As published in Computers and Education 1995. Draft. Do not quote this (but feel free to quote published article)
Children as computer users: the case of collaborative learning
Department of Human Sciences, Loughborough University
WIthout doubt, research into humam-computer interaction (HCI) hasprospered in recent years. How to make computer systems accessibleand attractive to the people who use them has become a topic of practical and theoretical interest. Yet it is surprising that thislively research tradition has been so preoccupied with the use ofcomputers by adults: there has been little interest in design issuesthat arise when the "users" are children. The aim of the presentpaper is to explore this shortfall in relation to one particular HCIchallenge: namely, the effective design of computing resources thatsupport collaborative learning. Studies of the user-as-collaboratorillustrate the HCI bias I have identified: the topic certainly hasreceived attention, but it is adults in their workplace that have beenmost closely considered. So, there is now a substantial literature onhow to design systems for computer-supported collaborative working[1,2]. More recently, HCI researchers have considered"computer-supported collaborative learning" . This is atopic that might foreground the needs of younger users. In fact,it has not: the dominant research themes being design for the moreadult communities of college students and distance learners [4,5].
Evidently, this neglect of children as computer users arises fromtheir less significant influence within a market place of softwareproducts. However, it may also reflect that fact that few HCIresearchers will have enjoyed an academic grounding in the relevantareas of developmental or educational psychology. Their theoreticalframeworks are more likely to derive from the the adult-orientedtraditions of cognitive science. In what follows here, I shall reviewsome findings from those areas of psychology that do concern children. In particular, I shall consider findings that serve to locatechildrens' status as collaborative learners. Computers may wellbecome a powerful resource whereby teachers may supportmore classroom. My purpose is to contextualise any ambitions of thiskind by suggesting how research in developmental psychology can informthe design enterprise. Firstly, I shall identify the generalsignificance of collaborative structures for early learning byreference to research on educational practice and by reference totheories of early cognitive development. This will demonstrate thatcollaboration in classroom settings is highly valued but hard toignite and sustain. Secondly, I shall interpret this tension throughan analysis of the psychological processes involved in a collaborativeexchange. Finally, I shall identify implications for designers -drawing upon some pointers arising from current classroom research inthis area.
The place of collaborative experiences in early learning
Recently there has been extensive research on early learning that isorganised to take advantage of interaction among peers. Partly, thisinterest follows from claims made within influential theories ofcognitive development. Piaget supposed that young children benefitfrom peer-based learning because a natural egocentrism necessarilygets challenged - through 'the shock of our thought coming intocontact with that of others' [6, p. 204]. Moreover, the influentialsocio-cultural theorising inspired by Vygotsky's writings stresses thecentral place of participatory learning - which must be managed withinsocial interaction. Cognitive development in Vygotsky's zone ofproximal development depends upon either 'adult guidance' or'collaboration with a more capable peer' [7, p. 86].
Such theorising has inspired research that evaluates peer-basedlearning in terms of its outcomes. "Outcomes" here may mean thequality of some product arising from learners' joint (as opposed tosolitary) activity. Or it may mean the relative performance ofindividuals on tests that follow a period of them learning eithercollaboratively, or alone. Reviews of such research indicate adistinct outcome advantage for peer-based conditions [8,9]. Somestudies have examined interaction during group work in order toexplore correlations between interactional styles and qualities ofoutcome. Evidently, such research could furnish insights into theprocesses that mediate an advantage for working collaboratively. Three lines of interpretation have emerged. The first dwells upon thefact that collaborators will usually *articulate* their thoughtspublically [10,11]. This may be advantageous 'because it forces thesubjects to bring to consciousness the ideas that they are justbeginning to grasp intuitively' [10, p. 152] - thereby making suchideas potential objects-for-exploration. The second line ofinterpretation stresses the productive value of *conflict* that canarise as partners negotiate a consensus . The thirdinterpretation stresses the possibility of *co-constructions* withincollaborative problem solving. Working with a peer may support acreative process of converging upon a single, systematising object - ahypothesis, a prediction, a model, or whatever .
Within the same period of this research, educational practice hasdeveloped in complementary ways. Teachers have been encouraged tofoster more collaborative structures for learning. The verylayout and furnishing of most primary school classrooms stronglyaffords this style of working. Indeed collaborative work is astatutary requirment in some areas of the UK National Curriculum. McMahon  has summarised official documents that promoted thisinterest in Britain. Although such developments occured in parallelwith developments in psychological theory, it is not clear that theorywas the driving force. For the British experence, Galton andWilliamson  find the origins of enthusiasm for group workelsewhere: in a more basic commitment to mixed ability classroomswhich, in turn, arose from the politically-managed breakdown of ahighly-streamed educational system. In North America, a similarexternal pressure to support group work seems to have arisen from adetermination to sustain and succeed with ethnically diverse classroomcommunities. This fostered influential programmes of interventionresearch - particularly on "cooperative learning". Such researchtypically considers either optimal strategies for an over-archingclassroom organisation , or optimal strategies for rewarding smallgroups . In both cases, it seems that cooperative classroomstructures are very effective in terms of both learning outcomes andpupil engagement .
