Adaptive Learning Opportunities

Gary Natriello Teachers College, Columbia University

Technical Report

Draft for Comment

September 2010

1

Abstract

This report reviews a wide range of work to create adaptive learning applications both within the education sector and the research community and beyond. The review is organized according to a model of the processes leading to adaptive learning opportunities that calls attention to seven categories of work. One type of work focuses on what we term the “natural domain,” or the broad set of tasks in which individuals may be involved as they encounter adaptive systems in everyday interactions. A second type focuses on what we refer to as the “learning domain,” or tasks that had a deliberate learning goal. Applications growing out of knowledge of the learning process constitute a third type of work based on what we call the “learner model” tradition. We identify adaptive testing efforts as the fourth type of work and refer to these in the assessment tradition. The fifth type of work is based on knowledge of the relationships among content elements or curriculum components and constitutes what we term work according to a content model. Development activities rooted in an understanding of effective pedagogical activities form a sixth type of work that we label those based on a teacher model. The seventh type of development work which we refer to as network oriented draws on the possibilities for adaptive learning opportunities growing out of new social networking options. We identify research and development activities in each of these traditions, note some promising directions for additional work, and conclude with a discussion of the structures that might be used to connect work both across different development communities and between these communities and practicing educators interested in applying adaptive learning opportunities in their own educational settings.

2

Introduction

Personalizing education by adapting learning opportunities and instructional practices to individual abilities and dispositions has been a long-standing objective among educators and indeed, among all who seek more powerful learning experiences. Indeed, anyone who has been in the role of a student understands the need for attention to specific forms of knowledge and skill and particular learning orientations. Corno (1995) cites Snow (1982) who found references to the need to adapt education to learner differences in Chinese, Hebrew, and Roman texts as early as the first century BC. Snow also identified a specific fifth century BC passage from Quintilian discussing the different patterns of student learning, noting that some students work best with a free rein while others work best under some threat and that some make progress over time while others achieve more through a single burst of energy.

Interest in adapting learning opportunities to meet the needs of learners has continued up to the present time. This enduring interest has appeared in multiple stands of thought reflected in diverse writing ranging from formal scholarly papers to popular commentaries. Although the strands take quite different forms, they each signal a continuing quest for powerful learning experiences rooted in addressing distinctly individual states and characteristics. The strands can be characterized as comprising five different foci we can define as: relationships, institutional settings, communities, and mechanical affordances. Brief explanations of each strand provide background for the model of adaptive learning opportunities

The Relationship Strand

One strand of thinking has focused attention on the unique relationship between two individuals, one in a teaching role and the other in a learning role. The interest is in the relationship between individual students and their tutors or mentors. Examples of students and their mentors run throughout history, e.g., Plato and Socrates (Plato, 2003), Arthur and Merlin (Lupack, 2007). The primacy of the one-to-one relationship between learner and teacher has been captured perhaps most graphically in James Garfield’s image of the ideal college as a log hut with a bench and student at one end and Mark Hopkins at the other (Mark Hopkins, 2009). Examples in this strand of thinking contemplate the possibilities for enhancing learning through a masterful teacher providing undivided attention to am individual student.

3

The Institutional Strand

A second strand of thinking has sought to address limitations on adaptive learning posed by the increasingly institutional forms of education in the modern era. The organization of modern educational systems aims to configure classes of students who must meet common goals; their individual needs must be addressed by a single teacher presenting a standard curriculum. Thus, there continues to be an interest in the power of pedagogies that can be adapted to individual needs. In the latter part of the twentieth century, this interest in adaptive learning opportunities has taken form in the movement toward individualize instruction (Glaser, 1977; Tomlinson, 1999; Betrus, 2002) even in schools in which students are organized by age-grades and classrooms.

The Community Strand

A third strand of thought has focused on addressing the strengths and weaknesses of individual learners by connecting them to broad networks of resources that extend beyond the boundaries of particular relationships and particular institutions. Such learning opportunities have been seen as abundant in the study groups of colonial America (Cremin, 1972), in the envisioning of elaborate society-wide learning webs in a post-schooling era (Illich, 1972), and more recently as elements of communities of practice (Wenger, 1999). In each of these cases the power of adaptive learning opportunities grows out of the assemblage of vast and broad human resources and the highly refined informed connections between those resources and individual learners.

The Mechanical Strand

A particularly provocative strand of thinking has dealt with the shortage of human resources to support individualized learning by imagining some type of mechanical or non-human resource that would be responsive to learner capabilities and dispositions. This line of thinking is part of a broader quest for non-human resources to provide assistance of various kinds (Petrina 2008), where learning is but one. An early form of such a desired resource is seen in the Mechanical Turk, ostensibly a machine that could play chess, a skill demonstrated across Europe in the late 18th and early 19th centuries by besting a string of human opponents (Levitt, 2006; Standage, 2002). Later revealed as a hoax, this machine nevertheless evidenced a continuing desire on the part of human beings to believe in the power of mechanisms that would support or supplant intellectual work. In the mid-twentieth century this same desire for tools that would extend intellectual capacities was manifested in Vannevar Bush’s (1945) vision of the memex. The

4

memex was an early version of a hypertext system that would allow individuals to create and share paths through libraries of materials. The spirit of the memex was carried forward though multiple designs and iterations in the early days of computers as devices to expand the capacity of the human mind. It then moved into the era of computer networks linking individuals to machines, and eventually to other human beings, again as a way to address enhance capacities (Markoff, 2005). A parallel path motivated successive generations of technologies to be used in schools, including film-strips, films, radio, television, programmed materials, computers, interactive games, etc. In each case one goal was to assist learning through non-human resources.

The Current Press for Adaptive Learning Opportunities

The goal of developing means for responding to the states and learning orientations of individuals has received renewed emphasis in recent years both as a result of new demands and the prospect of enhanced possibilities for meeting such demands. These two forces are converging to make the present a particularly propitious time to realize significant progress in the development of adaptive learning opportunities.

New demands for learning are driven by the growing realization that advanced economies and their societies require greater proportions of their citizenry to achieve high levels of learning and the capacity to continue learning throughout the life course. The growth of “knowledge work,” i.e., those tasks that entail the symbolic analysis or the manipulation of information, means that it is increasingly important for individuals to have access to opportunities to acquire the intellectual capacities necessary to perform higher level tasks in all sectors, eg., health, education, government, high tech manufacturing, agriculture.

If the new post-industrial economies in nations throughout the world represent the pull for adaptive learning opportunities that facilitate the development of higher order skills essential to knowledge work, then it is developments in the technologies of computing and communications that represent the push for such opportunities. Progress over the past twenty years in computing and communications technologies has set the stage for the development of a new infrastructure that promises to support adaptive learning in ways never before possible (Computer Research Association, 2005; NSF Task Force on Cyberlearning, 2008). The Information Age has created opportunities for a generation of students to experience web-based learning environments (Strayhorn, 2006).  Students, especially those in secondary and post-secondary education, are growing increasingly

5

knowledgeable in computer sciences and technology (Strayhorn, 2006).  Many use email, word processing, search engines, and web resources to facilitate research and supplement their learning. This situation has served to further aid the creation of personalized learning experiences for individual learners (Oh & Lim, 2005). Particularly in recent years, the World Wide Web has become a popular platform for creating and hosting adaptive learning environments (Weber, 1999). This new infrastructure, anchored by the internet and growing upon it, promises to support the kind of individually responsive learning environments that have only been a dream to earlier generations (Maeroff, 2003; Gardner, 2009).

Method of this Report

Work to develop what we are calling personally adaptive learning opportunities has evolved in a variety of contexts, and progress has been achieved through a variety of methods and approaches.  Moreover, work related to adaptive learning opportunities has been performed by scholars, publishers, software engineers, practicing educators, and others who work in widely disparate professional communities. Because our longer-term goal is to foster the development of linkages among these communities, in this report we take a deliberately expansive perspective on adaptive learning reaching far beyond formal intentional teaching and learning efforts to include those that have evolved for a range of other purposes, at times quite incidentally.

Our activities to assemble a rudimentary base of information on these efforts have included three prominent efforts: 1) surveying the available literature, particularly in areas where the agenda has been driven by researchers; 2) examining a broad set of formal adaptive learning programs of disparate origin; and 3) gathering a wide range of technology mediated or supported experiences that are adaptive in nature and only secondarily involve intentions to support learning.

To facilitate understanding of this broad and disparate base of information, activities, and actors, we propose a model of the set of processes and interactions that can lead to adaptive learning opportunities. This model includes elements common to cybernetic feedback processes (Weiner, 1948) and general systems thinking (Bertalanffy, 1968) while also drawing on work on evaluation and control systems (Dornbusch & Scott, 1975) and adaptive technologies (Shute & Zapata-Rivera, 2007).

Figure 1 highlights the major features of the model. The learner is shown in the inner-most circle as the focal point of activity within the model. Three arrows

6

going to or from the learner represent the interactions between the learner and the adaptive system in which s/he is embedded. These interactions are anchored by three defining dimensions of the system shown on the outer circle. Beyond these are diverse resources that may be invoked by the system in operation. As will become clear, these seven classes of resources correspond to areas of development and research that serve to advance the state of the art of adaptive learning opportunities.

Figure 1: A Model of the Processes Leading to Adaptive Learning Opportunities

The elements of the model describe the major features of adaptive learning opportunities as follows.

1. Natural Domain

The term natural domain is used to refer to the broad set of tasks in which individuals may be involved as they encounter adaptive systems. Those concerned with learning deliberately are handled in the learning domain; all others are handled in the natural domain.

2. Learning Domain

7

Because we are concerned with adaptive learning opportunities it is reasonable to devote special attention to those efforts deliberately or intentionally focused on the promotion of learning. Thus, the domain of tasks deliberately concerned with the promotion of learning receives specific consideration.

3. Goals

The adaptive learning opportunity is initiated in the context of a goal or set of goals. These goals can be learning focused or more general in nature, i.e., pertain to the learning domain or the natural domain. In addition to variation in the substance of the goals, there can also be variation in the origin of the goals; in some cases goals are generated by learners or performers themselves, in other cases goals are generated by other individuals or systems. So, for example, learners can seek to locate themselves within adaptive learning opportunities to pursue goals (learning or otherwise) of their own choosing, or as prescribed by a curriculum.

4. Assign Task

Tasks within the model are important orienting devices that serve to evoke learner cognitive behavior or performance. Tasks are “assigned” (i.e., attached to learners) either by learners themselves or by other individuals or systems. So, for example, learners taking charge of their own learning opportunities can embrace a task thought to advance their learning by affiliating with an adaptive system.

5. Learner

The learner as the focal point of the model is the subject of all action within the system. More generally, the learner can be any individual performer targeted by the action of the system, whether or not s/he conceives of her/him self as a learner.

6. Capture Performance Sample

An essential element in any system intended to adapt to the individual learner is some way to capture a relevant sample of the performance of that learner. Further action within the adaptive system is based on this sample.

7. Appraisal

Once a sample of performance is taken, the information contained within the 8

sample must be analyzed or appraised from some perspective or according to some criterion or criteria, i.e., it must be made meaningful to guide further action.

8. Learner Model

A model or models of the learner or learning are drawn on as the basis of the appraisal of the learner performance. Both the nature of the learner models and the extent to which they are used to drive further action can vary substantially. Elaborate models can be tied closely to fine-grained action while simple models can trigger system actions of a rudimentary nature.

9. Selection

The process of selecting the next action of the adaptive system is governed both by the appraisal and by the resources that the system can draw upon. Selecting involves matching the state of the learner indicated in the appraisal with the appropriate resource or resources.

10. Assessment

One resource that can be presented to learners based on a sample of performance appraised with reference to a learner model is an assessment. Such assessments can be tailored based on the prior performance to achieve more refined understanding of the learner state.

11. Content Model

Content or curriculum can be used to foster learning, and the selection of content that is most appropriate in response to a sample of prior learner performance can be guided by a more or less elaborate content model. More elaborate content models can generate more fine-grained content decisions.

12. Teacher

Teacher is used to signify both the availability of a human teacher or an intelligent tutor or teaching agent and the specific behaviors evinced by either. A key defining feature of this resource is that a single individual or single intelligent agent is made available to the learner.

13. Network

9

In contrast to the single resource implied by the teacher category, the network category refers to a multi-actor set of interactions. The network can be bounded in a deliberate way as defined by the system or it can be open. In either case, information on the actors in the set drive the selection process to address learner needs. 14. Present

The third action impacting learners is the act of presenting the resource selected (assessment, content, teacher, network set) to foster learning. The presentation act concludes a single adaptive cycle. The system can be configured to continue with another cycle as it repeats the stages already reviewed.

Areas of Development and Research on Adaptive Learning Opportunities

The model of adaptive learning opportunities highlights the kinds of resources that shape and define different approaches to the construction of adaptive learning systems. The seven classes of resources in the model and our review of existing development and research on adaptive systems suggests seven major types of work that provide perspectives on contemporary adaptive learning opportunities.

