blogger counters

Sunday, January 13, 2013

Adaptive Learning: Feedback and Mastery - Where Are We Today?


If you’ve ever had private instruction, a tutor, or coach, you’ve been involved in adaptive learning. When you hit the ball late, or are using the wrong grip, your coach notices the problem, lets you know what you’re doing wrong, shows you how to correct it, and then, works with you until you have it right.

Similarly, if you’re working with a tutor to prepare for a standardized test, such as the LSAT, your tutor will acknowledge where you have mastered the content, and help you pinpoint where you have knowledge gaps. She will then work with you in those areas until you have mastered all the areas.

The fact that you receive instant feedback is very motivating, and until now, it has been hard to find good computer-based educational programs / e-learning that provides specific, personalized instant feedback on personalized content.

One of the reasons why tutors and coaches can be effective is that they provide instant, relevant, and detailed feedback.

Adaptive Learning – Vator News

Adaptive: Feedback and Mastery

However, it has not always easy to find a way to provide the same high-quality feedback in an online environment, and to see the same performance improvements. To try to replicate the interactive nature of the student-mentor relationship, computer programmers and designers have developed what is called “adaptive learning.”

Adaptive learning hinges on the idea that the course content should adapt to each user in order to achieve a desired outcome (often mastery).  It is a way to transform a student’s learning into a unique, individualized experience for each learner and by accommodating the diverse needs of individual students, it combats the tendency for “one size fits all” solutions to be boring, unengaging, and ineffectual. 

While you might read articles that focus on complex artificial intelligence systems and the
mathematics involved in creating “smart” educational software, it’s helpful to step back a moment and recognize that adaptive learning is based on the old, familiar one-on-one, mastery-centered learning.

The problem with individual coaching, private lessons, and tutoring is that they are expensive—when dependent on humans to provide the individualized feedback.  They are also not available where and when you need them. So, this is extremely important when it applies to mobile learning.
In many cases, the problem with computer-based tutoring is that there may be too much focus on using technology to automate, mechanize and reduce costs by becoming very efficient at delivering content, but there is little paid to the students’ actual outcomes and or performance.  It is also very important to have a way to align the learning goals with individual capabilities.

Adaptive learning was envisioned to meet and overcome this deficit. Often considered “smart” programs, the online or computer-based education programs adapt or modify the sequence and content of the lessons and courses based on the learner’s responses to assessment (formative and summative) questions that can take place during the lesson. It is useful because the experience replicates having a live tutor or mentor who can diagnose and guide students.

Many argue that adaptive learning is even better and more effective than “live” tutoring because the depth of assessment can be much more extensive. In addition, e-learning plus m-learning make access ubiquitous (assuming connectivity).

In addition, because of the personalized nature of the learning experience, and the fact that succeeding in mastery items that are not too hard, but yet not too easy, can be very intrinsically rewarding and motivating.

One of the primary allures of adaptive elearning is that it promises to be highly engaging, with extremely motivated learners who complete their work in a positive manner.

Defining “Adaptivity”

For clarity, it is useful to take a close look at the term, “adaptive,” and to specify its meaning in an elearning environment.

“Adaptivity” is the term used to indicate the “adjustment of one or more characteristics of the learning environment” (Wauters, 2010), and the adaptive actions can take place in one of three areas:
o Appearance or form (how it is presented to the learner; adding text, graphics, video, etc.)
o Adaptive content (how the instructional materials are presented, and cues / help given)
o Adaptive curriculum sequencing  (changes according to level of difficulty and the learner’s knowledge or skill level).

History of Adaptive Learning

Adaptive learning, which would result in a “smart” learning environment has been a dream of most educational developers, beginning in the early years of computer-based-learning which often did have some flexibility with the learning pathway, which is to say that you could not progress unless you demonstrated mastery, and if you did, you could ascend to the next level.

One can make the argument that early skill and mastery games that involved tests could be considered adaptive in the sense that the player automatically advances once he or she demonstrates mastery, it is not necessary to go through every stage, and the ones corresponding to the lower levels can be skipped.

Early interactive CD-ROM language courses (approximately 1998) that compared one’s pronunciation with the audio profile of a native speaker’s pronunciation and would not allow one to proceed unto a 80% “match” was achieved is another example of early adaptive learning environments.