Thus, both theories of cognitive development and researchon classroom practice tends to favour arrangements for learning thatentail peer interactions. Yet, there are other research outcomesthat do not sit easily with these conclusions. For example, atension is evident between what has been said above and the findingsfrom some interview studies of how pupils perceive schooling. Manyyoung pupils do not regard the circumstances of group work verypositively [19,20]. Naturalistic studies of small group work tend toreinforce this observation. Thus, ethnographies of classroom life inearly education reveal that effective pupil collaboration isstrikingly rare [15, 20, 21]. It is as if the seating has beensocialised into collaboration, while the pupils have not.
So, there is a discrepancy between positive outcomes from empirical studiesof collaboration, and the limited spontaneous collaboration observedwithin classroom group work. This discrepancy is partly clarified byconsideraing a subset of the research discussed above: studies ofcooperative learning. That research suggests that collaborationrequires explicit orchestration. These classroom programmes impose aformal structure on occasions of learning: such organisation appearsto be necessary if group-based work is to be productive. A supportingstructure might be created in various ways. For example, managementof incentives for group participants is one approach . Anotherarises from patterns for decomposing and distributing a problem taskamong the members of different cooperative learning groups - such asoccurs in the 'Jigsaw Classroom' . However, I wish to direct thepresent discussion towards a further kind of structuring: one thatarises from the inherent format in which a problem is presented tocollaborators - the *resources* it provides for them. After all, wemay not always be able, or inclined, to manage the rewards associatedwith problem solving, or to manage the way in which work is parcelledout. So, it will be necessary to understand more of how the formatand materials of a task constrain or facilitate collaborativeengagement around it. In short, I wish to argue that children'ssuccess as collaborative learners depends a lot on the character ofthe resources at hand to mediate the interaction. Potentially,computers are an intriguing class of resource in this context. However, this argument first requires some consideration of thepsychological processes we suppose underpin a collaborativeengagement: I shall consider this in the following section.
The social dynamic of collaborating
It was quite natural that outcome-oriented research should dominatethe literature in this area. It represents a straightforward effortto determine if learning through collaborating is "a good thing". Apre-test/intervention/post-test research design defines a natural approach to that question. It turns out that interventions comparingindividuals who have studied something in groups with inviduals whowere solitary learners usually show an outcome advantage for theformer. Under controlled observations, the socially-organisedlearning generally works better. In the face of these findings, it isthen natural to co-relate patterns of interaction within such groupswith variations in the degree of advantage they enjoy. Briefly, thisallows us to ask: given some group learning, what are theinteractional predictors of good (individual) performance atpost-testing? This second research move has usually entailed seekingcorrelations between post-test outcome measures and codings ofparticipants' speech during the collaborative intervention. Webb'sstudies  are a model of this approach and there is no doubt thatit can be informative. However, it has served to cultivate too narrowa perspective on the experience of collaborating.
Firstly, the design sketched above tends to equate "collaboration"with "putting into small groups". This operational definition of thephenemonon is convenient for research design but tends to deflect usfrom asking the question: "in this small group, did collaborationoccur?" If such a question is considered at all it will be byreference to quantitative surveys of certain kinds of speech. Moreof certain classes of utterances equates with "more collaboration"This identifies a second characteristic of the classical researchdesign: defining differences in collaborative activity through thecategorisation and enumeration of various speech acts. Finally, thereis a third factor apparent within this research tradition that mayfurther mitigate against addressing the real social dynamic ofcollaborating. There is too narrow a concern for conceptualisingsocially-organised learning in terms of its potential for moulding*individual* (cognitive) structures. Even supporters of Vygotsky'ssocio-cultural approach to cognition tend to frame it as a theory within which 'individual structures are actually formed by socialinteraction' [24, p.224]. For many purposes this may provide ahelpful model, but not if it distracts us from attending directly tothose *social* structures that arise within joint activity.
In sum, I am arguing that exposing the social dynamic of "learningtogether" has been neglected because of at least three researchpreoccupations: (i) defining a collaborative situation in merelyoperational terms, (ii) characterising the participants' activity interms of speech act catalogues, and (iii) intepreting the significanceof these activities exclusively in terms of impacts upon individualpsychological functioning. How could it be otherwise? A startingpoint would be to address more seriously the implicit core of anycollaborative effort - namely, the creation of some knowledge basethat comes to be shared by the collaborators. Analsysis ofcollaborative occasions needs to dwell upon the discursive moveswhereby this sharing is achieved. Certainly, such matters have beenconsidered in other areas of psychological research. Thus, theconstruction of common knowledge is a key concern in studies ofroutine conversation management  and also in studies of formalinstruction. Yet only a very few studies of collaborative learningtake seriously how participants reveal and manage an investment increating mutual understanding [26, 27]: how their talk makes thishappen and reveals a degree of determination that it should happen..
