crossing Accident Prediction Models

Reliability of fovea palatinea in determining the posterior palatal seal

Percentage of Non-Caucasians in Clinical Trials from 2000 to By Meghanadh Yerram E Special Project

Evaluation Of Logistic Regression In Classification Of Drug Data In Kwara State

Combined use of calcipotriol solution (SOp.g/ ml) and Polytar liquid in scalp psoriasis.

MUDRA PHYSICAL SCIENCES

Machine Learning Approach to Identifying the Dataset Threshold for the Performance Estimators in Supervised Learning

IBM Research Report. A Method of Calculating the Cost of Reducing the Risk Exposure of Non-compliant Process Instances

Design of a Low Noise Amplifier in 0.18µm SiGe BiCMOS Technology

YOUR VIEWS ABOUT YOUR HIGH BLOOD PRESSURE

Adaptive Load Balancing: A Study in Multi-Agent. Learning. Abstract

AMIA 2009 Symposium Proceedings Page - 109

COURSES IN FOREIGN LANGUAGES for ERASMUS INCOMING STUDENTS. at Sofia University FACULTY OF CLASSICAL AND MODERN PHILOLOGY

Design and simulation of the microstrip antenna for 2.4 GHz HM remote control system Deng Qun 1,a,Zhang Weiqiang 2,b,Jiang Jintao 3,c

Rearview Video System as Countermeasure for Trucks Backing Crashes

Statistical Techniques For Comparing ACT-R Models of Cognitive Performance

Measuring Cache and TLB Performance and Their Effect on Benchmark Run Times

PHA Exam 1. Spring 2013

A e C l /C d. S j X e. Z i

Input Techniques for Neural Networks in Stock Market Prediction Ensembles

CURRICULUM, ASSESSMENT AND REPORTING ARRANGEMENTS: A SUMMARY OF THE CHANGES FROM SEPTEMBER 2015

Research into the effect of the treatment of the carpal tunnel syndrome with the Phystrac traction device

A Comment on Variance Decomposition and Nesting Effects in Two- and Three-Level Designs

A Geographical Location Based Satellite Selection Scheme for a Novel Constellation Composed of Quasi-Geostationary Satellites

Car Taxes and CO 2 emissions in EU. Summary. Introduction. Author: Jørgen Jordal-Jørgensen, COWI

Identifying the Most Effective Model for Understanding the Growth Rate of Government e-transactions: Brown's Model of Exponential Smoothing

ON-LINE MONITORING AND FAULT DETECTION

Programme-Specific Examination and Study Regulations for the Consecutive Master Degree Programme

CARAT An Operational Approach to Risk Assessment Definitions, Processes, and Studies

Cattle Finishing Net Returns in 2017 A Bit Different from a Year Ago Michael Langemeier, Associate Director, Center for Commercial Agriculture

Catriona Crossan Health Economics Research Group (HERG), Brunel University

A Study of Factors Contributing to Low Retention Rates

Towards User-Adaptive Information Visualization

e/m apparatus (two similar, but non-identical ones, from different manufacturers; we call them A and B ) meter stick black cloth

AN ANALYSIS OF TECHNICAL ENGLISH COMMUNICATION SKILLS COURSE FOR ENGINEERING STUDENTS IN KONERU LAKSHMAIAH UNIVERSITY

Company registration number: ROI FRS 105 Demo Client UNAUDITED FINANCIAL STATEMENTS for the year ended 31 January 2018

Approximate Dimension Equalization in Vector-based Information Retrieval

A Practical System for Measuring Film Thickness. Means of Laser Interference with Laminar-Like Laser

The Monitoring of Machining Inherent

RULES REDUCTION AND OPTIMIZATION OF FUZZY LOGIC MEMBERSHIP FUNCTIONS FOR INDUCTION MOTOR SPEED CONTROLLER

Or-Light Efficiency and Tolerance New-generation intense and pulsed light system

THE CROSS-FLOW DRAG ON A MANOEUVRING SHIP. J. P. HOOFf. MARIN, Wageningen, The Netherlands

Presentation to the Senate Committee on Health & Human Services June 16, The University of Texas Health Science Center at Houston (UTHealth)

THEORY OF ACOUSTIC EMISSION FOR MICRO-CRACKS APPEARED UNDER THE SURFACE LAYER MACHINING BY COMPRESSED ABRASIVE

YGES Weekly Lesson Plan Template. Name: Kindergarten- ELA Date: Dec. 7-11, Monday Tuesday Wednesday Thursday Friday

Dr She Lok, Dr David Greenberg, Barbara Gill, Andrew Murphy, Dr Linda McNamara

ENCRYPTING OPTIMISATION TECHNIQUES WITH PARTIAL AUTHENTICATION

Magnetic Field Exposure Assessment of Lineman Brain Model during Live Line Maintenance

International Conference on Global Trends in Engineering, Technology and Management (ICGTETM-2016)

Industrial Power Demand Response Analysis for One-Part Real-Time Pricing

A multiple mediator model: Power analysis based on Monte Carlo simulation

How Asset Maintenance Strategy Selection Affects Defect Elimination, Failure Prevention and Equipment Reliability

A Robust R-peak Detection Algorithm using Wavelet Packets

Tests on a Single Phase Transformer

Company registration number: ROI FRS 105 Demo Client UNAUDITED FINANCIAL STATEMENTS for the year ended 31 December 2017

An Empirical Analysis of Software Productivity

Bionic Design for Column of Gantry Machining Center Using Honeycomb Structures

Using the Aggregate Demand-Aggregate Supply Model to Identify Structural. Demand-Side and Supply-Side Shocks: Results Using a Bivariate VAR

THE MOZART EFFECT: AN INVESTIGATION INTO CONTRIBUTIONS OF THE LEFT AND RIGHT HEMISPHERE EMILY BAUMAN AND JEHAN COUTU

Identification of Dangerous Highway Locations: Results of Community Health Department Study in Quebec

Automated Rust Defect Recognition Method Based on Color and Texture Feature

Artificial Neural Network to the Control of the Parameters of the Heat Treatment Process of Casting

The Preference and Environmental Perception of Building Contour Control of Mountain-landscape City

A Numerical Analysis of the Effect of Sampling of Alternatives in Discrete Choice Models

Eugene Charniak and Eugene Santos Jr. Department of Computer Science Brown University Providence RI and

COLORADO ENGLISH LANGUAGE AQUISITION PROGRAM. CELApro Final. General Research Tape Format. Technology CTB/McGraw-Hill Monterey, CA 93940

IV Congresso RIBIE, Brasilia 1998

How to Combine Expert (or Novice) Advice when Actions Impact the Environment?