Natural Tasks - A growing variety of everyday tasks that require or benefit from enhanced knowledge or skills on the part of participants have come to rely upon adaptive learning strategies to secure higher quality individual efforts.  In these settings learning is embedded in the broader goal of client engagement (e.g., TurboTax and other programs to guide tax return preparation).

Constructed (Learning) Tasks - Learning opportunities that are deliberately organized around tasks or similar application structures constructed for the purpose of promoting learning. With the growth in computing and communications capabilities, approaches that have traditionally relied on more deliberately constructed tasks such as problem-based learning, project-based learning, and scenario-based learning, are increasingly including personally adaptive elements.

Learning Research - Research efforts driven by a learning process perspective have been influenced by the base of knowledge on the patterns of human learning and strategies to apply technologies to leverage those patterns. (e.g., hypermedia and self-regulated learning)

10

Assessments - In the realm of assessment a considerable amount of work has been devoted to creating adaptive dimensions within single assessments or programs of assessments (e.g., computer adaptive testing)

Curriculum Development - Efforts growing out of a curriculum perspective have been rooted in an understanding of how content can be organized to facilitate mastery of material, whether based on traditional expert-driven approaches to generate curriculum scope and sequence plans or based on empirical investigations of sequences and their impact on learning as represented in learning progressions.

Teaching Research - Work oriented by the teaching perspective has focused on developing insight into differences in student learning that can be matched to a broadened repertoire of instructional strategies to achieve optimal learning. (e.g., adaptive teaching)

Social Networks – Technology mediated or supported social networks have offered the capacity to connect learners with other individuals who can support and enhance their learning (e.g., online professional learning communities).

Any parceling of the broad set of activities that may contribute to the evolution of adaptive learning opportunities is necessarily somewhat arbitrary. The seven perspectives identified are at times overlapping, and at least some adaptive learning activities draw upon multiple perspectives. Nevertheless, the seven perspectives identified provide a starting point for our effort to scour the terrain of activities that might be brought into a conversation to advance thinking about adaptive learning possibilities.

The remainder of this report is organized around these seven perspectives, and reviews the diverse landscape within which developments in adaptive learning opportunities are taking place. We taken into account the uneven theoretical and empirical progress in different areas, and assess the potential contributions of the various efforts for strengthening understanding of personally adaptive learning opportunities.  We conclude by identifying strategies for facilitating convergence in the field.

A. The Natural Task Perspective

Perhaps the broadest of the seven resource areas specified in the model of adaptive 11

learning systems is what we have termed the natural domain or natural tasks. We use the term “natural” to distinguish this domain of tasks in our model from those that are “constructed” with a deliberate learning goal or focus. “Natural” thus refers to the fact that these tasks (which may facilitate learning incidentally) are not designed deliberately or primarily to promote learning.

A wide range and growing number of other tasks in contemporary life are being developed in ways that make them personally adaptive. These naturally occurring tasks make use of learning (and are sometimes based on learning principles), but they are not explicitly conceived as educational tasks. Brusilovsky and colleagues (Brusilovsky, 1996; Brusilovsky, 1998; Brusilovsky, 2001) provide a key orientation to work in adaptive hypertext and hypermedia. Since the mid-nineties this work has largely concentrated on web-based media. Stand-alone hypermedia systems are still being introduced, but the growing capacity of web interfaces and the distributional advantages of the Internet make web systems an increasingly attractive avenue for development. Accompanying the shift to web-based adaptive systems has been the coming together of the user modeling and hypermedia community to allow for the development of systems that adapt in response to individual users.

Brusilovsky and Maybury (2002) distinguish between adaptable systems and adaptive systems. Adaptable systems require the user to specify the ways in which the system should be different for his or her use. For example, in an adaptable news source, the user would specify his or her location to add local news. Adaptive systems rely on an explicit model of the individual user that includes information on the user (e.g., knowledge, goals, interests, recent online behavior) that is then used to distinguish among different users. Brusilovsky and Maybury (2002) note that user data can be gathered by observing user interaction and by making direct requests for user input. In our model of adaptive learning opportunities, these data sources are organized through the task assignment process.

Brusilovsky (1996) identified five types of information on individual users that might be used in an adaptive system: goals, knowledge, background, hyperspace experience, and preferences. Brusilosvky (2001) identified two additional types of information about the user; namely, interests and individual traits. He adds environmental dimensions as another kind of information to drive adaptation. Consideration of these eight types of information that might be employed in a user model illustrates the scope of individualization that is possible.

Goals refer to the objectives the user is trying to accomplish in the context of the 12

environment in question. Adaptive systems typically support a set of possible user goals. Simple systems may support one or a small number of goals or tasks, while more complex systems may support a hierarchy of related tasks leading to goal accomplishment.

User knowledge usually refers to the knowledge that a user possesses in relation to a model of the structure of a bounded knowledge domain (Brusilovsky, 1996). The knowledge domain is represented by a structure of concepts and their relationships, and the knowledge of the user is represented in an overlay model that seeks to assess the state of user knowledge in reference to the domain structure. The assessment can be binary, a qualitative measure, or a quantitative measure used to represent the knowledge state of the user. In some cases a simpler stereotype model with several typical or stereotype users is specified as the reference for assessing user knowledge

Background refers to all of the available information related to the user’s previous experience outside the subject of the adaptive system. Included under background are factors such as the user’s profession, work history, and point of view (Brusilovsky, 1996).

Hyperspace experience refers to the user’s familiarity with the structure of the hyperspace system as reflected in the ease of navigation. Such familiarity with the structure of the system is distinct from the user’s knowledge of the subject, and has direct bearing on the most appropriate adaptive navigation techniques. Experienced users of systems that are not adaptive may encounter the effects of this kind of hyperspace experience as they become less patient with the very navigation prompts that they found useful as novice users.

Preferences refer to the fact that a user can prefer some areas of a system over others and some parts of an individual page over others. Brusilovsky (1996) points out that the user has to inform the system about these preferences and that adaptive systems can then generalize such preferences in new contexts within the system. He also suggests that when user preferences are represented numerically they can lead to the development of group user models that accumulate the preferences of a particular group and lead to systems optimized for collaboration.

Interests refer to the long-term or enduring interests of the user (as opposed to the short-term task or user goal). Interests are an increasingly important dimension of information retrieval systems as the systems combine interests with short-term goals or tasks to develop models to improve information filtering and recommendations. As Brusilovsky (2001) notes, user interests are a part of

13

adaptive bookmarking systems and some other systems that rely on interests for making recommendations.

Traits, as conceptualized Brusilovsky (2001) refer to user features such as personality factors, cognitive factors, and learning styles that are stable or only change over long periods of time. There is controversy among psychologists over use of this term; some reserve the term trait for personal qualities that cannot be changed, and prefer to use propensity in referencing features that Brusilovsky calls traits (Stanford Aptitude Seminar, 2002). Nevertheless, relatively stable qualities such as these present special challenges to adaptive systems because they often require special psychological tests to extract.

The environment in which the user is accessing the adaptive system offers possibilities for adaptation along several dimensions. User environments can differ, for example, in terms of the software platform, hardware capacities, and network bandwidth. Adapting successfully to variables such as these can influence comprehension of material presented. The growth of mobile devices adds two other aspects of the environment connected to the smaller screen and the capacity to identify location based on the global positioning system. Location information has adaptive potential for delivering guide information in terms of both the general area and the finer-grained positioning within buildings.

Having addressed the broad range of types of information to drive adaptation, we can turn to the complementary issue of what can be adapted in adaptive systems. Brusilovsky (1996; 2001) is instructive in this regard as well. Brusilovsky (1996) notes the limitations of adaptive hypermedia systems, a point that will be highlighted by our later consideration of adaptive teaching where there is a much more expansive repertoire of adaptivity. He identifies two broad classes of adaptive strategies: adaptive presentation and adaptive navigation.

Adaptive presentation includes adaptive multimedia presentation, adaptive text presentation, and adaptation of modality. Adaptive multimedia presentation involves tailoring the presentation of content within online media based on user characteristics. A familiar example might be a system that recognizes the bandwidth of a user and alters the density and size of multimedia files to accommodate bandwidth limitations to provide a viewable media asset to the user. Adaptive text presentation entails presenting users represented by different user models with different text. Brusilovsky (2001) divides adaptive text presentation into canned text adaptation and natural language adaptation. Canned text adaptation can involve inserting/removing text fragments, altering fragments, stretchtext (i.e., providing more detailed text treatment of particular content to

14

facilitate user understanding), sorting text fragments, or dimming fragments. Natural language adaptation may make similar use of fragments through the use of natural language technologies.

Adaptive navigation support refers to those techniques that assist users in finding their paths through hypermedia content adapting the way that links are presented in response to differences in user characteristics. Brusilovsky (2001) identifies six techniques: direct guidance, adaptive link sorting, adaptive link hiding (including hiding, disabling, and removal), adaptive link annotation, adaptive link generation, and map adaptation.

Direct guidance involves showing the user the best link or node based on user characteristics as represented in the user model. Adaptive ordering or sorting positions the best link according to the user model at the top of the list of links. Adaptive hiding restricts the navigation possibilities available to the user by hiding links to pages that are not relevant or appropriate according to the user model. Adaptive link annotation provides additional information to the user about the node or page behind the link either through additional text or visual cues. Map adaptation entails ways of adapting the form of hypermedia maps based on the characteristics of the user as represented in the user model.

Brusilovsky and Maybury (2002) and Brusilovsky (2004) discuss the challenge of adaptive navigation support in situations other than assisting users in moving through a closed body of hypertext. Two situations which press the limitations of adaptive systems are the open body of resources represented by the web and the complex graphical dimensions in virtual reality environments. Providing adaptive support in these situations pose new challenges and opportunities. Brusilovsky (2004) discusses the use of both content-based and social technologies to address the open body of resources on the web. Brusilovsky (2004) notes that adaptive navigation support technologies developed for hypertext have analogs in similar approaches in virtual environments.

Adaptive hypermedia systems based on user characteristics and types of adaptivity discussed in this section have taken diverse forms. Brusilovsky (1996) identifies six kinds of adaptive hypermedia systems: online information systems, online help systems, information retrieval systems, institutional hypermedia, systems for managing personalized views of information spaces, and educational hypermedia. Online information systems operate to provide reference access to information for diverse users. Online help systems operate much the same way, but the information they provide is directly related to the online system the user is accessing. Information retrieval hypermedia systems combine information

15

retrieval techniques with access from an index of terms to a set of documents. Institutional hypermedia systems seek to provide online access to all of the information necessary to support the work of an institution. Systems for managing personalized views of information spaces provide access to a subset of hyperspace selected as a result of user characteristics represented in a user model. Finally, educational hypermedia seek to provide adaptive access to the hyperspaces representing a particular course or part of a curriculum. We consider educational hypermedia under the learning domain or learning task discussion in a later section.

Writing in 2009, Knutov, De Bra, and Pechenizkly consider progress in adaptive hypermedia systems over the preceding twelve years. They organize their discussion around six major questions of adaptation:

Why do we need adaptation (Why?)What can we adapt? (What?) What can we adapt to? (To What?) Where can we apply adaptation? (Where?)When can we apply adaptation? (When?)How do we adapt? (How?) (pp. 7-8)

Knutov, De Bra, and Pechenizkly (2009) note that reference models used to describe adaptive hypermedia architectures employ a tower model that represents different elements as different levels or layers. Along these lines, they discuss different layers as they relate to the key questions.

The Goal Model (GM) layer (addressing the “Why?” question) potentially involves not just goals, but also a hierarchical structure that includes “goals, objectives, tasks, requirements, workflows” (Knutov, De Bra, and Pechenizkly, 2009, p. 23) of the sort shown in our model of adaptive learning opportunities as (3) goals, (4) tasks, (5) learner, and (6) performance. A generalized view of the goal centered approach might include a “hierarchy of goals and corresponding tasks comprising this goal, workflows that need to be followed to complete a requirement (Knutov, De Bra, and Pechenizkly, 2009, pp. 23-24). In addition to systems in which the goal is assigned by some expert or selected by users from the outset, Knutov, De Bra, and Pechenikly (2009) note that it is possible for a system to infer a goal from the interactions of other users of the system.

The Domain Model (DM) layer (addressing the “What?” question) consists of concepts and the relationships among concepts, most often organized in a hierarchy. When the Domain Model is structured so that concepts are fine grained

16

and structured hierarchically it is possible to apply adaptive techniques. As Knutov, De Bra, and Pechenizkly (2009, p. 14) observe, “Each concept relationship has a type (e.g., direct link, inhibitor, ‘part-of’, or prerequisite) which may play a role in the adaptation.” Two issues receiving current attention are the challenge of dealing with an open body of documents and the idea of higher order adaptation. Domain Models applied to a closed body of material can be defined in advance by a domain expert, but when we are dealing with an open body of material, as in the case of the entire Internet, then the problem of mapping concepts to content must be addressed as elements of the open body of material are brought into play. Thinking about higher order adaptation presents its own challenges. Systems with higher order adaptation are able to monitor the user as a basis for determining the parameters around which adaptation will occur.