The theoretical concept of "co-construction" comes closest to helpingwith these issues. Yet, typically the objects of co-constructiveactivity are identified in terms of certain abstractions in theproblem solving process - an hypothesis, a model, a prediction etc. Such strategic achievements are certainly likely to be an importantpart of what might be valuable in a collaboration, and their socialconstruction is likely to be part of what we might want to study. However, they will usually represent the particular outcomes of a moresustained social process - namely, talking and acting in ways thatreveal problem-solvers' concern for coming to "have things in common". Outside of psychological research, this is just what professionalcollaborators seem to identify in what they do . More casually,it also matches our intuitive sense that the term "collaborating" doesrefer to some more substantical state of *engagement* (and not merelyto a collection of speech acts, or the convergence on some formalproblem-solving abstraction).
There are further grounds for suspecting that collaborative learningwill need to be understood with more reference to the creation of thismutual engagement: the current style of analysis has notclearly answered the questions that are most urgent for practitioners. For example, the traditional analysis of collaborative interactionsgenerates a confusing mixture of findings for the effects of groupcomposition variables. Teachers in particular have been interested inwhat research can reveal about the relevance of gender and ability forthe formation of classroom groups. Yet studies in which thecomposition of groups have been manipulated with attention to thesevariables provide contradictory results [25, 29, 30]. Suchdiscrepencies suggest there may be a volatility associated with eventsthat arise within group work. Moreover, this may be particularly thecase for young children. Children may less readily suspend theircurrent interpersonal attitudes and their current enthusiasms in theinterest of wholeheartedly "being collaborators" (for the transientpurposes of teachers or researchers). Thus, the success of anyparticular occasion of collaborating may depend on rather fluidfriendship patterns, on momentary classroom tensions and, perhaps,upon pupils' social construction of the intervention or experiment. These are contextual factors that are hard to capture. Unfortunately,they obstruct strong generalisations about good practice (eg. inrelation to gender or ability). Perhaps we should not be surprised tofind that apparently minor variations in experimental proceduresometimes can lead to differing conclusions about the efficacy ofgroup work . Under these circumstances, how might the concept of co-constructionhelp clarify the social dynamic of collaboration - particular whereyoung children are participants? I would suggest that the concept hasto be unpacked to refer to the following: how collaborators reveal aconcern for achieving mutual knowledge, and how they realise thisinvestment through deploying particular discursive devices. What gets"constructed" is a certain state of mutuality; this achievement maythen be explored by researchers in terms of its interpersonalcorrelates and their relation to material or institutional contexts. More particularly, I would propose three features of socialinteraction that are central to the creation of collaborativeengagement.
FIrstly, productive collaborations may be characterised by a sense ofintimacy that arises in situations where we are aware of communualpurpose. Anderson  touches on this theme, but at a more macrolevel: noting the potency of "imagined community" as it arises fromour shared indulgence in the institutions of cultural life (newsstories, soap operas etc.). Krauss and Fussell  make the point inrelation to our routine social category memberships (as in being 'aNew Yorker'). Oddly, the affective tone of shared experience hasrecieved little research attention at the more micro level - as itmight arise in the bounded circumstances of creating a distinctivehistory of experiences within a small collaborating group. Yet onecreative force driving a collaboration may be a certain sense ofcommunity that emerges within the overlapping experiences of solvingsome problem. This point makes contact with the contemporary interest in relating our human ability to learn from collaboration to ourhuman capacity for intersubjectivity - projecting states of mind intoother people. Collaborating is not simply having some set ofexperiences or understandings in common, it is knowing that this stateof mutuality exists. Mobilising shared experience into a genuinecollaboration depends upon such projective capabilities; it dependsupon knowing (recursively) that the other knows what you know. However, the particular point I am foregrounding here is thatexploiting such mutuality may have an affective dimension - as well ascarrying a cognitive significance for the problem solving enterprise. There may be something motivating about a certain sense of harmony andintimacy that can arise when pursuing a collaborative encounter. Inrelation to the challenge of designing for collaborators, thisencourages us to fashion problem solving situations such thatparticipants have a sharper sense of jointly owning a common anddistinctive set of experiences.
Secondly, the chances of creating a productive situation of commonknowledge is likely to be improved by a rich supply of externalresources. In addressing some problem space, collaborators willbenefit from vivid and accessible referential anchors. The moreabstract the terms of the problem, the more helpful it may prove tohave external representations that resource the construction of ashared understanding . If those representations permitmanipulation - if they exist in some interactive media (such ascomputers) - then the potential for jointly witnessing the behaviourof such representations will doubtless enrich the sense of sharedexperience. It is recent observations of computer-supportedcollaborating that have brought this possibility into focus[27, 36].