The Diaphyseal nutrient foramina architecture - a study on the human upper and lower limb long bones

The Strengths and Limitations of the Statistical Modeling of Complex Social Phenomenon: Focusing on SEM, Path Analysis, or Multiple Regression Models

L4-L7 network services in shared network test plan

Research on reforming the teaching of satellite communications based on STK

Breeding new sugarcane clones by mixed models under genotype by environmental interaction

Emerging Subsea Networks

Computation and Analysis of Propellant and Levitation Forces of a Maglev System Using FEM Coupled to External Circuit Model

Improving the Surgical Ward Round.

Transcription:

CUSTOIZED INSTRUCTIONAL PEDAGOGY IN LEARNING PROGRAING PROPOSED ODEL 1 UHAED YOUSOOF, OHD SAPIYAN 1 Dhofar Univrsity, Dpartmnt of IS, Salalah, OAN GUST, Dpartmnt of Computr Scinc, Kuwait, KUWAIT E-mail: 1 m_yousoof@du.du.om, sapiyan@gust.du.kw ABSTRACT Computr programming is a highly cognitiv skill. It rquirs mastry of many domains. But in rality many larnrs ar not abl to cop with th mntal dmands rquird in larning programming. Thus it lads to rot larning and mmorization. Thr ar many rasons for this situation. Howvr on of th main rasons is th natur of th novic larnrs who xprinc high cognitiv load whil larning programming. Givn th fact that th novic larnrs lack wll dfind schma and th limitation of th working mmory, th studnts could not assimilat th knowldg rquird for larning. It is to b notd that som larning support in th form of visualization may hlp in larning programming, as tachrs ar always rmindd that us of visual aids could nhanc larning in studnts. Th ffct of visualization in larning is not clarly tangibl. This papr addrss this issu by mploying NASA TLX rating scal to masur th cognitiv load in larning programming using visualizations. Th masurmnt of cognitiv load could hlp to undrstand th difficulty lvl of th larnrs. Th larnrs vary on anothr in trms of thir larning styl and capabilitis and hnc th load xprincd during larning programming may diffr significantly from on anothr in a sam homognous group. This papr will propos a modl to optimiz th instruction to larnrs basd on thir background profil and will mploy nural ntwork to optimiz th instruction by suggsting th bst visualization tool for ach larnr. Kywords: Programming, Visualization, Cognitiv Load, NASA TLX scal, Nural Ntwork 1. INTRODUCTION As discussd in th abstract this papr addrsss to rsolv th cognitiv load in larning programming. Th study also aims at concluding th ffctivnss of th visualizations in rducing th cognitiv load. Thr ar two mthods to masur th load namly physiological masurs and non physiological masurs. W dcidd to masur th cognitiv load using non-physiological masurs. Non-physiological masurs ar mostly basd on th rating scal. Ths masurs ar dvlopd using th foundations of psychology whrby th chancs of strotyping and biasnss ar liminatd. Th study is carrid out in such a way that it includs th factors that impacts th larning procss of novic programmrs. Th study is basd on th Cognitiv Load Thory (CLT) which involvs th two aspcts of mmory namly Long Trm mory (LT) and Working mory (W). Th factors considrd for th LT includd th grads scurd in Pr Univrsity athmatics, atriculation athmatics, prior computr knowldg, atriculation English scor, Pr Univrsity English scor and th grads of th IELTS typ xamination. Th Cognitiv load is masurd using th rating scal NASA TLX. Th studnts hav ratd th difficulty on th six diffrnt dimnsions of th difficultis facd. Thy also gav wight-ag for th difficulty. Th cognitiv load is calculatd using th standard procdur of cognitiv load calculation for NASA TLX scal which is discussd in th latr sctions. Th subjcts ar wll organizd in trms of thir prior knowldg background which rlats to LT and also th cognitiv load is masurd by th NASA TLX rating scal which corrsponds to th W.. SAPLE DISTRIBUTION Th xprimnt was carrid out with 40 studnts in Chnnai, India. Ths studnts wr in th first yar of Bachlor of Computr scinc studying introductory programming. Th studnt s profil comprisd of various s dmographical aras such as 309