The User Model (UM) layer (addressing the “To What?” question) is made up of entities that have values stored in a structure (typically a table) that overlays the Domain Model, in effect creating a map of some attribute (such as user knowledge) over the domain. Systems can use both domain dependent properties such as user knowledge and domain independent properties such as user characteristics. Systems can respond to both dynamic properties such as user knowledge gathered from user interaction with the system and static properties such as user personal characteristics. Systems can simply draw on the static properties but they must process ongoing changes in dynamic properties (Knutov, De Bra, and Pechenizkly, 2009, pp.19-20).

The Context Model (CM) layer (addressing the “Where?” and “When?” questions) refer to system features that take account of the context in which the adaption is applied and the environment in which the application is used. Adaptation processes and decisions that are dependent on an application address the “Where?” question. Adaptation processes and decisions that are independent of an application (e.g., time, day of the week, network bandwidth) address the “When?” question. Although Knutov, De Bra, and Pechenizkly (2009, pp. 24-25) discuss these issues, they do not treat them as a layer in the general model.

The Adaptation Model (AM) layer (addressing the “How?” question) deals with techniques and methods of providing adaptation within a system. According to Knutov, De Bra, and Pechenizkly (2009, pp. 25), “Methods represent generalizations of a technique.” They identify techniques of content adaptation (inserting/removing fragments, altering fragments), techniques of adaptive presentation (dimming fragments, sorting fragments, stretchtext, zoom/scale, layout, link sorting/ordering, link annotation, combinatorial techniques), and techniques of adaptive navigation (link generation, guidance, link hiding).

17

Knutov, De Bra, and Pechenizkly (2009, pp. 30-34) conclude their review of progress in adaptive systems by highlighting a number of emerging trends likely to drive work on such systems in the immediate future. These include:

Using ontologies to support bringing together user models from different applications so that user experience in one application can be shared with another application

Applying adaptive methods to an unknown or open document space such as the web

Extending adaptive processes by applying them not just to individuals, but to groups of users

Using data mining techniques to support adaptation, notably the updating of event-condition-action rules that drive the adaptive system

Applying higher order adaptation in which monitoring of the user results in modifications of the adaptation itself.

Developing the Context Model to allow the system to operate under different rules in different contexts

Extending adaptation techniques to multimedia content

To these anticipated areas of development we would add the possibility of extending adaptive systems to include the altering of not only content, but applications or processes. As applications come to be a more prominent element of the world wide web, the ability to alter applications based on knowledge users before they initiate interaction with an application will become an important way to enhance the user experience.

B. The Constructed (Learning) Task Perspective

The Constructed Task Perspective calls our attention to development and research that includes construction of tasks specific to the application in question. Such tasks can be more or less learning oriented. For example, some elaborately constructed task environments are created for purposes of entertainment rather than learning. But learning oriented tasks are well represented among constructed

18

tasks, and a good deal of theoretical and practical work focuses on developing task environments that promote learning.

As we move from the domain of natural tasks to the domain of constructed tasks, it is immediately evident that developers of such systems have conceived and planned the task environment presented to user/learners/students in a very deliberate manner. Indeed, work in this area places great emphasis on the conscious creation or design of tasks and task environments. A key aspect of the deliberate nature of the design of these environments is the connection to different theoretical traditions for guiding the design of adaptive systems.

For our present purposes we can divide the theoretical traditions guiding the design of tasks for adaptive learning opportunities into two major categories: objectivist and constructivist. These traditions differ in their major assumptions about how individuals learn and consequently they carry different implications for the design of adaptive systems. One of the major traditions is rooted in what Jonassen (1999) terms “objectivist conceptions” of learning. This tradition has been dominant in instructional design; authors such as Papanikolaou, Grigoriadou, and Samarakou (2005) refer to it as the “instructional design approach.” In the objectivist tradition knowledge is assumed to be transferred from teachers to students or transmitted from instructional technologies and acquired by learners. Papanikolaou, Grigoriadou, and Samarakou (2005) describe such approaches as entailing definition of specific learning objectives, the design of materials and procedures to address these objectives, and assessments to determine if the objectives have been met. Instructional systems with these elements go back to early military training programs and the curriculum tradition of Ralph Tyler (1969). Papanikolaou, Grigoriadou, and Samarakou (2005) note that such approaches entail the definition of specific learning objectives, the design of materials and procedures to address these objectives, and assessments to determine if the objectives have been met. Papanikolaou, Grigoriadou, and Samarakou (2005, p. 2) note that in the instructional design or objectivist approach learners are provided with structured content consisting of various knowledge modules differing in format and interactivity level along with individually tailored support for acquiring the intended knowledge, skills, and/or attitude objectives.

Numerous adaptive learning systems, or what Brusilovsky and Peylo (2003) term adaptive and intelligent web-based educational systems (AIWBES), and Papanikolaou, Grigoriadou, and Samarakou (2005) refer to as adaptive educational hypermedia systems (AEHS)), have been developed based on the objectivist perspective. Papanikolaou, Grigoriadou, and Samarakou (2005) argue that the majority of early adaptive learning systems, including DCG (Vassileva, 1997),

19

ELM-ART (Weber & Bruskolvsky, 2001), INSPIRE (Papanikolaou, Grigoriadou, Kornilakis, & Majoulas, 2003), AST (Specht, M., Weber, G., Heitmeyer, S., & Schöch, V.,1997), and KnowledgeSea (Brusilovsky & Rizzo, 2002), were designed from the objectivist perspective. A similar claim may be made about WebMA, a system designed to provide drill and practice in mathematics (Nguyen & Kulm, 2005).

Constructivist approaches have been employed in a range of projects to design technology mediated or enhanced learning environments. Central to constructivism is the assumption that knowledge and understanding are constructed individually and collectively by learners as a result of their experiences. The implication of this view is that learning is facilitated by experiences that allow individuals to construct knowledge. Jonassen (1999) summarizes some key aspects of the design of constructivist environments. He highlights the problem-focused or case-oriented approach in which learners are presented with an interesting problem or a challenge and provided relevant information along with tools to scaffold required skills, tools to support communication and collaboration, as well as social or contextual support for learning.

A variety of related emphases are captured in work on constructivist learning environments. For example, McClintock (1971) and Black and McClintock (1995) have envisioned and developed digital information infrastructures that constitute environments for supported study opportunities for students. Schank and colleagues (Shank, Fano, Bell, & Jona, 1993/94 emphasize the use of goal-based scenarios to orient and guide activities in learning environments. Bransford and colleagues (Bransford, Sherwood, Hasselbring, Kinzer, & Willams, 1990) used anchored instruction in which learning is focused through a problem entailing realistic tasks. Work on problem-based learning (Baron, et al., 1998) and project-based learning (Blumenfeld, Soloway, Marx, Krajcik, Guzdial, & Palincsar (1991) leads to the design of learning environments oriented around realistic and more open-ended challenges to engage learners along with resources to support their learning. Shaffer (2004; 2006) focused on development environments and experiences for students that resemble real world work environments associated with the professions as well as games that hone skills that will be useful in real world applications. Papert and colleagues (Harel & Papert, 1991; Kafai & Resnick, 1996) championed a constructionist perspective stressing that individuals learn effectively when they are engaged in building, in making things in the real world. All of these approaches in the constructivist tradition incorporate the idea of situated learning, i.e., learning in the context of realistic situations, and scaffolding, i.e., personalized support from an individual or a resource that scaffolds or assists in learning. In a web-based environment, the scaffolding is provided by the website and its multimedia options (Own, 2006).

Efforts at developing adaptive learning opportunities in the constructivist tradition locate the adaptive dimensions in the environments provided for learners rather than in a direct instructional interaction. As Lee and Park (2008) observe, when designers of computer-

20

based learning environments begin with a constructivist perspective instead of an objectivist approach, they emphasize the facilitative role of the system in place of the tutoring role. For example Akhras and Self (2002a) note that while environments rooted in objectivist conceptions of knowledge have architectures that emphasize representations of the knowledge to be learned (i.e., a domain model), inferences regarding the current knowledge state of the learner (i.e., a learner model), and instructional activities for the learner (i.e., a teaching model), environments based on constructivist perspectives might have quite different architectures.

Akhras and Self (2002a) propose such an architecture that includes different but parallel components: situations or interactional contexts that facilitate the activities of learners to construct their own knowledge, properties of interactions and sequences of such interactions that reveal the processes by which learners are constructing knowledge, and opportunities provided by affordances of learning situations to learners whose learning process is in a particular state. Akhras and Self (2002a) view the domain, learner, and teaching models characteristic of systems grounded in the objectivist perspective as being contained within broader categories of the situation, interaction, and affordance models they offer as compatible with the constructivist approach. Akhras and Self (2002a) illustrate their approach with two systems: SAMPLE, a system to help students learn concepts of salad making, and INCENSE, a system to help students learn software engineering.

The proposal for a constructivist architecture by Akhras and Self (2002a) has been challenged from several perspectives. Azevedo (2002) called on Akhras and Self to enhance the conceptual clarity of their approach and apply their constructivist perspective to systems intended to foster self-regulated learning in students. Young Depalma and Garrett (2002) drew on an ecological psychology perspective to call attention to some additional complexities inherent in understanding a learning context. Akhras and Self (2002b) responded to these challenges by clarifying their original arguments, and in a later paper Akhras (2005) offered additional formalization to strengthen the approach to modeling learning interactions.

Henze and Nejdl (1999) and Henze, Nejdl, and Wolpers (1999) used a constructivist approach that incorporates elements of project-based learning, group work and discussions in the design of the KBS Hyperbook System. The system allows students to select their own learning goals and the review suggestions for appropriate projects along with the information necessary to achieve their learning goals.

Papanikolaou, Grigoriadou and colleagues (Papanikolaou, Grigoriadou, 2006; Papanikolaou, Grigoriadou, Alexi, & Samarakou, 2005; Papanikolaou, Grigoriadou, & Samarakou, 2005) discuss the development of ProSys, a web-based adaptive educational system to support project-based curricula and case-based learning. In ProSys learners are engage in project work. As they engage, the system presents them with learning aids to support their completion of the project. Among the aids is a library of case structures to

21

be consistent with the conceptual structure of the knowledge domain. Learners are provided with navigation aids adapted to their knowledge level and degree of progress. In addition, learners are offered a sequence of activities tailored for them.

It is clear that constructed tasks offer important points of entry into research and development of adaptive learning opportunities. Whatever the theoretical perspective underling the tasks, the structure of the task environment plays a defining role in many adaptive systems. Moreover, tasks and task environments are targets for the creative energies of designers and developers in all areas.

C. The Learning Research Perspective

Many research and development efforts to create adaptive learning opportunities have been driven by a learning process research perspective; that is, they have been influenced by the base of knowledge on patterns of human learning and strategies to apply technologies to leverage those patterns. Inherent in these approaches is a model of the learner maintained by the adaptive system and the capacity of the system to adjust itself in response to the range of processes employed by individual learners (Shute & Zapata-Rivera, 2007).

An understanding of learner differences, resources, and responses to those differences is essential to creating an effective adaptive system. Several broad classes of factors on which learners may be anticipated to differ have been considered over the years. These include, for example (as defined by Shute & Zapata-Rivera, 2007): 1) cognitive factors such as knowledge, skills, and abilities (see also Carroll, 1993; Stanford Aptitude Seminar, 2002); 2) motivational and affective factors such as interest, attachment, and success expectations (Deci, Koestner, & Ryan, 2001; Pintrich & Shunk, 1996); 3) demographic factors such as race, gender and social class (Natriello, 1994), and 4) cultural factors such as ethnicity and language (Boykin, 1994; Hurley, Allen, & Boykin, 2009; Rogoff, 2003).

Developments in several areas promise increasingly detailed information about a growing range of learner characteristics. First, more and finer grained information on learners is available as a result of students spending more time in online learning environments. Online environments capture massive amounts of information on learners by tracking student use of the learning resources through their direct interactions with the software programs and augmented by such biologically based means as eye-tracking, speech-patterns, and head- and facial-

22

gesture captures. These large data pools can be explored through data mining techniques that can be used to uncover patterns and trends in student behavior (Shute & Zapata-Rivera, 2007).