Thirdly, productive collaborations are likely to depend upon thequality of interpersonal relations already in place at the time somenovel collaborative encounter is initiated. In short, participantsusually will bring to these exchanges histories of joint activityconducted on other occasions and at other times. Indeed there isempirical evidence to suggest that friendship status is a significantfactor in dictating the character of collaborative problem solving. Yet, as Pozzi, Healy and Hoyles comment '..the fact that pupilsknow each other, have likes and dislikes of each other and haveexpectations of each other and themselves is rarely considered in theresearch' [24, p.239]. The report of these researchers pays closerattention to such dynamics. They describe group interactions thatseemed to have a rather serendipitous quality: suggesting a stronginfluence from existing relationhships. In fact, such an influencemay have overriden any effects from traditional experimental variablesalso present in their research design - gender and ability differencesin particular.
If we take these distinctions seriously, then there are implicationsfor how we study occasions of collaborative learning. They musteach be evaluated as a more or less successful enterprise of creatingcommon knowledge. Analysing some potential "collaboration" willinolve considering whether there has been an investment of this kind,how it was managed, and whether it was creatively exploited. "Exploitation" of the investment is of special concern: it is whatwill underpin those occasions where (perhaps in post-testing designs)we find favourable learning outcomes. For the generative nature ofcommon knowledge arises from it serving as a platform forcollaborators to reason publically about some problem in increasinglymore intimate and careful ways. The more elaborate the backlog ofcommon knowledge, the more effective and the more motivated thereasoning.
In summary, this discussion encourages a particular strategy for theanalysis of collaborative exchanges. It requires documenting anydiscursive effort that indicates an active concern among thecollaborators for creating common knowledge. Any effort of that kindis not so readily abstracted from the traditional coding of discreetutterances. It is a more extended social achievement: demandingthat researchers attend to the collaborative discourse at a lessmolecular level. When this is done, it becomes quite possible thattwo collaborating groups get evaluated very differently - even while amore traditional coding of utterances conceals any difference. Thatis, in terms of the pragmatics of their conversation (the occurrencein speech of predictions, challenges, hypotheses etc), two groupsmight appear very similar; while in terms of their intersubjectiveattitudes they might be characterised as very different. For example,equally animated partners may be differ in intersubjectiveterms: they may be variously competitive, or collaborative, or simplyoriented towards finishing the task as quickly as possible [27,157-175]. Such distinctions touch on a dimension of interactionswithin collaborative learning that traditional analyses overlook. This observation reinforces a point made earlier: the impact on grouplearning of traditional social psychological variables (gender,ability etc) may often be overriden by variations in whether there wasreal collaborative engagement, in the sense discussed here. Suchvariation may be traced to rather slippery aspects of the learningcontext - friendships, social representations of a task, momentaryenthusiasms or attitudes, and so on. Our best approach may not be totry and predict or manage all these contextual features (although wemay usefully document the general class of things that can exert suchinfluence). Rather, we might do better to concentrate on optimsingthe material situations in which pupils are encouraged to learncollaboratively. If the circumstances of collaboration are well*resourced*, then it is this that may optimise the chances that acollaborative attitude will be fostered. Defining such optimisationmay be our best research strategy - one which may be well served bynew technology, as I shall argue in the final section of this paper. First, however, the analysis developed above can be usefully directedat an important question - the answer to which dictates whether thismatter of young children collaborating at computers is worth pursuingat all: "Can collaborative learning ever work with children duringtheir early stages of cognitive development?"
Can younger children learn collaboratively?
Some research traditions that make a case case for promotoingcollaborative learning have also been mobilised to argue that youngerchildren find learning this way intrinsically difficult. Thus,Piaget's theorising supposes cognitive change occurs within peerinteraction but it also suggests that, during the early school years,children lack the necessary socio-cognitive resources to learneffectively in collaborative arrangements. In addition, naturalisticobservations of classroom group work reinforce this belief: they showthat productive interaction in small groups is very infrequent.
It is not unusual for commentators on this situation to suggest thatchildren lack certain psychological "skills" - cognitive prerequisitesfor effective collaboration. For example, Dauite commments on children's limited performance during collaborative writing:'...although children enjoy working together, they may not have thesocial or cognitive prowess to collaborate effectively at first'[38, p.47]. Gauvain and Rogoff describe young children's efforts insimilar terms: 'Perhaps children as young as 5 years of age are notskilled at collaborating even when told to do so and even when workingwith an adult' [39, p.148]. Reference to the developmental status ofsome cognitive "skill" is a popular theoretical move for manypsychologists. For example, this move might be made in relation tothe recent interest in children's developing theories of mind. Brownell and Carriger comment: 'As they get older, children becomebetter able to comprehend the behavior and intentions of another, aswell as better able to affect the other's behavior and to communicateabout their own behavior and desires... age-related social andcognitive skills contributing to peer collaborations influence whatand how children learn from these collaborations' [40, p. 381]. Thisperspective is ellaborated by Tomasello and colleagues. They assertthat younger children can not partake of collaborative learning in anyproper sense, because they await the emergence of certainsocio-cognitive skills: 'Collaborative interactions of the typeperformed by preschoolers thus do not constitute collaborativelearning as we define it' [34, p.501].