Rural, Urban and Smi Urban. Th sampls wr qually distributd by gndr. Th first stp in conducting th xprimnt was to collct th basic information of th sampls. Th basic data includd th gndr of candidat, ara of origin namly Rural, Urban and Smi Urban. In addition to th abov basic data, th English and athmatics knowldg masurs wr also collctd. Th scors scurd by th sampls in th High school and Pr Univrsity courss for English and athmatics was collctd. Th sampls wr givn a comprhnsiv English languag proficincy tst to masur thir rading, listning, spaking and writing skills. Th English tst is similar to IELTS tst. It is obsrvd in th slctd sampl that girls mak up 60% of th sampl. Th sampls ar also profild qually basd on th dmographics of th studnts, som hail from th city as wll as thos who ar from th rural ara who stay in th collg hostl. Th studnts ar classifid according to th dmographic locality of thir schools prior to ntring th trtiary ducation as urban, smi-urban and rural. Th distribution is as follows,37.5% from rural,35% from smi urban and 7.5 % from urban Th studnts wr givn a choic to choos th programming languag to larn for th xprimnts. Thr wr thr groups basd on th choic namly th first group to larn, th scond group to larn and yt anothr group with to larn as thr ar som studnts who had som knowldg of in Pr Univrsity cours. So th third group optd to larn. Th distribution of th sampl was that 15 Studnts took cours about 37.5%. Anothr 14 studnts about 35% optd for th languag. Th rmaining 11 studnts narly 7.5% optd to study, as thy had studid th computr programming languag in thir Pr Univrsity lvl. English languag skills play a major rol in th computr languag larning, as computr programming. Programming involvs complicatd trminologis and jargons which rlat to th English skills of th larnrs. Th lvl of English is dtrmind by administring an IELTS typ of xamination. If th studnt s ovrall prformanc is lss than Grad 5, thn thos studnts ar not considrd for our study. Th xamination consists of th rading, listning, spaking and writing skills. According to th obsrvation of th languag skill tst prformanc, th studnt s ovrall avrag was 7.46. 3. EXPERIENTAL DESIGN Th studnts usd th two visualization tools to larn ithr or programming. A control group larnt th sam programming without any visualization tool. Thy larnt th concpts using class room taching mthods. This control group could hlp to study th impact of using th visualization tools in rducing th cognitiv load. In th total of forty studnts, fiftn studnts slctd programming languag and fourtn studnts slctd languag. It is to b notd that lvn studnts had alrady compltd languag in thir Pr Univrsity and thy slctd languag. Ths thr groups as mntiond had to larn th various programming concpts using th visualization tools or traditionally. Th larning consists compriss of 48 hours for all groups including th introductory sssion. Th introductory sssion gav an ovrviw of th xprimnt and orintation of th visualization tools to b usd in th xprimnt. Th studnts wr tstd for thir undrstanding of th concpts larnt by a short tst at th nd of larning for ach concpt. Th scors in th short tst in ach concpt is th masur of prformanc of th larnrs. 4. COGNITIVE LOAD CALCULATION NASA TLX workload valuation procdur is a two-part procdur rquiring th collction of both wights and ratings from th studnts and th manipulation of th collctd data to provid wightd subscal ratings and an Ovrall Workload scor. Thr ar fiftn possibl pair-wis comparisons of th six scal lmnts that contribut to cognitiv load. Th subjcts ar givn a flip book which has th pairs of two lmnts that constitut th load. Th subjct chooss th lmnt which h fls and has contributd to th load and that lmnt is calculatd as a factor for cognitiv load masurmnt. Th lmnt that constituts th cognitiv load is slctd th load which prsnts ach pair to th subjct on pair at a tim. Th ordr in which th pairs ar prsntd and th position of th two lmnts (lft or right) ar compltly randomizd and diffrnt. Whn all fiftn possibl pairs hav bn prsntd, th scond part will b continud. Th scond rquirmnt is to obtain numrical ratings for ach difficulty attribut that rflcts th magnitud of that factor. Th subjct rats btwn 0 to 0. This is trmd as raw rating for ach 310

lmnt of th cognitiv load. Th wightd workload rating for ach lmnt in a task is simply th Wight (tally) for that lmnt a numbr btwn zro and fiv, multiplid by th agnitud of load, a numbr btwn zro and on hundrd. Thrfor, if th subjct had xprssd four tims for th wight of Tmporal Dmand and indicatd a magnitud of Tmporal Dmand in a particular task to b 45, thn th wightd workload du to Tmporal Dmand for that particular task would b 90. Th ovrall workload for a particular task is dtrmind by summing all of th wightd workload ratings for an individual subjct for th particular task and dividing by 15.Th abov mntiond procdur is adoptd for masuring cognitiv for ach concpt larnt in th cas of all th larnrs. 5. LT(LONG TER EORY CALCULATION) As mntiond arlir in this papr, w calculatd th numrical valu for th LT from th basic information about ach studnt. Th paramtrs considrd for calculating th valu for LT is Pr Univrsity athmatics marks, atriculation athmatics marks, Pr Univrsity English marks, atriculation English marks and th scor got in th English languag comptncy tst. Th calculation is don by assigning a fixd wight-ag for ach of th aspcts considrd for th LT. Th languag wight of th LT includs th grads scurd in th Pr Univrsity English, atriculation English and IELTS tst scor. Th othr aspct of th LT is rprsnting th analytical wight of th Long Trm mory which includs th grads of atriculation athmatics, Pr Univrsity athmatics and wight-ag of Pr Univrsity Computr Scinc grads. Th input for th languag wight is calculatd by assigning 50% wight-ag for th English Languag skill tst which was administrd to thm prior to larning programming, 5% wight-ag for th English languag grad in atriculation and anothr 5% is considrd from th Pr Univrsity English languag grad. Th sum of ths wights is hundrd. Th calculatd valus of LT ar shown in th subsqunt sctions. 6. RESULTS AND ANLSYSIS Th xprimnt was carrid out as statd in th prcding sction. During th xprimnt, th cognitiv load xprincd was masurd using NASA TLX and larning prformanc in ach catgory was calculatd for ach concpt larnt by studnts. Th rsults ar tabulatd in Tabl 1.Th masurs mntiond in th tabl for both cognitiv load and prformanc is th avrag scor for all th studnts in ach catgory. Tabl 1 Analysis Of Th Rsults Basd On Programming Languag And Visualizations Cognitiv Work Load Jav a C+ + C+ + to Jav a Tools La rn rs in a x Av ra g Tool 1 5 40 70 53 1 Prfo rman c Tool 5 38 75 57 1 Class room 5 4 77 57 13 Tool 1 5 39 69 56 1 Tool 5 43 68 54 11 Class room 4 44 77 56 11 Tool 1 4 41 7 54 11 Tool 4 37 78 53 1 Class room taching 3 44 69 57 11 It is obsrvd from Tabl 1 that whil larning th avrag cognitiv load was 57% and an avrag prformanc of 13 ovr 0 whil using class room mthod. Th sam situation applis to and to group whr th avrag cognitiv load is 56% and 57% rspctivly. Th cognitiv load for class room was highr than th groups using visualization tools. So it can b concludd that visualization tools do hlp in rducing th load. Th minimum cognitiv load for larning is got whil using th taching machin tool with a scor of 38% and th highst scor of 77% whil using th class room mthod. Howvr, whn taking th avrag load xprincd for all th fiftn larnrs of th group, Tool1 has lss man cognitiv load of 53. So it can b concludd that Tool1 is mor appropriat to larn programming languag. On th othr hand, it is difficult to gnraliz this conclusion, as th load varis for ach studnt by using th sam visualizations. Th larning prformanc is dtrmind by th grads of th studnts for ach concpt. In cas of larning, th larning prformanc was vry high for class room mthod whn compard to Tool 1 and Tool. This is 311

