Adaptive Hypermedia Content

Adaptive educational hypermedia (AEH) has been described as a systemic process to tailor individual content to the user through websites that automate the learning process. A current goal of AEH is to provide alternatives in the field of digital learning to identify learner modes and present content material in a variety of ways to encourage the learner's progress. In the academic scholarship, two broad categories appear in the e-learning side of the software: adaptive presentation and adaptive interface. Papadimitriou [1] described adaptive presentation as the use of computer gathered information to build a learner profile of educational characteristics. An adaptive interface should allow learners to personalize complex interfaces to allow for easier access to use software for specific tasks [2].

Both techniques are organized around the automated process of a program presenting material to the learner, based on the scripted educational analysis. As this subject may be complicated, incorporating elements of machine learning and instructional website use, this overview is focused on the adaptive presentation side of the AEH process. It is likely worth noting that as AEH is based on web architecture, security concerns that affect the Internet are likely present. General searches did not address security directly, possibly as the architecture is still under development. There are currently three major web technologies referred to in AEH literature; for this paper, Paquette's [3] definition of web 2.0/ 3.0 was used. Web 2.0 was described as social networking and web 3.0 as a semantic web, which utilizes machine learning to connect data for the learner.

This review focuses on the current use of AEH and presents samples of how adaptive hypermedia might work in learning environments. Current research trends for future development are presented in the conclusion.

Historical Background

Weber and Brusilovsky [4] presented that the adaptive hypermedia paradigm began in the 1990s with the introduction of the Internet, though Paquette noted the field's origins were rooted in higher order machine learning first developed during the 1960s with the evolution of computers. Somyürek's [5] literature review on AEH trends noted that the field had evolved notably with the use of a learning-centered model. The author noted that one difficulty with the field that researchers likely needed to address was a lack of standardization among e-learning device platforms. Hwang [6] noted that a facet of mobile platform e-learning was to facilitate quick learning processes. Mobile learning has had potential with the use of intelligent tutoring systems to focus the learner-centered model with personalized development. Vermunt and Donche [7] would likely describe these models as individualized learner patterns.

Somyürek noted many of the early designs for AEH evolved in the 1990's; this period showed the growth of learning styles in education, which has evolved into more complex learning patterns [7]. It is not unusual to find AEH learner methodology based on learning styles, Al-Azawei and Badii [8] addressed the issue of learning styles and lack of precise terminology in an analysis of 76 published studies that used learning styles in AEH from 2000 to 2013.

Adaptive educational hypermedia was designed to address individual learning styles with a set of scripts that allowed the use of interactivity and personal learning preferences to be automated. One example of interactivity is recommender systems similar to advertising used by companies like Amazon or LinkedIn [9]. Another example of AEH interactivity is personalization: Interbook was an early version of this technology for interactive textbooks that allowed learners to customize the content to their learning patterns. Rather than a static reading of the online text, the program allowed for terminology linked to definitions, tutorials on concepts, and practical examples which could be used by students according to their individualized needs and preferences [10]. Both recommender systems and personalization illustrate two different approaches to AEH technologies that allow for engagement with the learner.

Current Research

One of the difficulties with the design of adaptive websites has been both the cost and effort needed to manage links and content rollout. Chae et al. [11] proposed a potential solution to the effort and cost of development with the use of semantic web via cell phone technologies. Their proposed structure uses some adaptive technology and several experts that interface with the system to provide additional support as might be needed among groups of elementary students; the experts interfaced with the learning content manager portion of the semantic database.

By using existing open content links for data, much of the cost and effort at building the database has been already designed. In the proposed case, the research was for elementary students' classes to use a moderated database and access the content through Android phones. Steichen, Ashman, and Wade [12] proposed a different solution to cost containment with research and design efforts through a redesign of how traditional web searches run; their example would likely help leadership initiatives with companies looking to embrace this technology. Instead of searches on presentation, as is typical in most areas of the web, content would be provided based on both presentation of the material and a ranking of content in the presentation. In this model, the authors advanced that designers would not need to reorganize datasets but instead change the descriptors of data sets that search engines use in hypermedia. The change would allow for personalized information retrievals of content based on past user experience. While not necessarily scalable to the entire web, the authors noted, the modified lists would be potentially successful in closed semantic web systems. The authors hoped that an adaptation of the design could be used through multiple web scaled uses to allow for eventual generalization.

Sample Cases

ELM-ART is a semantic web course designed to teach the programming language of LISP. Knowledge of programming is not necessary for the use of the site [4, 13]. This course is a free example of how adaptive hypermedia could work; for instance, the site tracks the user with a user name and password to present progress through the content. The learner has control over the advancement and review of content while the adaptive portion appears to remember the options selected to hone the lesson according to personalized learning patterns. A self-study course, ELM-ART may be an efficient design to engage learners and provide control-particularly in a formative assessment. This type of system could potentially shorten a student's study times. According to Brusilovsky [14], a reasonable second step might be for the program to determine learning patterns and further adjust the program based on that information.

WELSA used learning styles as an outcome to provide students and instructors with adaptation rules based on both content objectives and learner preferences. As Popescu, Costin, and Trigano [15] described, early adaptive education hypermedia usually used a singular learning style to deliver content. WELSA used learner behaviors to dynamically present material to students based on scripted metadata tags, a function of the semantic web as described by Paquette.

An example of a slightly different approach to AEH is ActiveMath. Ullrich, Libbrecht, Winterstein, and Muehlenbrock [16] presented the program, which used learner objects from user goals to provide dynamic interactivity based upon user actions. This approach did not appear to categorize learners based on learning theories, such as in WELSA or ELM-ART, instead relying upon content derived from goals and preferences the learner set and capabilities demonstrated in use of the program. Based partly on HTML, the program presented material that was similar with web pages that learners were likely familiar with.

Summary

Currently, adaptive hypermedia has moved beyond personalization and customization to the use of formative assessments that change on inputs from the learner and to guided learning programs that may incorporate learning patterns and preferences with data modeling structures. These automated program scripts are often recommender systems that provide learner selections. De Bra, Freyne, and Berkovsky [17] suggested four special issue paper topics to be published would be of interest: personalized navigation, customized content handlers, navigation of web pages by scripts, and automated production of material via web and search engines. Perhaps educational individualized learning plans may develop through this technology where students may customize content material and select individualized learning content presentation patterns through dynamic interaction that engages the learner and allows for efficient cognitive processes.

References

[1] Papadimitriou, A. Architecture Trends of Adaptive Educational Hypermedia Systems: The Case of the MATHEMA. American Journal of Artificial Intelligence 1, 1 (2017), 15-28.

About the Author

Derek Luch taught computers and multimedia courses for 12 years, developing a keen interest in pairing digital learning with instruction for adult students.