Such prescriptions are unfortunate. There may be discontinuities incognitive development that are relevant to the management ofcollaborative learning. But they should not distract us fromrecognising that there are also important *continuities*. Onesuch fundamental continuity arising within early joint activity isthe human social concern for joint reference. It was implied in theprevious section that any analysis of collaborative problem solvingshould foreground how collaborators purposefully create and elaboratemutual reference. Their concern to do this describes the core ofjoint thinking. It is this that may define for theorists an importantkind of continuity beneath the varying *conditions* under whichcollaborative problem solving gets managed during development. Certainly, the registration of joint reference is a pervasiveachievement that can be traced back to the very earliest of childhoodsocial experiencess. Within the very first year of life, there isevident a lively concern for managing mutual attention to theimmediate environment [41, 42]. Trevarthan terms this 'secondaryintersubjectivity'. It signals a capacity and a motivation tocoordinate with other minds. It might fairly be claimed that: '...only humans have the kind of appetite a one-year old begins to showfor sharing the arbitrary use of tools, places, manners andexperiences' [43, p. 55]
I suggest that it is useful to begin tracing the natural history ofcollaborative learning from this precocious appetite for socialengagement. As Trevarthan describes this early period: '..it seeschildren starting cognitive learning in a co-operative and imitiativerelationship to other more experienced companions and activelycontributing to the propagation of collective knowledge' [43, p.37]. Young children's interest in establishing mutual knowledge seems verystrong. It might even be reasonable to reverse the force ofthe question heading up this section - asking instead if there is verymuch early learning that is *not* collaborative. The child's masteryof language is the most remarkable achievement of early learning; yetthe manner in which the environment supports this achievement suggestsa strongly collaborative relationship between children and adultnative speakers . Such observations touch on the question ofwhether we are "basically" cooperative or competitive: a developmentalissue that enjoys a long history of discussion . Yet it may bebest disposed of by assuming that what we are "basically" is buildersof intersubjectivity: our concern to predict and understand othersdemands projective theorising about their mental lives and, withinthat process, we create mutualities. However, this is not to answerone way or the other in the particular debate about being acooperative versus a competitive species. Because the"intersubjective attitude"  can be mobilised into the service ofcolloboration or into the service of competition. Culturalcircumstances may vary in this respect: creating differentdevelopmental trajectories for the basic achievement of mutualknowing.
This same point about the fomrative role of culture can be applied toour concern over whether young children can be collaboborativelearners. It is important to remember that the reason for raisingdoubts about their potential as collaborators arises from observationsof what goes on in classrooms. But if we glance into a school*playground* (perhaps during the breaks in our research observations),we soon notice children actively managing collaborative routines. So,the challenge for research is how to understand a certaindiscontinuity of functioning. In the "playground" children mightmanage reciprocity very effectively; faced with the agenda of theirclassroom they may be much less successful and, thus, not veryeffective collaboraters. We can consider whether new technology canhelp us tackle this discontinuity.
Designing for collaborative learners in early education
I am suggesting that the way forward in this discussion is toconsider more carefully the contexts in which collaborations areorganised. Their fate as social interactions may have much to do withthe *resources* that are available to mediate the exchange. Researchthat has looked closely at children's more playful collaborationssuggests two particularly potent forms of such mediation. Thefirst is a jointly visible and manipulable array of playthings. So,small groupings of children as young as 20 months have been shown todisplay an active concern for developing shared reference in relationto a set of interesting objects accessible between them . Thesecond mediational resource is narrative. In the context ofrole-playing games, preschoolers can be very effective in managingmutual knowledge ; by the time they are of school age they canmanage such exchanges over very extended periods .
Of course, typically, each of these resources are well represented inthe contexts of young children's play. During play, children oftenwill be oriented towards making, controlling or exchanging *things*;they often will be exploring the possibilities of scripted pretense. Such support is less obviously central to the experience of being inclassrooms. It is true that popular images of primary classroomssuggest an active environment with much interest in material things -yet much of childrens' work will still occur in abstractrepresentational formats. This is exactly the obstacle identified byBennett  in his consideration of the meagre collaborationsobserved within small groups of primary pupils. This may be becausemuch of what schooling is about is confronting representationalformats that are inherently abstract (eg. mathematics), or becauselocating a problem in some narrative format demands too elaborate aninvestment in classroom organisation.
It is surely at this point that new technology can offer some supportfor what teachers will have to achieve. In particular, computersoffer a medium in which abstract material can be rendered accessiblethrough creating visible and manipulable representations. It is clearthat the technology has a considerable potential for externalisingmodels of complex systems in this manner. It is also clear thatcollaborative learning is all the more effective when the participantsare able to orient towards such referential anchors and jointlywitness the effects of manipulations upon them [36, 27]. Is itpossible to be more explicit: to specify more about design principlesfor software that is to mediate pupils' collaborative learning?