Second, progress in the development of models of learning and learners will lead to the identification of both new learner characteristics and new modeling techniques. For example, in response to questions posed by Shute and Zapata-Rivera, (2007, p. 287), a panel of experts in the field identified the following learner variables as promising for consideration:

Cognitive abilities (e.g., math skills, reading skills, cognitive developmental level, problem solving, analogical reasoning)Metacognitive skills (e.g., self-explanation, self-assessment, reflection, planning)Motivational and Affective states (e.g., engagement, flow, anxiety level)Additonal variables (e.g., personality types, learner styles, social skills such as collaboration, and perceptual skills)

In their consideration of constructivist learning environments, Akhras and Self (2002) identify four dimensions of learner behavior, or what they term “properties of interactions and of sequences of interactions,” (p. 8) that may be more likely to lead to learning: cumulative, constructive, self-regulated, and reflective. Lee and Park (2007) call attention to systems that include motivational variables, citing De Vicente and Pain (2002), whose motivation model includes control, challenge, independence, fantasy, confidence, sensory interest, cognitive interest, effort, and satisfaction as variables to indicate a learner’s motivational state. In considering the issue of control, Lee and Park (2007) noted the distinction made between adaptable systems that allow users to choose certain parameters, and adaptive systems where options are determined by the system based on measures of user needs. Lee and Park (2007) also highlight approaches and systems that consider metacognitive skills such as SCIWISE (White, Shimoda, & Frederiksen, 1999), which encourages students to express their metacognitive ideas and Help-Seeking Tutor Agent (Aleven, Stahl, Schworm, Fischer, & Wallace, 2003; Aleven, McLaren, Roll & Koedinger (2004), which supports learners to become better help seekers.

Third, more and better data on students along with new conceptual and theoretical models together with new modeling techniques (e.g., Bayseian networks), can only be employed to create more powerful adaptive learning opportunities centered on learning if the bandwidth available to exchange the information

23

increases. Fortunately, the bandwidth available for such systems is increasing although the distribution of capacity is uneven throughout the world.

With rapid advances in computing and data transfer technologies, a broader set of research efforts to develop systems that intentionally involve adapting to individual learning patterns, and attendant increases in learner agility in the use of these technologies, it is becoming evident that, not only can the broader set of tools be better leveraged to help provide more effective adaptable learning environments, but also that additional learner attributes might well be taken into consideration in the quest for evolving better and more effective adaptive teaching and learning practices.

As Shute and Zapata-Rivera (2007) note, adaptive learning technologies can be considered in terms of a four-process or four-stage adaptive cycle. The four processes are capture, analyze, select, and present. The cycle begins with the capturing of personal information on the learner as the learner interacts with the environment. Such information is used to update internal learner models in the system. A model of the learner organizes the analysis process used to understand the current state of the learner in relation to the domain. In work representative of the learning research area, these learner models are well developed and substantial in driving the adaptive cycle. In the selection stage content is identified in terms of the model of the learner held by the system and the overall goals. In work representative of the learning research area, the emphasis in the selection process is on the learner model. This drives the presentation of the next sequence of information to the learner.

In their discussion of technology-rich environments, Lajoie and Azevedo (2006) stress the utility of research and development driven from well-developed learning theories. They highlight approaches that involve both cognitive and metacognitive tools to support learners. In particular, they argue that in the wake of the growing development of open electronic learning environments accompanying the spread of the Internet, students are increasingly being called upon to regulate their own learning more effectively. This leads them to call attention to the substantial body of work on self-regulated learning (e.g., Boekaerts, Pintrich, & Zeidner, 2000; Butler & Winne, 1995; Paris & Paris, 2001; Schunk & Zimmerman, 2001; Zimmerman & Schunk, 2001).

Lajoie and Azevedo (2006) draw on the synthesis and conceptual work of Pintrich (2000) which they offer as a sound foundation for organizing extant work on self-regulation that can be relevant to technology-rich environments. Four assumptions common to models of self-regulated learning identified by Pintrich are that: 1) learners are active participants in the learning process, 2) effective learners are able to monitor, control, and regulate aspects of their own cognition, motivation, behavior, and context, 3) various

24

constraints, whether biological, developmental, contextual, or otherwise individual, can interfere with learner control and regulation; and 4) there is a goal or standard used by learners to determine whether to continue or alter the current approach to learning. In addition to these assumptions, Lajoie and Azevedo (2006) report on a four stage conceptual framework that guides much research on learner self-regulation: Phase 1 – planning and goal setting, Phase 2 – monitoring, Phase 3 – efforts to control and regulate, and Phase 4 – reactions and reflections on the self, the learning task, and the context.

Research on the impact of hypermedia environments on student self-regulated learning provides a good example of work that expands the view of student learning while simultaneously attempting to create more powerful learning opportunities. The work attends not only to learning gains, but also to the learning process. Self-regulated student learning refers to a host of processes (e.g., planning, monitoring, strategy use, handling of task difficulty and demands) (see, e.g., Greene & Azevedo, 2009) that learners can employ to become more effective. However, as Lajoie and Azevedo (2006) note, despite the range of work being done on technology-rich environments from a self-regulation perspective (e.g., Hill & and Hannafin, 1997; Chou & Lin, 1997; Jackson, Krajcik, & Soloway 2000, Hadwin & Winne, 2001; Azevedo, Guthrie, & Seibert, 2004), there has been little research that would allow us to understand the entire cycle of activities involved in self-regulated learning in technology-rich environments. Pursuing such work as part of a broader class of efforts to enhance adaptive learning opportunities could be very fruitful.

D. The Assessment Perspective

In contrast to other perspectives where the adaptive elements are designed to alter learning resources (e.g., relationships, content) in response to learner/user problem solving and behavior, sometimes gathered in an assessment episode, the assessment perspective focuses our attention on those instances where the adaptation is directed to the assessment itself, i.e., where the later elements of an assessment are altered in light of learner/user performance on a previous assessment element. Such arrangements can be multistage tests in which a set of tests or “testlets” of differing difficulty are organized along various paths with performance at one stage determining the assignment of the testlet in the next stage (Mead, 2006; Luecht, Brumfield, & Breithaupt, 2006). They can also be computer adaptive tests where the test items are of varying known difficulty, and students are assigned them in the course of the test; the assignment is determined by performance on earlier items of known difficulty (Kreitzberg, Stocking, & Swanson, 1978; Wainer, 2000; Challis, 2009). As Van Horn (2003, p. 567) clearly puts it: “If a student misses an item, a slightly easier one is given. If a student gets an item correct, a slightly more difficult one is given.” The estimate of the level at which the student is functioning is adjusted with each answer to guide the selection of the next test item from the item pool.

25

In these assessment situations, there are typically no learning resources per se since the goal is to assess learner/user capabilities in the absence of such resources. In the realm of assessment a considerable amount of work has been devoted to creating adaptive dimensions within single assessments or programs of assessments, and our consideration of adaptive learning opportunities can draw on the unique developments in the assessment field to enrich adaptive learning.

Computer adaptive testing has been employed successfully in large-scale testing programs such as the Armed Services Vocational Aptitude Battery (Sands & Waters, 1997) and the Graduate Record Examination (Schaeffer, Steffen, Golub-Smith, Mills, & Durso, 1995; Mills & Steffen, 2000). Such large-scale programs have both the resources and size to support the necessary test development activities. To address some of the constraints on the application of computer adaptive testing, Welch and Frick (1993) consider several methods for using computer-adaptive testing in instruction settings such as corporate education and training and public school classrooms. Although they emphasize the possibilities for employing the various approaches they review, they also identify numerous barriers such as the need to generate data on test items from large numbers of students prior to using the approach in the context of ongoing programs. Indeed, these kinds of barriers have tended to limit the use of computer-adaptive approaches to more formal testing programs.

The situation is changing with the introduction of learning systems that involve large numbers of teachers and students with the capacity to capture and retain large amounts of performance information. Lilley, Barker, and Britton (2004) highlight some of the advantages of using computer-adaptive testing as part of the managed learning environments made possible by online systems. These advantages include greater motivation for students who are not confronted with test questions that are too easy or too hard, greater efficiency in testing requiring less time from learners, and better information on learner performance to drive the educational experience.

Nirmalakhandan (2007) describes the use of computer-adaptive testing in the context of a computerized adaptive tutorial and assessment system used for a university level hydraulic engineering course. Triantafillou, Georgiadou, Anastasios, & Economides (2008) offer a design in which computer-adaptive testing is implemented on mobile devices that can be accessed by students in diverse settings. Speaking to the increasingly diverse applications that may be possible for computer-adaptive testing, McGlohen &Chang (2008) and Cheng (2009) present data on the use of computer-adaptive testing in the context of cognitively diagnostic assessments that can provide not just an assessment to drive the next test item, but also diagnostic feedback to learners. Although problems with computer adaptive testing remain to be addressed (e.g., Rulison & Loken,

26

2009), continuing work in this area of assessment may offer important benefits to the overall goal of providing adaptive learning opportunities for students.

E. The Content Model Perspective

The content model perspective calls attention to a variety of strategies and approaches to controlling and shaping content and the presentation of content to achieve a variety of goals. Such work places emphasis on the selection, organization, and presentation of content to learners. In the present case the most relevant goals for driving actions to control content are learning goals. Attempts to control and shape content for educational purposes have a long and robust role in the history of education and formal schooling, and they have been the subject of a good deal of analysis and theorizing (e.g., Tyler, 1969; Pinar, 2004; Tanner & Tanner, 2006; Schiro, 2007). We briefly consider some of this work as a foundation for the insights it might provide to contemporary efforts to create adaptive learning opportunities.

While our immediate interests are more focused on those dimensions of curriculum or content models that form the basis for the kinds of differentiation that can be associated with adaptive learning opportunities for students, it is important to note that much of the effort around theorizing and policy making concerning curriculum has been devoted to specifying commonalities of content experience (Tyack, 1974; Reese, 2005) rather than diversity. Exposing individuals to common content with the intention to accomplish common learning goals has been a strategy to build group cohesion and solidarity. This strategy has been applied at the level of the nation state (Meyer, Tyack, Nagel, & Gordon, 1979), the community (Servgiovanni, 1996), the professional group (Boyer & Mitgang, 1996), the corporation (Chrusciel, 2006), and the classroom (Edward & Mercer, 1987). Even these attempts to convey a common curriculum or content to diverse groups have led to differentiated efforts to secure the attention and engagement of all concerned on the common content agenda.

The history of curriculum differentiation is of interest because it includes important dimensions for adapting content. The major dimensions along which content has been differentiated include student background, including cultural background (Ladson-Billings and Brown, 2008), student destinations, including career aspirations (Grubb, 1995; Gordon 2007), and differentiation by personal qualities such as interest (Tomlinson, 2001), and student intellectual ability (Reiser, 1987). Efforts to differentiate educational opportunities have differed along each of these dimensions and often among several at the same time.

Differentiation efforts have occurred at the system level, the school level, and the classroom level. At the system level, attempts to differentiate based on student characteristics have led to the development of different types of schools. For example, in some systems students are sorted into vocational and college preparatory schools with

27

possible gradations within each type. In the U.S. we have witnessed the emergence of special theme schools as part of magnet school policies or more recently as charter schools (Buckley & Schneider, 2009). At the school level instructional alternatives have been selected on the basis of learning goals and students' general cognitive abilities and achievement levels in the curriculum structure. Emerging in the early 1900s, this approach was meant to be an alternative to the traditional lock-step system, and so different instructional and assessment methods were adopted to better accommodate students' learning abilities (Reiser, 1987). The ultimate result, however, was that students were simply grouped by grades or scores on standardized tests (i.e., “tracked” by ability or achievement level), while instructional strategies that generally involved the instructor presenting information, asking questions to monitor learning and then providing feedback based on students' individual performances, remained essentially the same. It is therefore not surprising that school level adaptive instructional systems are not well-received in mainstream education today; nevertheless, some form of tracking continues to exist nationwide (Oakes, Gamoran, & Page, 1992; Natriello, 1994; Lucas, 1999; Oakes, 2005). At the level of the classroom attempts at differentiation include both strategies of grouping and individualization. For example, an early paper by Glaser (1977) recommended a system combing several modes of instruction suitable for individuals and small groups, ranging from tutoring to cooperative learning. These included managerial steps to be taken by teachers – curriculum planning, pacing, task assignment, and monitoring, into a program designed to encompass an entire school day.

Efforts to facilitate learning growing out of a content model perspective have been rooted in attempts to generate some understanding of how content can be arranged to facilitate learning through mastery of material. Such efforts connected with the development of adaptive learning opportunities are only a small and recent subset of the full range of approaches connected with preparing content as part of educational processes and systems. These approaches have been influenced by variety of forces and factors over the years. Some influences over the school curriculum may be described as the result of social processes related to the general organization of societies, such as the prominence of religions (Young, 1971; Bernstein, 1975; Meyer, Kamens, & Benovot, 1994). Other influences have a more decidedly political dimension, such as the roles of public and private entities (Whitty & Young 1976; Eggleston, 1977; Cody 1990). The content of education has also been shaped by professional perspectives, both those proffered by the academic disciplines (Gumport & Snydman, 2002) and those emanating from professional educators (Tyler 1969).

