Optimising Student Cognitive Load in Computer Education Juhani E. Tuovinen, Centre for Leaming and Teaching Support, Monash University, Churchill, Vic. 3842. Senior Research Fellow in Interactive Multimedia, Monash Univemlty
[email protected] NDstract
The work descn'bed in this chapter is a synthesis of recent insu'uctional cognition research implications for the planning of computer education. The research described here has been conducted in the Cognitive Load Theory context. The leading research group in this area is located at the University of New South Wales, in Sydney, and a complementary cognition research program is based at the Open University of Netherlands. The work ew~nafing f ~ m these groups and allied efforts is applied in this paper to improving the plarmin~ and /mplementation of computer education.
Cognitive Load Theory provides a coherent way to opt/m/~e student cognitive processing load during lear,ln S, A range of principles identified in this theory can be appfied to invrove student processing during learning computing content. These principles range ffi~m the goal free problem solving to worked e~amples, split-attention, redundancy and variability effects. However, these principles need to be applied strategically. In this paper the most ;,.,:,ozlant considerations for slz'ategjc plannin.~ of computer education, ranging from the content element interactivity, mental effort measurement to student prior knowledge, are organised into a set of instructional choices. These choices are summarised in a flow chart, which can be used in the ecb~cafionalplanning, as a tool to help ensure the identified issues are considered in an optimal sequence.
The objectives of this paper axe to suggest a theoretical fo,mdation for compWong instruction, and to distil from the relevant cognition research a m~mher of practical implications for educational computing planning, design aJld use.
1 Introduction In this paper the su'ucture of human cognitive architecture will be described fi'om an infonmtion processing l ~ e c t i v e . Then the cognitive load theory will be introduced. The interaction of various possible forms of compu~ng instruction and learning materials development based on cognitive load theory and material complexity or element intemctivity plus prior knowledge will be considered. Guidelines for the effective design and use cognitive principles will be noted in each section. Finally a general smTn~ry flow chart of the instructional design choices in optimal application of the cognition prin~les will be developed and future research interest will be cfiscussed.
2
Background
2.1 Oognitive ~wchlteetum The research discussed in this paper views the human mind as an information processing system, The architecture of the b,nn~n mind is thought to consist of three basic components, Sensory Memory, Work~n~ Memory and L o n g - t ~ Memory [1], which roughly conespond to the input, processing and storage component stages of computers. This view has provided a useful basis for developing theories with sisnificant learning and teaching imptications. Figure 1 3 component model of the ] ~ processing system
information
W o r k i n g Memory [
,~ (PROCESSING) ]~ (STORAGE)
The b.rnan mind receives information fi'om the outside world through the senses (Input stage) which are decoded in the Sensory Memory. The information fzom the Sensory Memory is then processed in the Working Memory (Processing stage) and stored in the Long-term Memory (Storage). Of course the previous information stored in the long-term memory can also be accessed, or activated, to help deal with the processing in working memory [13].
Pemd.tskm to make digilal o~ hanl ropier of all or p m of this work for permml ~ ~ me b granted ~ fee i~ovided tl~t e e l ~ me not mtde ef dism~uted for profit or connnemial advanlage, amd that copies bear tim nmke and the full citation on the ~ t t imSe. To copy ofl~rwi~., W republ/.~h, to pod o9 ~ , ~ r s or w redinn'btiR W ACE 2000 12/00 Me~oume, A u ~ m l b © 2000 ACM I-$8113-271-9/00/00! 2 ... ~.G0
235
One of the most interesting and significant aspects of the human mind is the very small capacity of the working memory. In 1956 G. A. Miller coined the famnus term, "the magical n,mher seven plus or mi-us two" to descn'be the number of ~ items he thought b,,m~n~ could hold in the working memory at any time [14]. Since then the exact number of items has been shown to depend on a ,,reSet of factors, ranging fi'om age, health, level of fatigue, the type of item, familiarity with the content, trainin S, etc.[3, 23]. However, without doubt the capacity of the worifing memory to deal with distinct items is quite 1/,-ited, whereas the capacity of the long-teim memory is very large, in fact, no clear boundm7 has been esmbli~h~ for it [22]. There are tWO ~ ~ nsech,,i~re~ that help to overcome the working memory limitations. They are Schema Fonmtion [11] and Automation [20, 21]. If the items of infu,~tion are grouped together in a me~ninLfful way, e.g. 16031977 vs. 16.03.1977 (my son's birthday in the Australian dd -.,._yyyy format) they become easier to remember and use as one item, called a chunk [10, 14]. of right separate items, they can be treated as a single entity by work/r,~ memory. The second mechanism is automation, i.e. processing that is so familiar that one does not have to th/rdc about the ccvm~onents of the processing consciously. This is the type of processing fluent readers use when rcading text, where they do not try to rrmk~ out hxlividual letters, but process larger groups, words or groups of words, without attend/rig to individual letters or even words separately. As one develops better schemas and automation one gains expertise in a given field. Then one is able to select and use more elaborate schemas and automated processes to avoid the bottleneck of processing in the working memory, with too many individual separate items of information leading to confimion and poor processing due to a ~ overload.