Simple rules of thumb are unlikely to be forthcoming, butclassroom research can help identify the broad themes that designersmay need to keep in mind as they prototype ideas. I shallillustrate this by citing two examples of such themes that arose in myown observations of computer-supported collaborations inprimary schools [27, 49]. The first is a pervasive tendency amongyoung children for appropriating tasks to a *turn-taking* structure ofinteraction. Much software for early education allows this. Forexample, control of a single input device may encourage it  orit may be adopted because tasks are often encountered as sequences ofinitiate-respond-evaluate packages . There may be no simple rulesfor circumventing this, but sensitivity towards it as an issue mayencourage more promising design decisions [27 p. 183-187]. My secondexample concerns the design of interfaces to activities where progressdepends upon careful reading of screens. This demand is paticularlycharacteristic of adventure-format problems that are popular inprimary schools: here written language carries an (unusal)instrumental role in directing the decisions that users mustnegotiate. Some of my own observations of younger children usingsuch software drew attention to how small differences in readingability could serve to marginalise slower members of a small workinggroups.
Such examples may serve to bring into focus aspects of acomputer-supported collaboration that need to be carefully managed -at least, if the interaction is to be a productive engagement of thesort defined in this paper. Yet these examples also remind us thatcontrbutions of effective HCI "design" are not exclusively a matter ofspecifying what happens on the screen or at the input device. So ifturn-taking is an issue that is strongly marked among young childrenas users, this is not necessarily tackled at the level of traditionalHCI interface concerns. For the problem has as much to do with stylesof work or assessment in the educational contexts within whichcomputer-using pupils have been socialised.
This need not seriously restrict HCI researchers. However, it doessuggest that they must approach the design enterprise within a broadresearch framework. Their interest must embrace an understanding ofthe culture of learning within which computer artefacts may bedesigned to function. As it hapens, the nature of the tools they maydevelop for the support of collaboration increasingly will draw themcloser to the social and institutional contexts of studentcollaborators. This may not be so evident if the circumstances of"collaborating" are always the narrow variety of occasion we think offirst - that is, situations where two or more pupils are jointlyacting *at* some computer terminal. Settings for organised teachingand learning do involve a more extended and richer sense ofcollaboration than this. Primary schools work by fostering smallcommunities (of pupils and teacher) that make their progress bycultivating structures of common knowledge . "Collaborativelearning" is a phenomenon manifest in the diffuse and extendedpatterns of interaction that go to make up this community. Suchlearning is not only and simply what happens when two or more peoplesit together and interact at a particular problem-solving location. Designing for collaboration now goes beyond designing for pairs ofpupils briefly congregated at a specified classroom location to solvea circumscribed problem. It extends beyond this to service the moreloosely coupled interactions that go to make up communities ofeducational practice.
Some have already documented their experience of deploying newtechnology to empower the broader collaborative atmosphere of earlyeducation. The most discussed examples arise from efforts to usewide-area computer networking to draw schools into patterns ofcross-site communication [52, 53]. This is an intriguing enterprisealthough research around it has not, so far, generated a great deal ofinsight as to how computer resources are best designed for children. Moreover, it would be unfortunate if our model for supporting thecommunal quality of education was limited to this format oflong-distance, asynchronous communication. There are other modelsthat forcus upon computer-supported communicaiton managed at a morelocal level .
At the start of this discussion, I noted the lack of theorising inrelation to children as computer-users. Nevertheless, the trends ineducational computing resources have reflected well the preoccupationsof developmental and educational theorists. The tradition ofintelligent tutoring systems has drawn from theorising withinArtificial Intelligence. The complementary tradition of computermicroworlds (Turtle Logo in particular) has drawn from Piagetian,construcitivst thinking. An interest in meeting the needs oflearners-as-collaborators reflects the emergence of a thirdtheoretical strand: namely, socio-cultural perspectives on developmentand learning. At first this socio-cultural influence was expressedthrough the promotion of computers as locations which might supportcollaboration in the sense of short-term, small group problem solving. In the terminolgy of Vygotsky, we might say that computers wereallowed to enter zones of proximal development - as they were createdamong co-working peers . More recently, socio-cultural theorisinghas shifted towards an emphasis on athe role of institutional andcommunity structures, rather than the more intimate interpersonalformats of interacting *at* computers. A recent symposium on newtechnology in education  reveals a strong sense of researchersshifting their interest towards the design of more integrated"learning environments".
At the leading edge of these developments we find a number of venturesin which computers are being utilised to capture and extend thecommunity quality of educational practice [57, 58, 59]. There arereal issues of design to be addressed here but they will be informedby research that is oriented to the workings of communities - ratherthan to the traditional psychological concern with thedecontextualised user-computer exchange. In the longer term, somebelieve that these developments will support the "deschooling" ofsociety predicted by Illich . However, it is unlikely thatdesigners will create compelling "virtual communities" in educationalcontexts - at least, not before we are more generally comfortable withsuch patterns of communication in other areas of our lives . Inthe meantime the challenge is to understand more of the communalnature of educational practice in order that new technology maysupport rather than undermine it.