Within education traditional curriculum development has relied on logical examination of content or professional judgment to organize sequences of materials for students to move through as a way of developing competence in a subject. More recently, efforts have been made to base the development of such content sequences on empirical studies of student experiences moving through particular content sequences or along particular paths. These empirically based sequences are referred to as “learning progressions” (Corcoran, Mosher, & Rogat, 2009), and they constitute an evidence-based method for

28

identifying effective patterns for the organization and presentation of content.

The various factors influencing the nature of content or curriculum highlight the malleability of content, ranging from decisions governing the selection of content to be included in any learning system or environment, to the shaping, sequencing, and leveling of content within a curriculum, to the matching of content to particular students or groups of students. It is this malleability of content that can be an advantage within an adaptive learning system. Content can be arranged in any number of ways, and the content model perspective sensitizes us to the prospects for arranging content to achieve particular learning ends.

Web-based education seems to offer some particularly useful options for adapting content to students, and we have discussed much of this work in earlier sections. Here we mention only a few approaches that seem particularly connected to content and its use to foster learning.

Brusilovsky, Eklund, and Schwarz (1998) suggest that course material becomes more flexible through the use of adaptive hypermedia systems (AHS), which allow students access to personalized content and personalized order of presentation. This flexibility manifests in adaptive hypermedia systems through a network of static hypertext pages, some of which include the use of media. Adaptive hypermedia systems present material in two ways: as adaptive presentation and as adaptive navigation support (Brusilovsky, Eklund, & Schwarz, 1998).  Adaptive presentation adapts the page content to such individual characteristics as student knowledge and goals. Adaptive navigation support assists students in finding the best path, based on their growing subject matter knowledge, through the learning material.  An example of adaptive hypermedia system technology is when a user is given an online presentation in a course that caters to his or her knowledge of the topic and then receives relevant and learning-level appropriate links that provide further information on the subject with which students can follow-up (Brusilovsky, 2003).  Any one or more of many curricular organizational frameworks can be employed in creating adaptive systems. A particularly intriguing approach might be to combine the empirical methods of learning progressions with the meta-adaptive strategy advocated by Brusilovsky (2003). This would allow empirically derived paths for learning to be associated with learners with different profiles so that the method of adaptation itself is adapted.

Beyond distributing content and navigating content in ways that adapt to students, a more micro strategy focuses on the adaptation of small parts of text itself in response to learners. Researchers have created a development framework (Biedert, Buscher, Schwarz, Moller, Dengel, & Lottermann, 2010) and a reading device (Biedert, Buscher, & Dengel, in press) that uses eye tracking data to adapt text to the reader to create a more powerful reading experience. Adaptations include elements such as annotations keyed to the sections of text being read, the addition of sounds and images, alterations in color, and translations. Because the connections between eye movements and cognitive

29

processes are well established, future developments in eye tracking equipment and algorithms that support immediate text adaptations and the tailoring of text on-the-fly may increasingly come to characterize the reading experience. Such micro adaptations of content can be linked to broader content models and curriculum strategies.

Earlier we highlighted some of the dilemmas for controlling content and its presentation in learning environments posed by the open environment of the web. Indeed, as communication networks and utilities advance, it is increasingly clear that not even the content of live talks can be constrained from network effects (Parry, 2009). With learners increasingly in control of the content that supports their learning, the pressure to develop content perspectives and models that support high performance learning will continue to grow. It seems that it is no longer possible or worthwhile to attempt to wall out content; the only strategy is to create more engaging and persuasive curricular experiences.

F. The Teaching Perspective

Work oriented by the teaching perspective has focused on developing insight into differences in students that can be attuned to a broadened repertoire of instructional strategies to achieve optimal performance. At least three broad approaches can be considered. First, there have been numerous efforts over decades of research on teaching and instruction that examine the interaction of instructional methods and student abilities. Second, more recent research addresses the micro strategies of teaching that can be used to adapt to student differences in real time in the context of real classrooms, assessing their efficacy and feasibility. Third, technology based solutions to creating adaptive learning opportunities derived from instructional models have taken the form of intelligent tutoring systems (Winne, 1989).

A substantial body of work follows the logic of research in the tradition of aptitude-treatment interactions (ATIs) (see Cronbach, 1957; Cronbach & Snow, 1977). ATI studies conducted throughout the sixties and seventies illustrated the importance of varying instructional methods for students of different ability and motivation levels; this research also showed that a lack of main effects in studies of instructional treatments could frequently be explained by aptitude (student characteristics such as ability and motivation) treatment interaction effects. Cronbach & Snow (1977) define aptitude as any individual characteristic that increases or impairs the student's probability of success in response to instruction. Aptitude generally comprises three major categories - cognitive abilities, which includes intellectual abilities and declarative knowledge; affective/conative aptitude, which includes motivation, such as interests and efficacy, and action-control constructs, and affective factors such as temperament and mood; and physical and psychomotor abilities, including sensory-perceptual abilities. The classic “ATI hypothesis” is that some portion of students at a high end of a given aptitude measure may be found to do well with a given form of instruction, while other students,

30

scoring at lower levels, do poorly. The logical implication from an ATI perspective is that specific instructional procedures and strategies should be adapted by teachers to specific student characteristics. However, implementing an ATI-based approach to adaptation requires identifying the most relevant learner characteristics (or aptitudes) for a given form of instruction and then selecting instructional strategies that best facilitate the learning processes of particular students exhibiting those aptitudes. Unfortunately, at the conclusion of their exhaustive review of hundreds of ATI studies, Cronbach and Snow (1977) concluded that there were so few student aptitudes able to be linked systematically to particular methods, that matching was an impractical goal.

Other scholars have developed theoretical notions about linkages between different learner aptitudes and corresponding appropriate instructional strategies that will facilitate learning (see, e.g., Gagne, 1967; Gallagher, 1994). These ideas have resulted in the proposal of several broad strategies, guidelines and activities that instructors can adopt based on the unique aptitudes of different students involved in a given learning context. For example, Jonassen (1988) proposed a taxonomy of instructional strategies theorized to correspond to different processes of cognitive learning.

Another issue that came to light in the review of research in the ATI tradition is that ATI research tended to lay emphasis on exploring simple input/output relations between measured learner characteristics and learning outcomes based on some instructional interventions (Tobias, 1989). However, individual difference variables are difficult to measure; therefore, results may not only be questionable in terms of validity, they may also not be generalizable to other areas not covered by the particular research studies. Indeed, (Carrier & Jonassen, 1988) recommend that teachers become proficient teaching subject matter using several types of instructional strategies, pointing out that since several individual differences between students do influence learning, selecting a specific strategy for a given situation will inevitably be impractical; the outcomes may be different for different instructional contexts. Research is ongoing, however, as Park & Lee (2004) indicate, "the outlook is not optimistic for the development of a comprehensive ATI model or set of principles for developing adaptive instruction that is empirically traceable and theoretically coherent in the near future" (p. 661).

Some researchers have attempted to develop more micro-adaptive instructional models using on-task measures of individual student behavior and performances. The belief is that diagnosing students' learning preferences, needs, and styles through on-task monitoring of how they learn including cognition, behavior, and emotional states, can lead to instructional strategies based on these diagnoses. This then is the basis for making adaptive instructional decisions, which does not rely on holistic models and guidelines prescribed prior to tasks. This micro-level approach is hypothesized to ensure learning outcomes by continuously monitoring, manipulating and tailoring instructional treatments based on the diagnosis of student learning concerns, information processing, and behavior during each ongoing learning cycle. One-on-one tutoring best exemplifies the model underlying this micro-level adaptive instruction approach, and indeed the

31

effectiveness of one-on-one tutoring in promoting learning has been empirically supported (Bloom, 1984; Winne & Nesbit, 2009).

As Corno (2008) notes in her discussion of adaptive teaching, one-on-one tutoring or any form of true individualization is practically impossible to implement in current formal educational settings (e.g., public school classrooms) where there are both inadequate resources and few available qualified tutors. Corno described how practicing teachers make such micro-adaptations on an ongoing basis, based on their own formations of student strengths and weaknesses. Other research programs have come at the micro-level adaptation question by capitalizing on advances in computing and communication technologies. These bodies of research are aimed at developing instructional models that include the use of new technologies in ways that are expected to support effective individualized learning and instructional activities.

Technology based solutions grow out of a history of work on computer assisted instruction. Computer assisted instruction/learning (CAI/CAL) marked an effort to use computers to enhance learning based upon the psychological theories extant at different periods of time. For example, in the 1920’s Sidney Pressey created a machine that claimed it “tested and also taught” (Skinner, 1986). This machine directed the student through a multiple choice test, and monitored responses. If the student chose the correct answers the machine progressed to a higher level. However, if some responses were determined to be incorrect, the student had the opportunity to repeat the test, but the machine only presented the items that the student got incorrect on the first trial. This simple tracking of a student's responses was considered a form of adapting future content to students' personal areas of weakness. This idea remained foundational to teaching machines as they developed.

In the 1950's, as Skinner’s idea of behaviorism was reflected not only in instructional methods in schools but also in design frameworks for educational technology systems, learning processes were generally conceptualized as forming connections between stimuli and responses, hence the drive towards the reward and punishment frameworks (Bransford, Brown, & Cocking, 1999). Skinner himself dabbled with teaching machines and showcased in 1954 his arithmetic machine at the University of Pittsburgh (Skinner, 1986). The machine was an awkward box that had a knob that students turned to have questions scroll into a designated area with sliders that the students used to create their numerical answers. Once they were done with answering one question, they would try to turn the knob and get another question, and if they had placed the sliders in the correct places (a correct numerical answer) then they were given the next question. If what they produced was wrong, they had to keep repeating the process of placing the sliders until the right response was obtained before they could continue. This machine built upon Pressey's (1926) earlier work in two important ways. First, it was not only a test taking machine, in that students who had not previously studied material could also use it, and so the machine both tested and taught. Second, the students composed their responses instead of choosing from a preexisting set, and the material was arranged hierarchically,

32

such that once learners completed the questions in the first frame, they were prepared to deal with questions in the next (Skinner, 1986).

Early prototypes of these programmable learning machines created during this time were "linear programs," in part because they neglected to provide rich feedback or individualization. They were not designed to develop a model of whom they were teaching, what they were teaching, and how to teach it (Nwana, 1990). The programs were organized into frames that were presented to the learner and were set up to take the learner toward a desired behavior. As the frames were presented the learner had to complete very simple operations such as filling in a blank space, and was immediately given knowledge of results. Regardless of the student response, frames would continue in a predetermined order, thus the computer was really doing little more than what a programmed textbook could do (Nwana, 1990), and early studies of student motivation with programmed instruction showed that boredom led to skipped items and lost interest.

The next step in the evolution of these machines involved taking the individual learner into consideration by regarding their input in terms of content and actions. This idea led to what is known as “branching programs” that modify the interaction between the system and the user by being able to comment on a response and also use it to choose a next frame. Branching programs still used the frame design but made decisions on what material should be shown to the student based on their knowledge and actions. Pattern matching techniques thus allowed learners' answers to be considered on an acceptability scale rather than the correct/incorrect model of the Skinnerian systems (Nwana, 1990).

The next development along the spectrum of intelligent tutoring grew out of the recognition that teaching material could itself be generated by the computer. This school of thought emerged by the late 1960’s and the corresponding systems that were developed were known as “generative” or “adaptive systems” (Nwana, 1990). Such systems were able to generate meaningful problems and monitor their solution as well. For instance, Urh (1969) implemented a series of such systems that generated problems in arithmetic that were tailored to a student's performance. Suppes (1967), and Woods and Hartley (1971) also produced systems with similar capabilities.

Generally however, these "generative" systems also relied on a frame-oriented design in which the course author specified the formats. The system generated parameters for these formats and upon receiving a response to a particular question, searched a database and presented students with information on whether the answer was right or wrong. Due to the obvious shortcomings of these systems, their usage was limited to domains with well-structured content like mathematics that allowed for drill-type exercises (Nwana, 1990).

Such computer assisted instruction provided the foundation for the development of intelligent tutoring systems (Winne, 1989). Through further development the systems gradually improved on their individualization and feedback features, though critics maintained that the systems could not fully mimic human knowledge (Winne, 1989;

33

Nwana, 1990), and also failed to incorporate many valuable learning principles and instructional strategies (Park, Perez, & Siedel, 1987). To such critics, computer tutoring systems should have a representation of what is being taught, who is being taught and how such a person should be taught. Thus, computer programs called Intelligent Tutoring Systems (ITSs) were devised, and these aimed to incorporate techniques from artificial intelligence (AI) in order to provide "tutors" that know what they teach and how to teach it in an ‘intelligent’ way (Nwana, 1990). Apart from the practical ideals of one-on-one customized tutoring and immediate feedback to learners, intelligent tutoring systems were thought to be more economical in the long run, and this contributed to heightened interest in these systems.