contradictory to th cognitiv load xprincd by th group. In ths xprimnts w hav considrd th cognitiv load xprincd as th main factor to masur th cognitiv load. It is du to th fact that th masurmnt of cognitiv load is don by a standard procdur. Th larning prformanc is usd as a control paramtr. It is usd to cross chck th rlation btwn th cognitiv load xprincd and prformanc. Th mismatch btwn th prformanc and cognitiv load is du to th fact it is avrag of all th larnrs. Thus, it is clar that thr is an individual diffrnc btwn th larnrs in trms of th cognitiv load xprincd. Th minimum cognitiv load of 39% whil larning is by using th Tool1 tool and th highst scor of 77% is obsrvd whil using th class room mthod. Howvr, whn taking th avrag load xprincd for all th fourtn larnrs of th group, Taching achin had lss man cognitiv load of 54%. So it can b concludd that tool is mor appropriat to larn programming languag. Howvr, it is difficult to gnraliz this conclusion as th load diffrs for ach studnt. In cas of larning, th larning prformanc was vry high for Tool1 whn compard to othr larning mthods which ar contrary to th highr cognitiv load xprincd by th group. Tabl givs th rsults of th xprimnts by concpt. Th tabl contains th information about prformanc, Cognitiv Load (CL) and Long Trm mory (LT) for Tool1, Taching achin(t) and Classroom. Tabl Summary Of Th Rsults Concpt Wis Tool1 Tool Classroom Pr CL LT Pr CL LT Pr CL LT 73 54.9 61. 63 56. 60.3 67.5 61. 57.6 3 9 54 54.6 61. 54. 53. 60.3 51. 55. 57.6 1 3 54.6 55.1 61. 56. 55. 60.3 48.7 54. 57.6 3 9 7 70.7 5.1 61. 73. 54. 60.3 75.4 54. 57.6 8 73. 54.4 61. 68. 53 60.3 73.7 56. 57.6 1 68.9 55.3 61. 70 56. 60.3 78.3 58. 4 57.6 It is obsrvd from th rsults in Tabl that Th minimum avrag cognitiv load for th first concpt was 54.9% whil using Tool1. Th avrag prformanc was also high whil using Tool1. Th lowst cognitiv avrag cognitiv load for th scond concpt is with th tool which is 53.1. orovr, th avrag prformanc is also highr for T. In th cas of third concpt, th lowst avrag cognitiv load whil larning using classroom mthod. In th cas of th fourth concpt, th avrag cognitiv load is lss whil using Tool1 and th highst prformanc avrag whil using class room mthod. For th fifth concpt, th avrag cognitiv load is lss whil using Tool and th avrag prformanc is high with th class room mthod. Th sixth concpt whil using th Tool1 and th highst avrag for prformanc is whil using th class room mthod. It is also obsrvd from th abov facts that thr is a variation btwn th cognitiv load xprincd and th prformanc lvl of ach individual studnt. This is du to th abov analysis don as a group study of all th larnrs larning a particular concpt. Individually thy ar uniqu by th background knowldg, dmographics and gndr. Th cognitiv load varis from larnr to larnr du to factors which ar intrinsic to th larnrs thmslvs. It is notd that th cognitiv load varis according to th concpts larnt for th sam studnt whil using th sam visualization tools. Th difficulty lvl of th concpts also altrs th cognitiv load of th larnrs. So w hav analyzd th data individually in th following paragraphs. Tabl 3 shows th cognitiv load xprincd and th prformanc by th individual larnrs whil using Tool1. Thr wr 3 larnrs ach from Urban and Smi Urban ara and thr wr 4 larnrs from rural ara. Thr wr 6 fmals and 4 mals who larnt using Tool1 31

Tabl 3: Summary Of Th Rsults Tool1 St Id Progra mming Langu ag Ara Gn dr 7 Urban F mal 15 al 18 F mal 3 S.Urb an F mal 4 F mal LT 6. 7 59. 9 76. 1 54. 9 69. 6 7 6. 8 9 Rural al 10 F mal 5 F mal 8 al 6. 3 65. 58. 4 56. CL Prform anc 55. 83 56. 4 61. 61 54. 11 51. 53.3 5.5 64.16 71.6 56.66 57 58.3 49. 89 49. 4 56. 94 53. 06 63.3 68.3 61.66 61.6 It is obsrvd from th tabl 3a that Th lowst cognitiv load was 55.83% and th highst cognitiv load was 61.61%. LT also varis from 76.1% to 59.9% for th urban larnrs Th lowst cognitiv load was 51.% and th highst cognitiv load was 57%.LT also varid from 54.9% to 6.8% for th smi urban larnrs Th lowst cognitiv load was 49.4% and th highst cognitiv load was 56.94%. LT also varid from 56.% to 65.% for th rural larnrs. Th following conclusions can b mad from th obsrvations Th cognitiv load xprincd by th individual studnt, whil larning using th sam tool and th sam concpts, varis vn though thy blong to th sam homognous group basd on thir dmographics. Ths variations ar du to th lvl of th LT of ach individual larnr. Ths variations in th LT also affct th lvl of cognitiv load xprincd. Thus th prformanc of th studnts also altrs accordingly. Tabl 3b: Cognitiv Load Whil Using Tool1 - Gndr Wis W dcidd analyz th rsults in thr aspcts whil using Tool1. Th thr aspcts includ dmographics, gndr and programming languag which ar shown in tabls 3a, 3b and 3c rspctivly. Tabl 3a: Cognitiv whil using Tool1- Dmographic wis Urban Smi Urban Rural LT CL P r LT CL Pr LT C L Pr al LT CL Prf LT CL Pr 6.7 55.83 53.3 59.9 56.4 5.5 76.1 61.61 64.16 6.8 57 58.3 54.9 54.11 71.6 6.3 49.89 63.3 69.6 51. 56.66 56. 53.06 61.6 65. 49.4 68.3 58.4 56.94 61.66 6.7 55.8 53.3 59.9 56.4 5.5 76.1 61.6 64.1 54. 9 69. 6 6. 8 54.1 71. 6 51. 56. 6 57 58. 3 6.3 49.8 63. 3 65. 49 68..4 3 58.4 56 61..9 6 56. 53 61. 6 It is obsrvd from th tabl 3b that Th lowst cognitiv load was 49.4% and th highst cognitiv load was 61.61%. LT also varis from 54.9 % to 76.1% for th fmal larnrs. Th lowst cognitiv load was 49.89 % and th highst cognitiv load was 57%. LT also 313