1. Bowers, C.A. The cultural dimension of educational computing. New York: Teachers college press (1988).
2. Grief, I. Computer-supported cooperative work: a book of readings. San Mateo, Ca.: Morgan Kaufmann (1988).
3. O'Malley, C. Computer-supported cooperative learning. Berlin:Springer-Verlag (1994).
4. Mason, R. and Kaye, A. Mindweave: Communication, computers and distance education. Oxford: Pergamon Press (1989).
5. Harasim, L. On line education: Perspectives on a new environment. New York: Praeger (1990).
6. Piaget, J. Judgement and reasoning in the child. London:Routledge and Kegan Paul (1928).
7. Vygotsky, L.S. Mind in society: The development of higherpsychological processes. Cambridge, Ma.: Harvard University Press (1978).(Original material published in 1930, 1933, and 1935).
8. Azmitia, M. and Perlmutter, M.. Social influences onchildren's cognition: State of the art and future directions. In H.Reese (Ed.) Advances in child development and behavior (Vol 22). NewYork: Academic Press (1989).
9. Damon, W. and Phelps, E. Strategic uses of peer learning in children's education. In T. Berndt and G. Ladd (Ed.), Peer relationships in Child Development. New York: Wiley (1989).
10. Hoyles, C. What is the point of group discussion in mathematics? Studies in mathematics, 16, 205-214 (1985).
11. Schunk, D. Verbalisation and Children's Self Regulated Learning. Contemporary Educational Psychology, 11, 347-369 (1986).
12. Doise, W. and Mugny, G. The social development of the intellect. Oxford: Pergamon Press (1984).
13. Forman, E. The role of peer interaction in the social construction of mathematical knowledge. International Journal of Educational Research, 13, 55-69 (1989).
14. McMahon, H. Collaborating with computers. Journal of Computer-assisted Learning, 6, 149-167 (1990).
15. Galton, M. and Williamson, J. Group work in the primary classroom. London: Routledge (1992).
16. Johnson, D. and Johnson, R. Motivational processes in cooperative, competitive, and individualistic learning situations. In C. Ames and R. Ames (Ed.), Research on motivation in education, Vol 2. New York: Academic Press (1985)
17. Slavin, R. Cooperative learning; engineering social psychology in the classroom. In R. Feldman (Ed.), The social psychology of education. Cambridge: CUP (1986).
18. Nastasi, B.K. and Clements, D.H. Social-cognitive behavioursand higher-order thinking in educational computer environments. Learning and Instruction, 2, 215-238 (1992).
19. Cullingford, C. The inner world of the school: children's ideasabout schools. London: Cassell (1991).
20. Galton, M. Grouping and group work. In C. Rogers and P. Kutnick (Ed.), The social psychology of the primary school. London: Routledge (1990).
21. Bennett, S.N. Cooperative learning in classrooms: Processes andoutcomes. Journal of Child Psychology and Psychiatry, 32, 581- (1991)
22. Aronson, E., Bridgman, D.L. & Gellner, R. The effects of acooperative classroom structure on student behavior and attitude. InD. Bar-Tal & A. Saze (Eds.), Social psychology of education: Theoryand practice. New York: Wiley (1978).
23. Webb, N. Student interaction and learning in small groups: A research summary. In Slavin, R. (Ed.), Learning to cooperate and cooperating to learn. New York: Plenum (1985).
24. Pozzi, S., Healy, L. and Hoyles, C. Learning and interaction ingroups with computers: when do abiltiy and gender matter? SocialDevelopment, 2, 222-241 (1993).
25. Clark, H.H. and Wilkes-Gibbs, D. Referring as a collaborativeprocess. Cognition, 22, 1-39 (1986).
26. Roschelle, J. Learning by collaborating: convergent conceptualchange. Journal of the Learning Sciences, 2, 235-276 (1992).
27. Crook, C.K Computers and the collaborative experience oflearning. London: Routledge. (1994).
28. Schrage, M. Shared minds. New York: Random House (1990).
29. Tudge, J. and Winterhoff, P. Can young children benefit fromcollaborative problem solving? Tracing the effects of partnercompetence and feedback, Social Development, 2, 242-259 (1993).
30. Hughes, M. and Greenhough, P. Gender and social interaction inearly LOGO use. In J. Collins, N. Estes, W. Gattis and D. Walker(Eds.) The Sixth International Conference on Technology and Education,Vol 1. Edingburgh: CEP (1989).
31. Littleton, K., Light, P.H., Joiner, R., Messer, D. and Barnes, P.Pairing and gender effects on children's computer-based learning. European Journal of Psychology of Education, 8, 311-324 (1992).