2.20ognltlvo Load Theory A major theory of learning ,rid problem solving, called the Cognitive Load Theory, is based on the above view of the mind [28]. A number of si~nificaut advances in learning have been developed in this theory context as briefly introduced below. Each of them has been shown to have s~ong effects on leanRug. 2.2.1 Goal-free effect Search in problem solving practice may be reduced by ch*-~-~ conventional problems with specific goals to reach after many steps, to simpler, more general explorations of the content area, leading to better learning [2, 27]. The search involved in the means ends analysis problem solving process, co ..... .nnly employed to solve goal directed practice problems, reduces the capacity for relevant learnln~. Le. schema formation. So reducing needless search mAk,S more effective use of the lirr~ed
236
worlang memory processing capacity resulting in better learning. For example, in the kinematics topic in physics a small number of equations of motion under constant acceleration are mually taught in senior high school. After being the taught the equations of motion, the smdenm are co ...... nnly given problems of the type: I f a body starts from rest under the action of a constant acceleration of 3 rn/sec2, find the distance moved over 4 seconds. This requires the students to man/pulate the equations of motion, combining them and sub~tuting known values for the variables until they find the value of the distance variable. The following alternative way (goal-free format) of presenfin~ the same problem was tried aud found to be more effective: I f a body starts from rest under the action of a constant acceleratkm of 3 m/sec2, flnd all the things you can know about the body after it has been in motion for 4 seconds (using the equations of motion). 2.2.2Worked example effect In a worked examples approach instead of students immediately practicing by solving problems after a lesson, they study the solu~ons to many s/rnilar example problems, before attempting own solutions [26, 40]. Studying the worked examples reduces the search involved in meansends analysis problem solving practice, leading to more effective schema development For e~,rnple the author in tese-hln~ dembasos to university students showed that worked examples were si~ificaut/y more effective for students learning complex content if they had little prior experience of databases [33]. After studying a sufficient mrmher of worked examples the students are ready to carry out the ~quired tasks for themselves. 2.2.3 Completion problem effect Instesd of students reading worked examples and then solving new problems, partially completed problems are introduced with gradually reducing amounts of assistance [17, 34, 36]. This approach was developed for students learning computer prograrrn'n/ng and it combines the benefits of the worked examples and helps to engage students in productive schema development. 2.2.4 SplR-atIenfion effect Difficulties of students needing to study physically separated materials, can be overcome by integrating learning materials, such as graphics and text [29, 40]. For example, ChA-mer [8] revised sttuty materials on programmin~ numerical control computers which included diagrams and separate text that referred to the graphics, by placing the text strategicallyon the diagram. This led m
more effective learning due to physical integration of the materials, loading to easier mental integration.
One of the main findings of the Cognitive Load Theory is that ff the learning content is easy, the format of instructional materials and the learning process do not matter much. If the content is den~nding, both the structuring of the learning materials and the learning process become more critical. Thus it is important to measure the complexity (element interactivity) of the learning content. We may estimate it from an examination of the material itselI~ as described in [25]. They analysed in close detail the steps involved in locating a point on x-y coordinate rods and found that seven distinct steps were involved. However, even though to a complete novice point location involves seven steps, as we work with these concepts we group them into larger cbn-k~, and develop schemas for these tasks, which our worlc/.~, memories can treat as single elements. In other words, our prior knowledge of the content to be learned critically affects our ability to process it.
2.2.5 Modality effects Separate processing of visual and auditory material in the working memory allows complementary materials to be utilised more effectively th.. material presented in a single modality [15, 30]. This is one of the key issues for multimedia learning. 2.2.6 Redundancy effect If an exiting presentation, such a circulatory diagram with arrows indicating blood flow direction, already conveys adequate information, physically integrating more textual infonmtion with the diagram, such as explmmtory text linked to the graphics, may cause problems for learning [7]. The s t u d e ~ are already able to gain the required information ffi'omthe diagram without the text, and the text becomes an mm~cessary burden to be processed.