varid from 56.% to 6.8% for th mal larnrs. Th following conclusions can b mad from th obsrvations Th cognitiv load xprincd by th individual studnts, whil larning using th sam tool and th sam concpts, varis vn though thy blong to th sam homognous group basd on gndr. Ths variations ar du to th lvl of th LT of ach individual larnr. Ths variations in th LT also affct th lvl of cognitiv load xprincd. Thus th prformanc of th studnts also changs accordingly. Tabl 3c: Cognitiv Load whil using Tool1 Programming Languag wis LT CL Pr LT CL Pr 54.9 54.11 71.6 76.1 61.61 64.16 6.7 55.83 53.3 69.6 51. 56.66 6.3 49.89 63.3 58.4 56.94 61.66 59.9 56.4 5.5 6.8 57 58.3 65. 49.4 68.3 56. 53.06 61.6 It is obsrvd from th tabl 3c that Th lowst cognitiv load was 49.4% and th highst cognitiv load was 55.83%. LT also varis from 54.9 % to 65.% for th larnrs. Th lowst cognitiv load was 51. % and th highst cognitiv load was 61.61%. LT also varid from 56.% to 76.1% for th larnrs. Th following conclusions can b mad from th obsrvations Th cognitiv load xprincd by th individual studnts,, whil larning using th sam visualization tool and th sam concpts, varis vn though thy blong to th sam homognous group basd on th programming languag. Ths variations ar du to th lvl of th LT of ach individual larnr. Ths variations in th LT also affct th lvl of cognitiv load xprincd. Thus th prformanc of th studnts also changs accordingly. Tabl 4 shows th cognitiv load xprincd and th prformanc by th individual larnrs whil using tool. Thr ar tn studnts who larnt using tool1 l as shown in th following tabl 6.8. Thr wr 3 larnrs ach from Smi Urban and Rural ara and thr wr 4 larnrs from urban ara. Thr wr 7 fmals and 3 mals who larnt using tool1. Tabl 4: Summary Of Th Rsults Using Th Tool Prog Languag Ara Gndr LT CL Pr Urban al 69.4 5.67 57.5 Urban 71.3 56.78 59.1 Urban 65. 54.5 61.6 Urban 56. 51.44 45.83 S.Urban al 58.5 57. 69.1 S.Urban 5 53.33 60.83 S.Urban 73.6 60.4 59.16 Rural al 51.3 61.39 61.66 Rural 70.7 56.11 54.16 Rural 57.5 49.94 6.5 W dcidd to analyz th rsults in thr aspcts whil using T tool. Th thr aspcts includ dmographics, gndr and programming languag which ar shown in th tabl 4a, 4b and 4c rspctivly. Tabl 4a: Cognitiv Load Whil Using Tool - Dmographic Wis Urban Smi Urban Rural LT CL Pr LT CL Pr LT CL Pr 69.4 5.6 57.5 58.5 57. 69.1 51.3 61.3 61.6 71.3 56.7 59.1 5 53.3 60.8 70.7 56.1 54.1 65. 54.5 61.6 73.6 60.4 59.1 57.5 49.9 6.5 56. 51.4 45.8 314

It is obsrvd from th tabl 8.a that Th lowst cognitiv load was 51.44% and th highst cognitiv load was 56.78%. LT also varis from 56.% to 71.3% for th urban larnrs Th lowst cognitiv load was 53.3% and th highst cognitiv load was 60.4%. LT also varid from 5% to 73.6% for th smi urban larnrs Th lowst cognitiv load was 49.4% and th highst cognitiv load was 61.39 %. LT also varid from 51.3% to 70.7% for th rural larnrs. Th following conclusions can b mad from th obsrvations Th cognitiv load xprincd by th individual studnts, whil larning using th sam tool and th sam concpts, varis vn though thy blong to th sam homognous group basd on thir dmographics. Ths variations ar du to th lvl of th LT of ach individual larnr sinc th othr paramtrs lik concpts larnt and visualization usd ar sam. Ths variations in th LT also affct th lvl of cognitiv load xprincd. Thus th prformanc of th studnts also altrs accordingly. Tabl 4b: Cognitiv Load whil using Tool - Programming Languag wis LT CL Prf LT CL Prf 69.4 5.67 57.5 56. 51.44 45.83 71.3 56.78 59.1 5 53.33 60.83 65. 54.5 61.6 73.6 60.4 59.16 Tabl 4c: Cognitiv Load Whil Using Tool - Gndr Wis al LT CL Pr LT CL Pr 71.3 56.78 59.1 69.4 5.67 57.5 65. 54.5 61.6 58.5 57. 69.1 56. 51.44 45.83 51.3 61.39 61.66 5 53.33 60.83 73.6 60.4 59.16 70.7 56.11 54.16 57.5 49.94 6.5 From Tabl 4b and 4c, w can obsrv that th highst cognitiv load was 61.4 and th lowst load was 5.6 whil larning. On obsrvation of ths two cass, w could s that th LT is high for th larnr who xprincd lssr cognitiv load and vic vrsa. Th sam is th cas with larning languag. Th highst cognitiv load is 60.4 and th lowst cognitiv load is 49.4. Whn obsrving ths two cass, it is found that th LT valu is th lowst for th studnt who has highr valu for cognitiv load and vic vrsa. Th sam applis for th valus basd on gndr. So it can b concludd that th LT lvl affcts th lvl of th cognitiv load xprincd. It is clar from th abov xampls that th cognitiv load varis from studnt to studnt vn though thy blong to a homognous group. Ths diffrncs ar du to LT and othr control factors such as gndr, dmographics and programming languag. Tabl 5 givs th Cognitiv load xprincd by th studnt and th prformanc whil using Class Room mthod. 58.5 57. 69.1 70.7 56.11 54.16 51.3 61.39 61.6 57.5 49.94 6.5 315