32. Anderson, B. Imagined communities. Londong: Verso (1983).
33. Krauss, R.M. and Fussell, S.R Constructing sharedcommunicative environments. In L. Resnick, J. Levine and S. Teasley(Eds.) Perspectives on socially-shared cognition. Washington, DC:American Psychological Association (1991).
34. Tomasello, M., Kruger, A.C. and Ratner, H.H. Cultural learning. Behavior and Brain Sciences, 16, 495-552 (1993).
35. Crook, C.K. On resourcing concern for collaboration within peerinteractions. Cognition and Instruction, 13, 541-547 (1995).
36. Roschelle, J. and Clancey, W.J. Learning as social and neural. Educational Psychologist, 27, 435-453 (1992).
37. Azmitia, M. and Montgomery, R.) Friendship, transactivedialogues, and the development of scientific reasoning. SocialDevelopment, 2, 202-221 (1993.
38. Daiute, C. Issues in using computers to socialize the writing process. Educational computing and technology Journal, 33, 41-50 (1985).
39. Gauvain, M. and Rogoff, B. Collaborative problem solving and children's planning skills. Developmental Psychology, 25, 139-151 (1989).
40. Brownwell, C. and Carriger, M.S. Collaborations among toddlerpeers: individual contributions to social context. In L. Resnick, J.Levine and S. Teasley (Eds.) Perspectives on socially sharedcognition. Washington DC: American Psychological Association (1991).
41. Bruner, J.S. (1983) Child's talk. Oxford: Oxford University Press (1983) .
42. Schaffer, H.R. Joint involvement episodes as context for development. In H. McGurk (Ed.), Childhood social development: Contemporary perspectives. Hove: Lawrence Earlbaum Associates (1992) .
43. Trevarthan, C. Universal cooperative motives: how infants begin to know the language and culture of their parents. In G. Jahoda and M. Lewis (Ed.), Acquring culture: cross cultural studies in child development. Beckingham, Kent: Croom Helm (1988) .
44. Pepitone, E.A. Children in cooperation and competition. Lexington, MA: Lexington Books (1980) .
45. Davidson, P.M. The role of social interaction in cognitivedevelopment: A propaedeutic. In L. Winegar and J. Valsiner (Eds.)Children's development within social context. Volume 1: Metatheory andtheory. Hillsdale, NJ: Lawrence Earlbaum Associates (1992) .
46. Verba, M. The beginnings of collaboration in peer interaction. Human Development, 37, 125-139 (1994) .
47. Goncu, A. and Kessel, F.S. Preschooler's collaborativeconstruction in planning and maintining imaginative play. International Journal of Behavioral Development, 11, 327-344 (1988) .
48. Kane, S.R. and Furth, H.G. Children constructing socialreality: A frame analysis of social pretend play. Human Development,36, 199-214 (1993) .
49. Crook, C.K. A Social Context for Classroom Computers. In J. Rutkowska and C. Crook (Ed.), Computers cognition and development. Chichester: Wiley (1987) .
50. Light,P and Glachan,M. Facilitation of Individual Problem Soliving through Peer Interaction. Educational Psychology, 5, 3&4, 217-225 (1985) .
51. Edwards, D. and Mercer, N.M. Reconstructing context: the conventionalization of classroom knowledge. Discourse processes, 12, 91-104, (1989)
52. Robinson, B. Communicating through computers in the classroom. In P. Scrimshaw (Ed.) Language, classrooms and computers. London:Routledge (1993) .
53. Keep, R. On-Line: Electronic mail in the curriculum. Coventry: National Council for Educational Technology (1991) .
54. Crook, C.K. Electronic communications in two educational settings: some theory and practice. In C. O'Malley (Ed.) Computer-supportive collaborative learning, Berlin: Springer-Verlag (1994).
55. Crook, C.K. Computers in the zone of proximal development:Issues of evaluation. Computers and Education. 17, 81-91, (1991)
56. Vosniadou, S. De Corte, E. and Mandl, H. Technology-basedlearning environments. Berlin: Springer-Verlag, (1994).
57. Brown, A.L. Design experiments: Theoretical and methodologicalchallegnes in creating complex interventions in classroom settings. Journal of the Learning Sciences, 2, 141-178 (1992) .
58. Bereiter, C. and Scardamalia, M. An architecture forcollaborative knowledge building. In E. De Corte, M. Linn, H. Mandland L. Verschaffel (Eds.) Computer-based learning environments andproblem solving. Berlin: Springer-Verlag (1992) .
59. Cognition and technology group at Vanderbilt From visual wordproblems to learning communities: Changing conceptions of cognitiveresearch. In K. McGilly (Ed.) Classroom lessons: Integrating cognitive theory and classroom practice. Cambridge, MA: MIT press
60: Illich, I. Deschooling society. Harmondsworth: Penguin Books (1973)
61. Crook, C.K. Schools of the future. In T. Gill (Ed) Electronic Children. London: National Children's Bureau.