A classic example of the prior knowledge of the domain affecting learning with multimedia was reported by [16]. They found that in a learning program employing m-ltim~dia with other" text and diagr~,,.,~tic information sources the multimedia was of no benefit to participants with good prior knowledge, but of si~ificaut benefit to students without prior knowledge.
2.2.7Vadability effect Provided the les.rnin~ coutent is not so difficult that it totally swamps the working tmmory [18], exercises that vary in their deep structure more effectively assist students to recognize appropriate solution schema classes than exercises with only one solution structure [12, 19].
If prior knowledge is so critically important, how do we find out the students' prior knowledge in a given area? Do we need to ask a huge number of questions to find out the relevant knowledge? For example, in a recent study of learmng to construct complex database components the students were asked more than a hundred questions about their background with computers, databases, etc, [31]. However, the analysis indicated only one question was critical to eliciting the relevant prior knowledge. It was:
2.2.8 Imagination effect M~.i=l rehearsal, 'irna#n~,tion', provides better learn/n~ th~, conventional study [6]. If students mentaliy rehearse the operation of procedural components, e.g. in computer spreadsheet operatiom, they learn better t ~ . through conventional study. 2.2.9 Rote vs meaningful learning Rote learning can be a useful basis for later learning with m~aning with d~r'nnndmg content, whereas constant learning with me~-/~g may overload the working memory at the earlier stages [6].
Please tell us about your experience with the following types o f sofhcare, indicating your usage and naming the sofhcare packages you most commonly use (iflmown): Databases, 1 = Never, 2 = Seldom, 3 = Sometimes, 4 = Often, 5 = Very often.
Principally the Cognitive Load Theory work has been aimed at reducing the cognitive load on the working memory dining learn/nE~ This theory is based on the universal structure of b~nre~ mind and so its implications can be readily applied in education generally. This paper deals mainly with the application of the cognitive load theory principles to computer education. Other educatioml implications of these findings are presented in [24, 28].
If the students had any experience with databases at all, i.e. ff they selected any of the options 2 -5 (the actual results were in the range 2 - 3) in the above question, after a lesson on the construction of complex fields in a new database and one example, they could learn productively by Uying out the procedures for themselves. However, for the students without prior exposure to databases the sel£directed exploration practice was ineffective, and they needed carefully explaine~d worked examples before they could solve typical calculation field construction problems.
Strategic use of Cognitive Load Theory principles 3.1 Element InteracUvfty and Pdor Knowledge
3
237
So what can we learn from this? •
We can identify relevant prior knowledge to guide appropriate level/type of instruction by asking students questions before the learning sessions. We might begin with , - , - y items on an ini~al questionnaire to elicit the prior knowledge, but we may be able to reduce it to a very small number of critical items after an analysis of their relevance to learning. If the complexity (element interactivity) of the material is high, and the students have poor prior knowledge of the area, they benefit from caiefully constructed multimodal instruction, heavy use of worked u-mples and other cognition research based materials and methods. On the other baud, developing nmltimodal instruction or olher instructional materials or methods based on the cognitive load theory principles for low complexity material or for students with good prior knowledge may be a waste of resources and could even be counterproductive for learning due to redundancy effects.
3.2 Formatof Materials and Extraneous Cognltivo Load
apply new content to many contexts. However, there is a ceiling on how difficult these tasks can be. 4 Strategic planning of computer education In pl,rming how to incoxpomte the cognition principles we could use a process such as shown on the following flow chart. By asking key questions the strategic directions for applying the cognitive load principles listed above may be developed. These are shown in Figure 2. 5
Future bends
The stlategic pi.nnin~ of i n s ~ o n a ] design to incorporate cognitive load theory principles in computer education rests on general educational theory, and on too numerous e x - ~ l e s of this theory into student learning of comput~.~ for space to allow to discuss them all, e.g. [4, 5, 9, 31, 33, 35, 37-393. What is desired is research to investigate how the above principles may be included in various forms of modern computer education, e.g. in both off-line and on-line modes [323, and the collection of empirical examples of how these principles can be best appfied.
We noted above that often the physical separation of materials, which need to be integrated during learning and the provision of izrelevaut materials reduce learning effectiveness. These inappropriate ways of formatting educational learning, content create an extra load for the students to process, over and above the necessary core content. Tr~...... ;rig the exuaneous load by various methods improves the learning [4, 24]. For example, Chandler [9] reduced the split.attention effect in learnln~ computer progr~rn~ from texts, by requiring students to attend to either the text or the computer screen only, rather than expectin S students to split their attention between two different locations for new information.