Research and development on intelligent tutoring systems spans three different disciplines - computer science, cognitive psychology and educational research, and this results in major differences in research prototypes and goals, terminologies, theoretical frameworks, and emphases among researchers (Nwana, 1990). Currently, prototype and operational intelligent tutoring systems provide practice-based instruction to support individualized learning in K-12 and college education, although on a limited scale.

Work from the teaching or instructional perspective shows promise along several lines. First, there seems to be the potential for intelligent tutoring systems to be developed and refined to include more complete adaptive elements and meta-adaptive dimensions. Second, strategies to prepare and equip classroom teachers to teach adaptively may be supported by the development of technology-based assistive devices and adaptive teaching tools and systems. We highlight some recent developments in each area to illustrate the potential for further progress from the teaching perspective.

Although current uses of intelligent tutoring systems remain modest, there is a good deal of ongoing research designed to make such systems more effective components of the learning process (Self, 1999). Recently, intelligent tutoring systems have moved towards reading the user's emotional state and conveying the emotional state of the intelligent tutor (Nkambou, 2006; Sarrafzadeh, Alexander, Dadgostar, Fan, & Bigdeli, 2008). The idea is to for the system to “sense” when learners are frustrated, confused, or angry and in need of remedial assistance and to respond appropriately, including an affective dimension. This added adaptivity is expected to improve students’ affective response or satisfaction as well as their performance. Other work to increase the sensitivity and responsiveness of intelligent tutoring systems is drawing on evolving techniques of data mining to (Nkambou, Fournier-Viger, & Nguifo, 2009) to take into account the wider array of data generated by students in the act of problem solving.

Efforts to support classroom teachers to adapt their instructional approach to individual learners have relied on professional development (e.g., Wang & Gennari, 1983; Wang, Rubinstein, & Reynolds, 1985), but as Corno (2008) notes, the conditions found in most schools make individualization at the level envisioned by fully adaptive teaching impossible. Fortunately, the same advances in computing and communications

34

technologies that support digital learning environments can also be employed to create a new generation of tools to support classroom teachers in successful adaptations. The in-classroom use of handheld computing and communications devices is one early example of a means for creating opportunities for teachers to engage in dynamic assessment of student performance that supports their efforts to adapt instruction to individual students (Penuel & Yarnall, 2005; Hupert Heinze, Gunn, & Stewart, 2008).

G. The Social Network Perspective

Social relationships among individuals have always provided major opportunities for learning, and, indeed, the social groups in which we are all embedded have long been used to explain differences in educational attainment (Lin, 1999). Studies of the structures and processes of human social networks have been joined more recently by studies of the network processes that are growing along with new developments in computing and communications technologies. Moreover, the network possibilities connected with the Internet and the world wide web have brought forth a new generation of research on network processes and properties (Barabasi, 2003; Watts, 2004). Technology mediated or supported social networks offer the capacity to connect learners with other individuals who can support and enhance their learning. These networks present new possibilities and opportunities for adaptive learning.

There are at least three streams of activity that bear on adaptive learning opportunities with varying degrees of intention and structure. First, the increasing size and density of the Internet and the world wide web has the potential to offer expanded learning opportunities. Second, within the broader Internet there are naturally forming online communities that offer rich collections of individuals who can be resources for the learning of others within the same community. Third, there are deliberately assembled and purposively structured online communities of individuals committed to facilitating the learning of one another.

First, with the growth of the number of individuals actively online, there is an enormous set of individuals who may serve as resources to support the learning of others on the network. These individuals are not necessarily clustered within the network in proximity to any particular learner in need of assistance. Moreover, there is reason to believe that opportunities for learning from less directly connected individuals may be powerful (Granovetter, 1973). Opportunities for enhancing the learning value of connections on the open Internet seem to lie with the development of new understandings of the operations of structure and processes operating within networks, including those factors that affect the distribution of resources and access to them (Barabasi, 2003; Watts, 2004). If we are to enhance the learning value of elements of a network, we need to develop strategies for exposing the learning opportunities available in various resources (Alesso & Smith, 2009; Antiniou & van Harmelen, 2008), as well as more powerful search utilities to connect learner needs to far flung individuals (Morville, 2005).

35

Second, there are naturally forming communities within the larger Internet that offer more specialized collections of individuals. These communities form around common interests and pursuits such as common problems or professional interests (Wenger, 1999). With the rise of social networking services such as MySpace, Facebook, and Flickr, the opportunities for social processes leading to exchange of relevant information appear to be increasing. Such networks convey various benefits to participants (Ellison, Steinfield, & Lampe, 2007), including learning opportunities (Constant, Sproull, & Kiesler, 1996). In these networks the process of assembling the appropriate resources predates particular efforts to access the resources for an immediate learning purpose. Nonetheless, with the continuing development of technologies for enhancing both the quality of information on resources and the quality of information on individual users and their learning needs (Breslin, Passant, & Decker, 2009), the opportunities to facilitate learning are increasing in substantial ways.

Third, there are more deliberately assembled online communities that create rich collections of individuals who can be accessed to assist each other’s learning. These more deliberate efforts encompass a wide range of activities from targeted research and development programs to the rapidly growing number of online courses and learning environments. These efforts all share the explicit goal of facilitating learning. Stahl, Koshmann, & Suthers (2002) and Lee and Park (2007) discuss the range of research and development efforts within the field of education that seek to enhance learning opportunities through the support of interaction and collaboration. Bruckman (2002) describes the learning possibilities of online communities deliberately organized for educational purposes.

Focusing more directly on the adaptive learning dimensions of deliberately assembled educational collectivities, Brusilovsky and Peylo (2003) discuss a group of technologies that they term “intelligent collaborative learning.” They view these approaches as lying at the crossroads of the fields of computer supported collaborative learning and intelligent tutoring systems. Brusilovsky and Peylo (2003, pp. 164-165) identify three distinct technologies underlying this work. Adaptive group formation and peer help approaches identify appropriate matches to support the specific learning needs of individuals and groups by drawing on knowledge about potential collaborating peers contained in their student models. Masthoff (2008) notes the task of dividing a group into smaller teams for optimal collaboration as a target for such support as well as the task of identifying a suitable partner for working together.

Adaptive collaboration support approaches draw on information about more or less successful collaboration patterns taken from communication logs to provide interactive support for collaboration processes. Masthoff (2008) distinguishes further between interaction support and decision support. Interaction support may

36

involve promoting equal and appropriate contributions through modeling and incentives as well as assisting learners in asynchronous collaboration environments, and sorting through a large number of contributions made in their absence by providing relevance information. Decision support may involve modeling common knowledge among group members, using individual user model information to guide adaptation to groups, and facilitating negotiation among group members.

Virtual student approaches rely on the introduction of virtual peers into the learning environment to support student learning. For example, Uresti and du Boulay (2004) introduced a learning companion requiring instruction and review from real students to promote their engagement. Gulz (2004) examined the evidence available on the learning impact of virtual students as of 2004, describing the overall results for the early research as uncertain. She identified a number of avenues for additional research as this approach is developed further.

CONCLUSION

As we have said, any parceling of the broad set of activities that may contribute to the evolution of adaptive learning opportunities is necessarily somewhat arbitrary. The seven perspectives we have defined are at times over lapping, and many adaptive learning activities will draw on multiple perspectives. Nevertheless, the seven perspectives identified provide a starting point for our efforts to scour the terrain of activities that might be brought into a conversation to advance thinking about adaptive learning possibilities. They may also serve to suggest new avenues for development, both for those starting on the path of developing adaptive applications and practices and for those who are already deeply engaged in such development.

The sheer breadth and extent of the development of adaptive learning applications suggests that we are ready to take the next step to bring such applications and their creators into a sustainable conversation with both the increasingly multi-disciplinary community of developers and with practicing educators in the evolving education and learning sector. The question before us now is how best to facilitate such a conversation. Our next step is to engage key constituencies (e.g., developers, researchers, publishers, educators, learners) in activities to develop prototype designs for these and other elements of the infrastructure for facilitating the growth of adaptive learning opportunities.

We believe that the current period is ripe for fundamental change in the education sector. Indeed, such change seems likely to involve what Lawrence Cremin termed the “configurations of education,” that is, the ways in which the various institutions involved in education interact with each other and with other major social sectors. If we are to

37

foster the development, refinement, delivery, and use of adaptive learning applications on a large scale, we might need to imagine one or more new institutional entities added to the configurations of the contemporary education sector. And we will certainly need to imagine changes in the relationships among the parts of the sector. It is time to begin, and we invite you to join us.

REFERENCES

Akhras, F. (2005). Modelling the context of learning interactions in intelligent learning environments. In A. Dey, B. Kokinov, D. Leake, & R. Turner (Eds.), Modeling and Using Context 5th International and Interdisciplinary Conference CONTEXT 2005 Paris, France, July 5-8, 2005 Proceedings. Berlin: Springer.

Akhras, F. & Self, J. (2002a). Beyond intelligent tutoring systems: Situations, interactions, processes and affordances. Instructional Science, 30, 1-30.

Akhras, F. & Self, J. (2002b). Situations, interactions, processes and affordances: A response. Instructional Science, 30, 65-69.

Alesso, H. & Smith, C. (2009). Thinking on the web: Berners-Lee, Godel, and Turing. Hoboken, NJ: Wiley.

Aleven, V., Stahl, E., Schworm, S., Fischer, F., and Wallace, R. (2003). Help seeking and help design in interactive learning environments. Review of Educational Research 73(2), 277-320.

Aleven, V., McLaren, B., Roll, I., and Koedinger, K. (2004). Toward tutoring help seeking: applying cognitive modeling to meta-cognitive skills. In Proceedings of the 7th Internation Conference on Intelligent Tutoring systems, 227-239.

Antoniou, G. & van Harmelen, F. (2004). A semantic web primer. Cambridge, MA: MIT Press.

Azevedo R. (2002). Beyond intelligent tutoring systems: Using computers as METAcognitive tools to enhance learning? Instructional Science, 30, 31-45.

Barabasi, A. (2003). Linked. New York: Penguin.

38

Baron, B., Schwartz, D., Vye, N., Moore, A, Petrosino, A., Zech, L., & Bransford, J. (1998). Doing with understanding: Lessons from research on problem and project-based learning. Journal of the Learning Sciences, 7(3-4), 271-311.

Beal, C., Walles, R., Arroyo, I. & Woolf, B. (2007) On-line tutoring for math achievement testing: A controlled experiment. Journal of Interactive Online Learning, 6(1), 43-55.

Bernstein, B. (1975). On the classification and framing of educational knowledge. In B. Bernstein, Class, Codes and Control, volume 3, Towards a Theory of Educational Transmissions. London: Routledge and Kegan Paul.

Bertalanffy, L. Von. (1968). General system theory: Foundations, development, applications. New York: George Braziller.

Betrus, A.K. (2002). Individualized instruction. Encyclopedia of Education. Ed. James W. Guthrie. Vol. 4. 2nd ed. New York: Macmillan Reference USA, 2002. 1129-1133.

Biedert, R., Buscher G., & Dengel A. (in press). The eyebook: Using eye tracking to enhance the reading experience. Informatik-Spektrum,

Biedert, R. Buscher, G., Schwarz, S., Moller, M., Dengel, A. & Lottermann, T. (2010). The Text 2.0 Framework. In: Proc. of the International Workshop on Eye Gaze in Intelligent Human Machine Interaction, Hong Kong, China: February.

Bloom, B. S. (1984). The 2 sigma problem: The search for methods of group instruction as effective as one-on-one tutoring. Educational Researcher, 13, 4-16.

Blumenfeld, P., Soloway, E., Marx, R., Krajcik, J., Guzdial, M. & Palincsar, A. (1991). Motivating project-based learning Sustaining the doing, supporting the learning. Educational Psychologist, 24, 369-398.

Boekaerts, M., Pintrich, P., and Zeidner, M. (2000). Handbook of self-regulation. San Diego: Academic Press.

Boyer, E.L. & Mitgang, L.D. (1996). Building community: A new future for architecture education and practice. A special report. Princeton, NJ: Carnegie Foundation for the Advancement of Teaching.

Boykin, A.W. (1994). Afro cultural expressions and its implications for schooling. In E. Hollins, J. Kind, & . Wayman, (Eds.), Teaching diverse populations: Formulating a knowledge base. Albany: State University of New York Press.

Bransford, J., Sherwood, R., Hasselbring, T., Kinzer, C. & Williams, S. (1990). Anchored instruction: Why we need it and how technology can help. In D. Nix & R. Spiro (Eds.),

39

Cognition, education, and multimedia: Exploring ideas in high technology (pp. 115-141). Hillsdale, NJ: Lawrence Erlbaum Associates.