Prog. Lang Tabl 5: Summary Of Th Rsults Using Class Room thod Tabl 5 shows th LT, CL and prformanc of larnrs whil larning with th classroom mthod. It is obsrvd that as in th cas of th Tool1 and Tool, th cognitiv load varis from individual to individual vn though thy blong to a homognous group. In th class room mthod, it is obsrvd that th highr cognitiv load is xprincd by larnrs who hav highr valu of LT. This is quit contrary to th othr two groups whr th studnts larnt through visualizations. Tabl 5a: Cognitiv Load Using Classroom- Program Wis Ara Gndr LT CL Pr Urban S.Urban S.Urban S.Urban S.Urban S.Urban Rural Rural Rural al al al al LT CL Pr LT CL Pr 64.7 5.56 64.16 61.5 5.33 60.83 55.7 57.5 60.8 55.8 55.83 53.33 60.6 58.7 78.3 41. 54.11 53.33 58.3 56.67 60 59. 63.5 57.5 64.6 61.39 5.5 61.5 5.33 60.83 64.7 5.56 64.16 55.7 57.5 60.8 60.6 58.7 78.3 55.8 55.83 53.33 41. 54.11 53.33 58.3 56.67 60 64.6 61.39 5.5 59. 63.5 57.5 As an xampl, lt us considr th highst cognitiv load in Tabl 5a whil larning which is 61.39. Th lowst cognitiv load is 5.5. Th corrsponding LT valus ar 64.6 and 64.7. In spit of th sam lvl of LT, thr is a diffrnc in th cognitiv load. It is clar that visualizations do hlp larnrs in rducing th load. It is obsrvd that th larnrs who had good LT valu, xprincd highr cognitiv load du to th larning mthod adoptd. 7. DISCUSSION ON RESULTS In this study, th cognitiv load is masurd by th larnrs xprssion of cognitiv load using th NASA TLX scal systm. Th larning achivmnt is masurd by mans of prformanc on th basis of grads scurd for ach modul. Th obsrvation provids multidimnsional facts of th cognitiv load xprincd by th larnr during th procss of larning which includs mntal load, prformanc, frustration, tmporal load and th ffort which corrlat to th working mmory. Th LT schma is also considrd in th study by collcting th basic background of th larnrs in athmatics and English languag and prior programming knowldg. It is clar from th analysis that th cognitiv load xprincd by th larnrs diffrs from individual to individual vn though thy blong to a homognous group. This fact is clar from our discussion of th rsults of cognitiv load whil using diffrnt visualization tools ar analyzd on th basis of gndr, dmographics and programming languag. Cognitiv load varis for th sam larnr whil larning diffrnt concpts in spit of using th sam visualization. This shows that th cognitiv load is also affctd by th lvl of difficulty of th concpt. So, w cannot gnraliz th ffctivnss of all th visualization tools for all th concpts. Som concpts ar mad asy by using crtain visualization and whras th sam tool is not ffctiv for som othr concpts. This could b du to varying lvls of difficulty of concpts. This is also clar from th study that it would b appropriat to visualiz diffrnt concpts with diffrnt lvls of usr s intraction in th visualization tool. Th lvl of difficulty for ach concpt also dtrmins th ffctivnss of various visualization tools. From th analysis of th xprimntal data, w dcidd to dvis a mchanism to slct th appropriat typ of visualization tool for studnts on th basis of cognitiv load xprincd and prformanc and taking into th considration th factors that contribut to th LT. So this approach will hlp in providing th appropriat tool for larning suitabl for individual larnrs. 8. FRAEWORK FOR OPTIIZING INSTRUCTION This variation of th load is du to many factors which includ dmographics of th studnt and th schmata of th Long Trm mory which is basd on th prior knowldg of th English 316

languag, athmatical background and programming knowldg. It is difficult to dtrmin th bst tool for larning for vry individual studnt as it dpnds on various factors mntiond bfor. Th nxt stp is to dvis a suitabl mchanism to prdict th bst tool for larning for ach studnt considring many factors that affct larning. So, using th data from th xprimnts don in th prvious chaptr w construct a tool that would b abl to rcommnd th bst tool for larning programming for vry individual usr. W startd to xplor on how such a tool could b implmntd. A problm whos output is associatd with many factors can b asily rprsntd using th Nural Ntwork. Thr ar many prior works availabl in th litratur whr nural ntwork is usd in ral tim xampls such as sals pric prdication, stock pric prdiction,forcasting financial conomical sris) and ral stat pric prdiction. Artificial Nural ntworks hav bn usd in many applications rlatd to ducation filds lik Intllignt Tutoring Systm. Artificial Nural ntwork is highly succssful in arriving at prdictions whr thr is a high chanc of uncrtainty. Also in our study highr lmnt of uncrtainty xists, as th cognitiv load is complx and varis according to larning lvls and ability of th larnr. So w dcidd to dvlop a tool using nural ntwork modl to solv th larning difficulty of th studnts basd on th cognitiv load and othr rlatd factors. Th studnts fdback using NASA TLX scal is considrd as on of th inputs to th systm in ordr to slct th bst tool for thir optimizd larning. Th othr factors such prior knowldg of English, athmatics and Computr will also b considrd as input for th systm. W dcidd to us th suprvisd larning mthod in our nural ntwork modl to dtrmin th bst visualization tool as an output. Th ntwork can b traind using input of cognitiv load from th NASA TLX scal and also th Long Trm mory (LT) which is basd on th prior knowldg of athmatics, English and programming. Th wights for ach of ths inputs ar simplifid and discussd in th subsqunt sctions. Figur 1: Proposd odl Of NN Implmntation Th abov diagram shows th framwork of th tool for optimizing larning is built. But at th prsnt th intgration of th visualization modl with th nural ntwork modl is not accomplishd. Both th modls work as sparat ntitis. Prdiction of th tool is don by th nural ntwork modl and on th basis of th rcommndation th appropriat visualization tool is rcommndd to th larnrs for larning th concpts. 9. NEURAL NETWORK ODEL FOR COPUTER PROGRA LEARNING Th Nural Ntwork (NN) modl basd is composd of diffrnt layrs. Th input layr paramtrs includ th various cognitiv factors calculatd using NASA TLX scal, athmatical background, analytical background and tst prformanc.th output layr of th ntwork includs th rcommndation of th tool for optimizd larning. Th prdictions ar basd on th cognitiv load of th larnrs as wll as thir prformanc in a particular task. Th input to th nural ntwork is chosn so that it accounts for both th working mmory and long trm mmory. Ths two mmoris play an important rol in larning as pr th Cognitiv Load Thory (CLT). Th major inputs givn for th nural ntwork modl ar basd on th prior knowldg of mathmatics and computr programming, prior English knowldg and in addition to th load xprssd using NASA TLX scal. Th first two inputs rprsnt th Long Trm mory (LT) and th last input rprsnts th working mmory. Th inputs to th nural ntwork ar simplifid which will b discussd in th subsqunt sctions. This systm has adoptd 317