Conclusion A review of educational cognition research has been as applied to co]~)utEI educsfion. A11 u I ~ n d i n ~ Of the human cognitive architecture provides a theoretical basis for appropriate design of computing instruction. For example, for low mental processing d ~ - , ~ content nndtimodal and other par~cular designs of instruction are not needed, and they may even reduce learning. However, for students with poor prior knowledge of complex issues, use of cognition based insUucfion, such as mnlti~dal instruction, provides an effective way to overcome some of the work'inS memory processing limitations.
So we note lhat: • If content is complex, try' to pres=~ material together rather than apart. • If the content is simple, presentation does not matter so
The s~cb~nt cognitive processing load in such contexts needs to be investigated and a strategic program for applying cognitive principlesadopted, as outlined in the diagram above. Appropriate methods need to be developed to adapt imu-uc~on to the students' cognitive processing needs and capacities, e.g. by employing intelligent computer aided i n - - d o n , which relies on effective meas~nt of the students' cognitive load during learning.
much.
3.3
Variability of L.amlng Tasks
Once the declarative content knowledge is mastered, the application of the new knowledge to a different of contexts (procedural knowledge development) is facilitated by that vary in their setri-~ and structure [18], provided the task total cognitive load is kept below the working memory limiL This indicates that there is good value in students att~u~tlng t~ks that challenge them to
238
6
Figure 2 StrategicPlanning of Computer Education
m Nc
Yes No
Variable Tasks: Apply variabit~j effect pr ,les to str spr tasks.
Yes
No
Materials MocUfieation: Modify the learningmaterialsaccording to cognitivelead theory.
References
[1] Atkln~n, R. and Shiffx~ R. Human memory:. A proposed system and its control processes, in The psychology of learning and motivation, K. Spence and J. Spence, Editors. (1968), Academic Press.
[5]
[2] Ayers, P. Why goal-free problems can facilitate learning. Contemporary Educational Psychology, 18 (1993). 376-381. [3] Baddeley, A. The magical number seven: still magical after all these years? Psychological Review, 101 (1994). 353-356. [4] Carroll, J.M. The Nurnberg Funnel. Designing minimalist instruction for practical and computer skill. (1990), The MIT Press.
[6]
[7]
[8]
[9]
239
Cerpa, N., Chandler, P., and Sweller, J. Some conditions under which imegrated computer-based lr~mln.~ software can facilitate learning. Journal of Educational Computing Research, 15 (1996). 345367. Chandler, P., et al. Applying cognitive psychology principles to education and t~t,;-~, in JARE98 Annual Conference. (1998). Adelaide: AARE. Chandler, P. and Sweller, J. Cognitive load theory and the format of instruction. Cognition and Instruction, 8 (1991). 293-332. Chandler, P. and Sweller, J. The spilt-attention effect as a factor in the design of instruction. British Journal of Psychology, 62 (1992). 233-246. Chandler, P. and Sweller, J. Cognitive load while learning to use a computer program. Applied Cognitive Psychology, 10 (1996). 151 - 170.
[23] Stoltz~, E_R., Hasher, L., and Zacks, R.T. Working memory and aging: current status of the inhibitory view, in Working memory and human cognition, T.E. Richardson, et aL, Editors. (1996), Oxford University Press: 66-88. [24] Sweller, J. Instructional design in technical areas. (1999), ACER Press. [25] Sweller, J. and Chandler, p. Why some material is difficult to learn. Cognition and Instruction 12 (1994). 185-233. [26] Sweller, J. and Cooper, G.A. The use of worked ex~mples as a substitute for problem solving in learning algebra. Cognition and lnstruction, 2 (1985). 59-89. [27] Sweller, J. and Levine, M. Effects of goal specificity on hEarts-ends anulysis and hmrning. Journal of Experimental Psychology: Learning, Memory and Cognition, 8 (1982). 463-474. [28] Sweller, J., van Mcrri~nbocr, JJ.G., and Paas, F.G.W.C. Cognitive architectm'e and instructional design. Educational Psychology Review, 10 (1998). 251-296. [29] T a r ~ i . IL and Sweller, J. Guidance during mathematical problem solving. Journal o f Experimental Psychology, 80 (1988). 424--436. [30] Tindall-Ford, S., Chandler, P., and Sweller, J. When two sensory modes are better than one. Journal of Experimental Psychology: Applied, 3 (1997). 