Bransford, J. D., Brown, A. L., & Cocking, R. R. (1999). How people learn: Brain, mind, experience and school. Washington DC.: National Academy Press.

Breslin, J., Passant, A., & Decker, S. (2009). The social semantic web. New York: Springer.

Bruckman, A. (2002). Learning in online communities. In R. K. Sawyer, (ed.) The Cambridge handbook of the learning sciences (pp. 461-475).

Brusilovsky, P. (1996). Methods and techniques of adaptive hypermedia. User modeling and user-adapted interaction, 6, 87-129.

Brusilovsky, P. (1998). Methods and techniques of adaptive hypermedia. Pp. 1-43 in P. Brusilovsky, A. Kobsa, and J. Vassileva (eds.). Adaptive hypertext and hypermedia. Dordrecht: Kluwer,

Brusilovsky, P. (2001). Adaptive hypermedia. User modeling and user-adapted interaction, 11, 87-110.

Brusilovsky, P. (2003). Adaptive navigation support in educational hypermedia: the role of student knowledge level and the case for meta-adaptation. British Journal of Educational Technology, 34(4), 487-497.

Brusilovsky, P. (2004). Adaptive navigation support: From adaptive hypermedia to the adaptive web and beyond. PsychNology Journal, 2(1), 7-23.

Brusilovsky, P., Eklund, J., & Schwarz, E. (1998). Web-based education for all: A tool for development of adaptive courseware. Computer Networks and ISDN Systems, 30, 291-300.

Brusikovsky, P. & Peylo, C. (2003). Adaptive and intelligent web-based educational systems. International Journal of Artificial Intelligence in Education, 13, pp. 159-172.

Brusilovsky, P. and Maybury, M.T. (2002). From adaptive hypermedia to the adaptive web. Communications of the ACM, 45(5), 30-33.

Brusilovsky, P. and Rizzo, R. (2002) Map-based Access to Multiple Educational On-Line Resources from Mobile Wireless Devices, In: Paterno, F. (ed.) Mobile Human-Computer Interaction. Lecture Notes in Compuer Science, Vol. 2411, (Proceedings of 4th International Symposium on Mobile Human-Computer Interaction, Mobile HCI 2002), Pisa, Italy, September 18-20, 2002, Berlin: Springer-Verlag, pp. 404-408.

40

Buckley, J. & Schneider, M. (2009). Charter schools: Hope or hype. Princeton, NJ: Princeton University Press.

Bush, V. (1945). As we may think. Atlantic Monthly. Available at:http://www.theatlantic.com/doc/194507/bush

Butler, D. and Winne, P. (1995). Feedback and self-regulated learning: A theoretical synthesis. Review of Educational Research, 65, 245-281.

Carrier, C., & Jonassen, D. H. (1988). Adapting courseware to accommodate individual differences. In D. Jonassen (Ed.), Instructional design for microcomputer courseware. Mahwah, NJ: Lawrence Erlbaum Associates.

Carroll, J.B. (1993). Human cognitive abilities. New York: Cambridge University Press.

Challis, D. (2009). Committing to quality learning through adaptive online assessment, Assessment & Evaluation in Higher Education, 30(5), 519-527.

Cheng, Y. (2009). When cognitive diagnosis meets computerized adaptive testing: CD-CAT. Psychometrika, 74(4), 619-632.

Chrusciel, D. (2006). The role of curriculum in organizational significant change planning. The Learning Organization, 13(3), 215-229.

Cody, C. (1990) The politics of textbook publishing, adoption, and use. Pp. 127-145 in D. Elliott & A. Woodward (eds.) Textbooks and Schooling in the United States. Eighty-Ninth Yearbook of the National Society for the Study of Education. Part I. Chicago: University of Chicago Press.

Computing Research Association. (2005). Cyberinfrastructure for Education and Learning for the Future: A Vision and Research Agenda. Washington, DC: Computing Research Association.

Constant, D., Sproull, L., & Kiesler, S. (1996). The kindness of strangers: The usefulness of electronic weak ties for technical advice. Organization Science, 7(2), 119-135.

Corcoran T., Mosher, F. & Rogat, A. (2009). Learning Progressions in Science: An Evidence-Based Approach to Reform. New York: Center on Continuous Instructional Improvement, Teachers College, Columbia University.

Corno, L. (2008). On teaching adaptively. Educational Psychologist, 43(3), 161-173.

41

Cremin, L. (1972). American education: The colonial experience. New York: HarperCollins.

Cronbach, L. J. (1957). The two disciplines of scientific psychology. American Psychologist, 12, 671-684.

Cronbach, L. J., & Snow, R. E. (1977). Aptitudes and instructional methods: A handbook for research on interactions. New York: Irvington.

De Bra P. & Brusilovsky, P. (2009). Introduction to Special Issue on Adaptive Hypermedia. New Review of Hypermedia and Multimedia, 15(1), 1-3.

De Vicente, A. and Pain, H. (2002). Informing the detection of the students’ motivational state: an empirical study. Lecture Notes in Computer Science, 2362, 933-943.

Deci, E.L., Koestner, R., & Ryan, R.M. (2001). Extrinsic rewards and intrinsic motivation in education: Reconsidered once again. Review of Educational Research, 71(1), 1-27.

Dornbusch, S.M. & W.R. Scott. (1975). Evaluation and the exercise of authority. San Francisco: Jossey-Bass.

Edwards, D. & Mercer, N. (1987). Common knowledge: The development of joint understanding in the classroom. New York: Metheun.

Eggleston, J. (1977). The politics of curriculum knowledge. Pp. 141-150 in J. Eggleston, The Sociology of the School Curriculum. London: Routledge and Kegan Paul.

Ellison, N. B., Steinfield, C., & Lampe, C. (2007). The benefits of Facebook "friends:" Social capital and college students' use of online social network sites. Journal of Computer-Mediated Communication, 12(4), article 1. http://jcmc.indiana.edu/vol12/issue4/ellison.html

Gagne, R. M. (1967). Learning and individual differences. Columbus, OH: Merrill.

Gallangher, J. J. (1994). Teaching and learning: New models. Annual Review of Psychology, 45(171-195).

Gardner, H. (2009). Personalized education: A quantum leap in learning will allow everyone to go to the head of the class. Foreign Policy, 172, 86.

Glaser, R. (1977). Adaptive education: Individual diversity and learning. New York: Holt, Rinehart & Winston.

42

Gordon, H. (2007). The History and Growth of Career and Technical Education in America. Long Grove, IL: Waveland Press.

Granovetter, M. (1973). The strength of weak ties. American Journal of Sociology, 78, 1360-1380.

Greene, J. A., & Azevedo, R. (2009). A macro-level analysis of SRL processes and their relations to the acquisition of sophisticated mental models. Contemporary Educational Psychology, 34, 18–29.

Grubb, W. (1995). Education Through Occupations in American High Schools. Vol. 1. New York: Teachers College Press.

Gumport, P.J. & Snydman, S.K. (2002). The formal organization of knowledge: An analysis of academic structure. The Journal of Higher Education, 73, 3, pp. 375-408.

Harel, I. & Papert, S. (eds.) (1991). Constructionism. Norwood, NJ: Ablex.

Henze, N. & Nejdi, W. (1999). Adaptivity in the KBS hyperbook system. In 2nd Workshop on Adaptive Systems and User Modeling on the WWW. Toronto, Canada.

Henze, N., Nejdi, W., & Wolpers, M. (1999). Modeling constructivist teaching functionality and structure in the KBS hyperbook system. Proceedings of the 1999 conference on Computer support for collaborative learning. Palo Alto, California.

Hupert, N., Heinze, J., Gunn, G., & Stewart, J. (2008). In E. Mandinach & M. Honey, Eds. Linking data and learning. (pp. 130-150). New York: Teachers College Press.

Hurley, E.A., Allen, B.A., & Boykin A.W. (2009). Culture and the interaction of student ethnicity with reward restructure in group learning. Cognition and Instruction, 27(2), 121-146.

Illich, I. (1972). Deschooling society. New York: Harper and Row.

Jonassen, D. H. (1988). Integrating learning strategies into courseware to facilitate deeper processing. In D. H. Jonassen (Ed.), Instructional design for microcomputer courseware. Mahwah, NJ: Lawrence Erlbaum Associates.

Jonassen, D. (1999). Designing constructivist environments. Pp. 215-240 in C.M. Reigeluth (ed.). Instructional-Design Theories and Models: A New Paradigm of Instructional Theory, Vol. II. Mahwah, NJ: Erlbaum.

Kafai, Y. & Resnick, M. (eds.) (1996). Constructionism in Practice. Mahwah, NJ:

43

Erlbaum.

Knutov, E., De Bra, & Pechenizkly, M. (2009). AH 12 years later: A comprehensive survey of adaptive hypermedia methods and techniques. New Review of Hypermedia and Multimedia, 15(1), 5-38.

Kreitzberg, C.B., Stocking, M.L, & Swanson, L. (1978). Computerized adaptive testing: Principles and directions. Computers and Education, 2, 319-329.

Ladson-Billings, G. & Brown, K. (2008). Curriculum and cultural diversity. Pp. 153-175 in F. M. Connelly, M. F. He, & J. Phillion, (ed.). The Sage Handbook of of Curriculum and Instruction. Thousand Oaks, CA: Sage.

Lajoie, S. P. & Azevedo, R. (2006). Teaching and learning in technology-rich environments. In P. Alexander & P. Winne (Eds.), Handbook of educational psychology (2nd ed.) (pp. 803-821). Mahwah, NJ: Erlbaum.

Levitt, G.M. (2006). The turk: Chess automaton. Jefferson, NC: McFarland and Company.

Lee, J., & Park, O. (2007). Adaptive instructional systems. In J.M. Spector, M.D. Merrill, J.V. Merrienboer, & M.P. Driscoll (Eds.), Handbook of research on educational communications and technology (3rd ed., pp. 469-484). Mahwah, NJ: Lawrence Erlbaum Associates.

Lilley, M., Barker, T, & Britton C. (2004). The development and evaluation of a software prototype for computer-adaptive testing. Computers and Education, 43, 109-123.

Lin, N. (1999). Social networks and status attainment. Annual Review of Sociology, 25, 467-487.

Lucas, S.R. (1999). Tracking inequality: Stratification and mobility in American high schools. New York: NY: Teachers College Press.Luecht, R., Brunfield, T. & Breithaupt, K. (2006). A testlet assembly design for adaptive multistage tests. Applied Measurement in Education, 19(3), 189-202.

Lupack, A. (2007). The Oxford guide to Arthurian literature and legend. New York: Oxford University Press.

Maeroff, G. (2003). A classroom of one: How online learning is changing schools and colleges. New York: Palgrave Macmillan.

Mallen, G. (2005). Reflections on Gordon Pasks adatpive teaching concepts and their

44

relevance to modern knowledge systems. Paper presented at the Creativity and Cognition London.

Mark Hopkins. (2009). In Encyclopædia Britannica. Retrieved September 06, 2009, from Encyclopædia Britannica Online: Available at: http:///www.britannica.com/EBchecked/topic/271551/Mark-Hopkins

Markoff, J. (2005). What the dormouse said: How the sixties counter-culture shaped the personal computer industry. New York: Penguin.

Masthoff, J. (2008). An exploration of adaptive collaboration support. Adaptive Collaboration Support Workshop Proceedings. Hanover, Germany: Fifth International Conference on Adaptive Hypermedia and Adaptive Web-Based Systems.

McGlohen, M. & Chang, H. (2008). Combining computer adaptive testing technology with cognitively diagnostic assessment. Behavior Research Methods, 40(3), 808-821.

Mead, A. (2006). An introduction to multistage testing. Applied Measurement in Education, 19(3), 185-187.

Meyer, John, David Kamens, and Aaron Benovot. (1992). School Knowledge for the Masses: World Models and National Primary Curricular Categories in the Twentieth Century. Washington, DC: Falmer Press.

Meyer, J., Tyack, D., Nagel, J., & Gordon, A. (1979). Public education as nation-building in America: Enrollments and bureaucratization in the American states, 1870-1930. American Journal of Sociology, 85(3), 591-613.

Mills, C.N. & Steffen, M. (2000). The GRE computer adaptive test: Operational issue. In W.J. Van der Linden, (ed.). Computerized adaptive testing: Theory and practice. (pp. 75-99). Hingham, MA: Kluwer Academic Publishers.

Morville, P. (2005). Ambiant findability. Sebastopol, CA: O’Reilly Media.

Natriello, G. (1994). Coming together and breaking apart: Unifying and differentiating processes in schools and classrooms. Research in Sociology of Education and Socialization, 10, 111-145.