th fd forward loop whrby succssiv itrations mak th systm mor fficint in prdictions. Th ntwork is as shown in Figur 7. Figur : Nural Ntwork odl For Th Slction Of Th Visualization Tool 9.1 Long Trm mory (LT) Schmata Input calculation Among th inputs for th nural ntwork ar valu associatd with th LT, th languag wight and th analytical wight of th LT. Th languag wight includs th grads scurd in th Pr Univrsity English, atriculation English and IELTS tst scor. Th analytical wight includs th grads of atriculation athmatics, Pr Univrsity athmatics and wight-ag of Pr Univrsity Computr Scinc grads. Th input for th languag wight is calculatd by assigning 50% wight for th English Languag skill tst which was administrd to thm prior to larning programming, 5% wight-ag for th English languag grad in atriculation and anothr 5% is considrd from th Pr Univrsity English languag grad. Th sum of ths wights is on hundrd. It rprsnts th input to a maximum of 1 which is th cas of normal input to any nural ntwork. W hav assignd mor wight for th IELTS typ xamination as it rflcts th currnt stat of th studnt s English knowldg. Th maximum grads for English languag skills tst is 10 Points. Th maximum marks for matriculation English and Pr Univrsity English ar 100 and 00 rspctivly. Th English languag is 50 % of th load of th LT. Th languag wight of th LT is calculatd using th formula givn blow. 4 8 whr a = IELTS marks b = atriculation English marks. c = Pr Univrsity English marks. Th analytical wight of th LT is also calculatd in th sam mannr as th languag wight by considring th atriculation athmatical marks Pr Univrsity athmatical grad and Computr Scinc marks in th wightag as mntiond. Th 50 % of wight is assignd to th matriculation mathmatics marks rmaining 50% is assignd qually to th analytical skill which is basd on th Pr Univrsity athmatics marks and Pr Univrsity Computr Scinc marks. Th maximum marks for atriculation athmatics is 100 and Pr Univrsity athmatics and computr scinc is 00.W hav assignd mor wight for atriculation athmatics marks as this mathmatics forms th fundamntal knowldg rlatd to athmatical concpts. Analytical wight of Long Trm mory is calculatd as shown in th following formula. 8 8 whr d = atriculation athmatics marks. = Pr univrsity athmatics marks. f = Pr univrsity Computr scinc marks. 9. Cognitiv Load Input Calculation Anothr input to th nural ntwork is th cognitiv load xprincd during th task. Th training data for th inputs of this nural ntwork is basd on th xprimntal data don in chaptr 6.Th cognitiv load has two dimnsions basd on NASA TLX scal namly dmands imposd on th subjct (ntal, Physical and Tmporal Dmands) and th raction of th subjct with th task (Effort, Frustration and Prformanc).Th dmand imposd on th subjct has to b minimal and th raction of th subjct must b positiv (low scors) for th ffctiv larning procsss. Th wight is th valu dtrmind basd on th frquncy of th particular attribut of th cognitiv load rportd to b challnging using th standard procdur of flip book. Thr ar 15 possibl combinations availabl in th flip book. Evry occurrnc of th particular 318