257287. [31] Tuovinen, J.E. The cognitive load of discovery learnln~ in School of Education Studies, Facul~, of Arts. (1999), University of New South Wales: 549. [32] Tuovinen, J.E. Research fi-amework and implications for online nmltim~lia education prance based on cognition research, in Communications and networking in education: Learning in a networked society Conference. (1999). Aldunk% I-Iggl~lmn~ Finland: [F[P. [33] Tuovinen, J.E. and Sweller, J. A comparison of cognitive load associamd with discovery lcarnlns worked examples. Journal of Educational Psychology, 91 (1999). 334-341. [34] Van Merrionboer, J.J.G. Strategies for programming instruction in high school: program completion vs. program generation. Journal of Educational Computing Research, 6 (1990). 265-285. [35] Van Merri&abocr, J.J.G. Teaching introductory computer progrh.,~min~. A perspective fi'om instructional technology. (1990), University of
[10] Chase, W.G. and Simon, H.A. The mind's eye in chess, in Visual information processing, W.G. Chase, Editor. (1973), Academic Press: 215-281. [11] Chi, M.T.H., Glaser, R., and Rees, E. Expertise in problem solving, in Advances in the Psychology of Human Intelligence, R.J. Sternberg, Editor. (1982), Lawrence Erlbaum: 7-75. [12] Jelsma, O., van Merri~nboer, JJ.G., and Bijlstra, J.P. the ADAPT design model: towards instructional control of transfer. Instructional Science, 19 (1990). 89-120. [13] Logic, R.H. The seven ages of working memory, in Working memory and human cognition, J.T.E. Richardson, et al.,Editors.(1996), Oxford University Press: 31-65. [14] Miller, G.A. The magical mnnher, seven, plus or mimls two: some lift613 on our capacity for p r o c ~ g information. Psychological Review, 63 (1956). 81-97. [15] Mousavi, S., Low, R., and Sweller, J. Reducing cognitive load by mixing auditory and visual pr~umtation modes. Journal of Educational Psychology, 87 (1995). 319-334. [16] Ollemhaw, A., Aidman, E., and Kidd, O. Is an illustration always worth ten thousand words? Effects of prior knowledge, leami,~ style and multimedia illustrations on text comprehension. International Journal of lnstructional Media, 24 (1997). 227-238. [17] Peas, F.G.W.C. Tr~inln~ strategies for attaining transfer of problem-solving .~kill in ~ti~ics: a cognitive-load approach. Journal of Educational Psychology, 84 (1992). 429-434. [18] Paas, F.G.W.C. and Van Merri~aboer, JJ.G. Variability of worked examples and transfer of geom~a'ical problmn-solving skills: A cognitive-load approach. Journal of Educational Psychology, 86 (1994). 122-133. [19] Quilici, J.L. and Mayer, R.E. Role of examples in how students learn to categorize sta~tics word problems. Journal of Educational Psychology, 88 (1996). 1 ~ 161. [20] Schneider, W. and Shiffrin, R.M. Controlled and automatic humsm in.formation processing: I. Detection, Search, and Attention. Psychological Review, 84 (1977). 1-66. [21] Shiffi'in,R.M. and Schneider, W. Controlled and automatic human information processing: H. Perceptual learning, automatic attending, and a general theory. Psychological Review, 84 (1977). 127-190. [22] Simon, H.A. and Gilmurtin, K. A simulation of memory for chess positions. Cognitive Psychology, 5 (1973). 29-46.
Twine. [36] Van Merfienboer, J.J.G. a,d De Croock, M.B.M. Strategies for computer-based progr~ ...... ;-£ instruction: program completion vs. program
240
[37]
[38]
[39]
[40]
generation. Journal of Educational Computing Research, 8 (1992). 365-394. Van Merti~nboer, J.J.G., K r a f t , H.P.M., and Maaswinke], R.M. Automating the planning and constzuction of progr~ ...... in~ assi~nm~ms for teachi-~ introductory computer progrRrra'nhlg, ill Automating instructional design, development, and delivery, R.D. Tennyson, Editor. (1994), SpringerVerlag: 61-77. Van Merri~nboer, J.J.G. and Luutesma, J.J. Tmplemenling instructional models in computer-based learning environments: a case study in problem selection, in Advanced educational technology: research issues and .future potential, T.T. Liao, Editor. (1996), Springer-Verlag: 184-206. Van Merri~nboer, JJ.G. and Paas, F.G.W.C. Automation and schema acquisition in learning elementary computer programming: implications for the design of practice. Computers in Human Behaviour, 6 (1990). 273-289. Ward, M. and Sweller, J. Stracturing effective worked examples. Cognition and lnstruction, 7 (1990). 1-39.
241