Nguyen, D. & Kulm, G. (2005). Using web-based practice to enhance mathematics learning and achievement. Journal of Interactive Online Learning, 3(3).

Nirmalakhandan, N. (2006). Computerized adaptive tutorials to improve and assess problem-solving skills. Computers and Education, 49, 1321-1329.

45

Nkambou, R., Fournier-Viger, P., Nguifo, E. (2009). Improving the behavior of intelligent tutoring agents with data mining. IEEE Intelligent Systems, 24(3), 46-53.

Nwana, H. S. (1990). Intelligent tutoring systems: An overview. Artificial Intelligence Review 4, 251-277.

NSF Task Force on Cyberlearning. (2008). Fostering Learning in the Networked World: The Cyberlearning Opportunity and Challenge. Washington, DC: National Science Foundation.

Oakes, J. (2005). Keeping track: How schools structure inequality. (2nd edition). New Haven, CT: Yale University Press.

Oakes, J., Gamoran, A., & Page, R. N. (1992). Curriculum differentiation: Opportunities, outcomes, and meanings. In P. W. Jackson (Ed.), Handbook of research on curriculum. New York: MacMillan Publishing.

Oh, E. & Lim, D. (2005). Cross relationships between cognitive styles and learner variables in online learning environments. Journal of Interactive Online Learning, 4(1), 53-66.

Own, Z. (2006). The application of an adaptive web-based learning environment on oxidation-reduction reactions. International Journal of Science and Mathematics Education, 4, 73-96.

Papanikolaou, K. & Grigoriadou, M. (2006). Towards a constructivist approach in the design of adaptive educational systems. In P. Kommers & G. Richards (Eds.) Proceedings of World Conference on Educational Multimedia, Hypermedia and Telecommunications 2006 (ED-MEDIA), 2596-2601 Chesapeake, VA: AACE.

Papanikolaou, K., Gritoriadou, M., & Smarakou, M. (2005). Learning activities and aids in adaptive learning environments. Proceedings of the IADIS International Conference “Cognition and Exploratory Learning in the Digital Age.” Porto, Portugal, December 14-16.

Papanikolaou, K., Gritoriadou, M., Alexi, A., & Smarakou, M. (2005). Designing project-based adaptive educational systems. 2005 WSEAS International Conference on Engineering Education. Athens, Greece, July.

Papanikolaou, K, Grigoriadou, M., Kornilakis, H., & Magoulas, G.D. (2003). Personalising the interaction in a web-based educational hypermedia system: The case of INSPIRE. User-Modeling and User-Adapted Interaction, 13(3), 213-267.

46

Papanikolaou, K. & Grigoriadou, M. (2006). Towards a constructivist approach in the design of Adaptive Educational Systems. In E. Pearson & P. Bohman (Eds.), Proceedings of World Conference on Educational Multimedia, Hypermedia and Telecommunications 2006 (pp. 2596-2601). Chesapeake, VA: AACE. Retrieved from http://www.editlib/org/p/23374

Paris, S. and Paris, A. (2001). Classroom applications of research on self-regularted learning. Educational Psychologist, 36(2), 89-101.

Parry, M. (2009). Conference humiliation: They’re tweeting behind your backs. Chronicle of Higher Education, November 17. Retrieved from http://chronicle.com/article/Conference-Humiliation-The/49185/

Patel, A., Scott, B., & Kinshuk. (2001). Intelligent tutoring: from SAKI to Byzantium Kybernetes 30(5/6), 807-818.

Park, O., Perez, R.S., & Seidel, R.J. (1987). ICAI: Old wine in new bottles, or Is it a new vintage? In G. Kearsley (Ed..), Artificial intelligence and education. Reading, MA: Addison-Wesley. 

Penuel, W. & Yarnall, L. (2005). Designing handheld software to support classroom assessment: An analysis of conditions for teacher adoption. The Journal of Technology, Learning, and Assessment, 3(5). Available at: www.jtla.org

Petrina, A.M. (2008). Robotics: Present and future. Scientific and Technical Information Processing, 35(2), 73–79.

Pinar, W. (2004). What is curriculum theory? Mahway, NJ: Erlbaum.

Pintrich, P.R. & Schunk, D.H. (1996). Motivation in education: Theory, research, and applications. Englewood Cliffs, NJ: Prentice-Hall.

Plato. (2003). (Tarrant H., ed.; Tredennick H., trans.). The last days of Socrates. New York: Penguin.

Randi, J. & Corno, L. (1997). Teachers as innovators. In B.J. Bittle, T.L. Good, & I.G. Goodson (eds.), International handbook of teachers and teaching, Volume 1 (pp. 1163-1122). Dordrecht, The Netherlands: Kluwer.

Reese, W. (2005). America’s Public Schools: From the Common School to “No Child Left Behind.” Baltimore, MD: Johns Hopkins University Press.

Reiser, R. A. (1987). Instructional technology: A history. In R. Gangne (Ed.), Instructional Technology: Foundations. Mahwah, NJ: Lawrence Erlbaum Associates.

47

Rogoff, B. (2003). The cultural nature of human development. New York: Oxford University Press.

Rulison, K.L, & Loken, E. (2009). I’ve fallen and I can’t get up: Can high-ability students recover from early mistakes in CAT? Applied Psychological Measurement, 33(2), 83-111.

Sands, W.A. & Waters, B.K. (1997). Introduction to ASVAB and CAT. In W.A. Sands, B.K, Waters & J.R. McBride (eds.). Computerized adaptive testing: From inquiry to operation. (pp. 3-9). Washington, DC: American Psychological Association.

Sarrafzadeh, A., Alexander, S., Dadgostar, F., Fan, C., & Bigdeli, A. (2008). “How do you know that I don’t understand?” A look at the future of intelligent tutoring systems, Computers in Human Behavior, 24, 1342-1363.

Schaeffer, G.A., Steffen M., Golub-Smith, M.L., Mills, C.N., & Durso, R. (1995). The introduction and comparability of the computer adaptive GRE general test. Princeton, NJ: Educational Testing Service.

Schank, R.C., Fano, A., Bell, B. and Jona, M. (1993/94),. The design of goal-based scenarios. The Journal of the Learning Sciences, 3(4), 305-345.

Schiro, M. (2007). Curriculum Theory: Conflicting Visions and Enduring Concerns. Thousand Oaks, CA: Sage.

Schunk, D. and Zimmerman, B. (1994). Self-regulation of learning and performance. Hillsdale, NJ: Erlbaum.

Self, J. (1999). The defining characteristics of intelligent tutoring systems research: ITSs care, precisely. International Journal of Artificial Intelligence in Education, 10, 350-364

Sergiovanni, T. (1994). Building community in schools. San Francisco: Jossey-Bass.

Shaffer, D.W. (2006). How computer games help children learn. New York: Palgrave Macmillan.

Shaffer, D.W. (2006). Pedagogical praxis: The professions as models for post-industrial education. Teachers College Record, 106(7), 1401-1421.

Shirky, C. (2008). Here comes everybody: The power of organizing with organizations. New York: Penguin.

48

Shute, V.J. & Zapata-Rivera, D. (2007). Adaptive technologies. In J.M. Spector, M.D. Merrill, J.V. Merrienboer, & M.P. Driscoll (Eds.), Handbook of research on educational communications and technology (3rd ed., pp. 279-294). Mahwah, NJ: Lawrence Erlbaum Associates.

Skinner, B. F. (1986). Programmed instruction revisited. Phi Delta Kappan, 68, 103-110.

Snow, R. [in collaboration with E. Yalow]. (1982). Education and intelligence. In R.J. Sternberg (ed.) Handbook of human intelligence (pp. 493-596). London: Cambridge University Press.

Specht, M., Weber, G., Heitmeyer, S., & Schöch, V. (1997). AST: Adaptive WWW-courseware for statistics. Proceedings of the workshop “Adaptive systems and user modeling on the World Wide Web. Six International Conference on User Modeling, Chia Laguna, Sardinia, June 2-5.

Stahl, G., Koschmann, T., & Suthers, D. (2002). Computer-supported collaborative learning. In R. K. Sawyer, (ed.) The Cambridge handbook of the learning sciences (pp. 409-425).

Standage, T. (2002). The turk: The life and times of the famous 18th century chess-playing machine. New York: Walker and Company.

Stanford Aptitude Seminar (2002). Remaking the concept of aptitude: Extending the legacy of Richard E. Snow. Mahway NJ: Lawrence Erlbaum Associates.

Strayhorn, T. L. (2006). College in the information age: The use of technology among undergraduates and general learning outcomes. Paper presented at the annual meeting of the American Educational Research Association (AERA), San Francisco, CA.

Suppes, P. (1967) Some theoretical models for mathematics learning, Journal of Research and Development in Education 1, 5-22.

Tanner, D. & Tanner, L. (2006). Curriculum development: Theory into practice. New York: Prentice Hall.

Triantafillou, E., Georgiadou, E., & Economides, A.A. (2008). The design and evaluation of a computerized adaptive test on mobile devices. Computers and Education, 50, 1319-1330.

Uresti, J. & du Boulay, B. (2004). Expertise, motivation and teaching in learning companion systems. International Journal of Artificial Intelligence in Education, 14(2), 193-231  

49

Van Horn, R. (2003). Computer adaptive tests and computer-based tests. Phi Delta Kappan, 84(8), 567, 630-631.

Watts, D. (2004). The “new” science of networks. Annual Review of Sociology, 30, 243-270.

Wenger, E. (1999). Communities of practice: Learning, meaning, and identity. London: Cambridge University Press.

Woods, P. & Hartley, J. R. (1971). Some learning models for arithmetic tasks and their use in computer-based learning. British Journal of Educational Psychology, 41, 35-48.

Tobias, S. (1989). Another look at research on the adaptation of instruction to student characteristics. Educational Psychologist, 24, 213-227.

Tomlinson C.A. (1999). The differentiated classroom: Responding to the needs of all learners. Alexandria, VA: Association For Supervision and Curriculum Development.

Tomlinson, C. (2001). The how to’s of planning lessons differentiated by interest. Pp. 52-59 in C. Tomlinson. How to Differentiate Instruction in Mixed-Ability Classrooms. Alexandria, VA: Association For Supervision and Curriculum Development.

Tyack, D. (1974). The One Best System: A History of American Urban Education. Cambridge, MA: Harvard University Press.

Tyler, R. (1969). Basic principles of curriculum and instruction. Chicago: University of Chicago Press.

Uhr, L. (1969). Teaching machine programs that generate problems as a function of interaction with students. Proceedings of the 24th National Conference, pp. 125-134.

Vassileva, J. (1997). Dynamic courseware generation on the WWW. In Proceedings of the workshop “Adaptive Systems and User Modeling on the World Wide Web.” Sixth International Conference on User Modeling, Chia Laguna, Sardinia, June 2-5

Wainer, H. (Ed.). (2000). Computerized adaptive testing: A primer (2nd Edition). Mahwah, NJ: Erlbaum. Wang, M. C, & Gennari, P. (1983). Analysis of the design implementation and effects of a data based staff development program. Teacher Education and Special Education, 6, (4), 211-226. 

50

Wang, M. C., Rubenstein, J. L., & Reynolds, M. C. (1985). Clearing the road to success for students with special needs. Educational Leadership, 43, 62-67.

Weber, G. & Brusilovksy, P. (2001). ELM-ART: An adaptive versatile system for Web-based instruction. International Journal of Artificial Intelligence in Education, 12(4), 353-384.

Weiner, N. (1948). Cybernetics. New York: Wiley.

Welch, R.E. & Frick, T.W. (1993). Computerized adaptive testing in instructional settings. Educational Technology Research and Development, 41(3), 47-62.

Wenger, E. (1999). Communities of practice: Learning, meaning, and identity. London: Cambridge University Press.

White, B., Shimoda, T., and Frederiksen, J. (1999). Enabling students to construct theories of collaborative inquiry and reflective learning: computer support for metacognitive development. International Journal of Artificial Intelligence in Education, 10, 151-182.

Whitty, G. and Young, M. (eds.) (1976). Explorations in the politics of school knowledge. Driffield, England: Studies in Education, Ltd.

Winne, P. H., & Nesbit, J. C. (2009). Supporting self-regulated learning with cognitive tools. In D. J. Hacker, J. Dunlosky & A. C. Graesser (Eds.), Handbook of metacognition in education (pp. 259-277). New York: Routledge.

Young, M. F. D. (1971). An approach to the study of curricula as socially organized knowledge. In M.F.D. Young (ed.), Knowledge and Control. (pp. 19-46). London: Collier-MacMillan Publishers.

Young, M., Depalma, A., & Garrett, S. (2002). Situations, interaction, process and affordances: An ecological psychology perspective. Instructional Science, 30, 47-63.

Zimmerman, B. and Schunck, D. (2001). Self-regulated learning and academic achievement (2nd ed.). Mahway, NJ:L Erlbaum.

51