attribut is ratd as on. Th total wight is th sum of occurrncs of that particular aspct of cognitiv load. Th total wight is convrtd to prcntag as shown blow in th formula. 15 100 whr WCLI = Wight of cognitiv load itm OW = Frquncy of th cognitiv load itm Th rating scal for ach of th six aspct of th cognitiv load is masurd using th scal of 0 in NASA TLX. Th rating scal valu is convrtd to prcntag by th following formula. whr 0 100 RP = Rating prcntag OR = Obsrvd rating 10. CONCLUSION AND FUTURE WORK Th abov proposd modl can addrss th individual diffrncs of studnts. Thus it catrs to optimiz th instruction according to th background profil and th schmata formation of ach individual studnt. Th proposd modl has to b validatd by conducting a study on th appropriatnss and accuracy of th modl in slcting th bst instructional tool for ach studnt considring th various paramtrs such as background mathmatical knowldg, English languag skills and xposur to IT skills tc.if th modl is validatd thn it could b xtndd to othr domains of study to addrss th difficultis of larnrs. REFERENCES [1] Alkinani,. (009). Softwar Visualization- Examination of its prsnt status and its futur applications. [] Amy B,W, Tracy C,G, Tai-Lung,C, Shrri, S. (000). Prsonality as a prdictor for studnt succss in programming principls I.Papr prsntd at th 7 th Annual confrnc of Southrn Association for Information systm. [3] Andrson E, W, Pottr K, C, atzn L, E, Shphrd J, F & Prston G, A, Silva C, T. (011). A usr study of visualization Effctivnss using EEG and Cognitiv Load. Papr prsntd at th IEEE symposium on Visualization. [4] Antonnko, P, Pass,F,Grabnr,R. (010). Using Elctroncphalography to asur CognitivLoad. Educational Psychol Rv,, 45-438. [5] Baddly, A,Dlla,S. (1996). Working mory and Excutiv control. Philosophical [6] Transactions: Biological Scincs., 351, 1397-1404. [7] Cartr, J, Jnkins, T,. (00). Gndr diffrncs in programming? Papr prsntd at th 7 th Annual confrnc on innovation and tchnology in computr scincd ducation. [8] Curtis S.Ikhara, & E.Crosby, artha. (005). Asssing Cogntiv Load with Physiologica Snsors. Papr prsntd at th 38th Hawaii Intrnational Confrnc on Systm Scincs,Hawai,USA. [9] du Boulay, B. (Ed.). (1989). Som difficultis of larning programming. Hillsdal,NJ: Lawrnc Erlbaum. [10] E.Tuovinn, Juhani. (000). Optimising Studnt Cognitiv Load in Computr Education.lbourn: AC. [11] E.Winslow, Lon. (1996). Programming Pdagogy- A Psychological Ovrviw. SIGCSE [1] BULLETIN, 8(3). [13] Essi, L, Kirsti,A,,Hannu,,J. (005). A study of th Difficultis of Novic Programmrs. Papr prsntd at th ITiCSE 005, ont d Caparica,Portugal. [14] Garnr.S. (009). A quantitativ study of softwar tool that supports a Part complt solutionthod on Larning Outcoms. Journal of Information Tchnology Education, 8, 85-310. [15] Goms, Anabla, & nds, A.J. (007). Larning to program-difficultis and solutions. Papr prsntd at th Intrnational confrnc on Enginring Education - ICEE 007,Coimbra,Portugal. [16] Hart, S.G., Battist. V., Chsny.. A., Ward..., and celroy,. (1986).Comparison of workload, prformanc and cardiovascular masurs:typ A prsonalitis vs Typ B.Working papr.offtt Fild,CA:NASA Ams Rsarch Cntr. [17] Hart, S.G. (006). NASA TLX Load Indx(NASA TLX):0 yars latr. Papr prsntd at th Ergonomic factors socity 50th ting, Santa onica. [18] Iain, & R, Gln. (00). Difficultis in Larning and Taching Programming- Viws of Studnts and Tutors. Education and Information Tchnologis, 7(1), 55-66. [19] Ikhara,S, Curtis, & Crosby.E, artha. (005). Assssing Cognitiv Load with Physiological snsor.papr Prsntd at th 38 th Hawaii Intrnational Confrnc on Systm Scincs 005,Hawaii USA. [0] Kaastra, I, Boyd,. (1996). Dsigning a nural ntwork for forcasting financial and conomic tim sris. Nuro computing, 10, 15-36. 319

[1] Kolb, D, A (Ed.). (1985). Larning Styl Invntory: Tchnical anual. Boston: cbr and company. [] utka, Kirsti Al. (007). Problms in Larning and Taching Programming. Institut of Softwar systms, 13. http://www.cs.tut.fi/~dg/litratur_study.pdf [3] P.Bruc-Lockhart, ichal, & S.Norvll, Thodr. (000). Lifting th Hood of th computr:program Animation with th Taching achin. Papr prsntd in Elctrical and Computr Enginring, 000 Canadian Confrnc. [4] Paas, F, Juhani E.T, & Huib, T. (003). Cognitiv Load asurmnt as a mans to Advanc Cognitiv Load Thory. Educational Psyhchologist, 38(1), 63-71. [5] Pnnington.N. (1987). Comprhnsion stratgis in programming S. S. E. S. G Olson (Ed.) Emprical studis in programming (pp. 100-11). [6] Rajala T Laakso J, Kaila E, Salakoski T. (007). ViLLE uti languag Tool for Taching Novic Programming TUCS Tchnical rport. Hlsinki: Tampr Univrsity,Finland. [7] Rajala T Laakso J, Kaila E, Salakoski T (008). Effctivnss of program visualization A cas study with ViLLE Tool. Journal of Information Tchnology Education, 7, 15-3. [8] Robins, A, Rountr,J,Rountr,N. (003). Larning and Taching Programming: A Rviw anddiscussion. Computr Scinc Education, 13(). [9] Schonburg, E. (1990). Stock Pric Prdiction Using Nural Ntworks :A Projct Rport. Nuro computing,, 17-7. [30] Shaffr, Dal, Doub, Wndy, & Tuovinn, Juhani. (003). Applying Cognitiv Load Thory to Computr Scinc Education. Papr prsntd at th Joint Confrnc EASE&PPIG 003. [31] Soloway, E (Ed.). (1983). What do novics know about programming?. Ablx: Norwood,NJ. [3] Soloway, E. (1986). Larning to Program = Larning to construct mchanism and xplanations.communications of AC, 9(9). [33] Soloway, E, Spohrr,J, (Ed.). (1989). Studying th Novic programmrs, L. Erlbaum Associats.Nw Jrsy.. [34] Swllr, J. (1988). Cognitiv load during problm solving : Effcts on larning. Cognitiv Scinc,1, 57-85. [35] Widnbck, S, Ramalingam, V, Sarasamma, S, Corritor, C. (1999). A comparison of th [36] comprhnsion of Objct Orintd and procdural programs by novic programmrs. Intraction with computrs, 11(3), 55-8. [37] Wilkowski, W, Budzynski,T. (006). Application of Artificial Nural Ntworks for Ral Estat Valuation. Papr prsntd at th XXXIII FIG Congrss, unich,grmany [38] Winslow, L.E. (1996). Programming pdagogy- A psychological ovrviw. SIGCSE BULLETIN, 8(3), 17-. [39] Yu Shi, & H.C.Choi, Eric. (007). Galvanic Skin Rspons (GSR) - as an indx of Cognitiv Load.Papr prsntd at th CHI 007, San Jos, CA, USA. 30