Currently, adaptive learning systems are usually we-based application programs that provide a personalized learning environment for each learner, “by adapting both the presentation and the navigation through the learning content” (Tseng, etal, 2008), 171).  Recently, research has reinforced the notion that the most natural and effective types of learning are highly situated, and that they are, in their very nature, adaptive (Mesoudi, 2010, 337).

Next-generation adaptive learning frameworks are likely to incorporate artificial intelligence (AI) models, with a goal of developing very fine-grained sub models, to be able to fine-tune decision points and pathways (Pedrazzoli, 2010, 225). One adaptive elearning framework, OPUS One, is being developed to provide an individualized adaptive elearning  that can cover a wide array of subject matters at a very  high level.

Two Types of Adaptive Learning

Two types of adaptive learning environments have emerged: adaptive hypermedia (AH) and intelligent tutoring systems (ITS).  Simply stated, adaptive hypermedia could be thought of as “smart searches” while intelligent tutoring systems can be thought of as “smart test prep.”

Adaptive hypermedia often presents a great deal of information in the form of hypertext, while ITS tends to use more selective materials (Brusilovsky, 2001). Examples include online encyclopedias, e-commerce, and online libraries, which build on the individual’s choices and then provide access to likely good matches. The “smart guides” and “smart shoppers” have become familiar to many learners, especially given their experience with user interfaces such as Amazon.

Intelligent Tutoring Systems (ITS) are more widely used in e-learning because they support learners in the process of problem solving. ITS systems often include incorporating computerized adaptive tests (CAT) in order to align the level of difficulty and the content knowledge areas with the student’s learning goals  (Wauters, 2010, 550). 

In addition, the use of Item Response Theory (IRT) is often used to develop sequencing options. For example, IRT looks not only at the answers to exam questions, but also time on task, skipped items, and the sequence in which exam items were answered.  IRT also presents questions in the same sequence that the learning objectives were presented, and in ascending order of cognitive complexity.  Knewton, which incorporates IRT, puts it this way: “IRT models student ability using question-level performance instead of aggregate-test-level performance” (Knewton, 2012, p. 4).  So, the program contains an algorithm that models probability of the student answering a problem based on that student’s performance on other questions possessing the same level of difficulty, level, and topic).

Uses of Adaptive Learning

Adaptive learning is very widely used in areas where competency is easily measured, and there are numerous small mastery points along the way to make high-frequency assessment fairly easy and justifiable for learning purposes.

From a design or programming standpoint, a robust adaptive learning environment can be built given the following conditions:

o Modular, flexible course framework
o Large array of learning objects that can be deployed in flexible sequences
o Teaching materials correspond to the assessment learning objects; in fact the assessment learning objects may be presented in certain situations as a part of the content / course materials
o The presentation of concepts, contexts, and assessment determined by learning objectives
o Learning objects graded to correspond to difficulty and learning level
o Materials can be organized for use with adaptive tutoring as well as instruction

Examples of Adaptive E-Learning

Kenny and Pahl (2009) describe their experiences in developing adaptive and intelligent tutoring for a computer programming course. In their view, computer programming is an excellent topic for adaptive learning because competency, skill, and knowledge of content are easily measured.

The adaptive element for the Kenny and Pahl model involves an interface that presents feedback loops that adaptively generate corrections, suggestions, recommendations, and recommended sequences (Kenny & Pahl, 2009, 183).

They found the following benefits to the approach (Kenny & Pahl, 2009, 189):
 Always available
 Self-paced learning
 Easy to use
 Enjoyable / rewarding
 Practical application rather than passive lecture

Adaptive learning in both CD-ROM and cloud-based elearning can be found in the following areas:
· Language learning (beginning, intermediate, advanced  - conversation, reading, writing)
· Trades and careers that require use of equipment and scaffolded learning (welding, nursing, HVAC,
electrical trades, etc.)
· Topics that require identification of different taxonomies along with functions (anatomy, yoga, massage therapy, veterinarian technologist)
· Skills that require building-block approaches to concepts (mathematics, geographical information systems, chemistry)
· Performance on standardized testing.

Adaptive Learning and Web 2.0 and Web 3.0:

Can adaptive learning be used in the context of interactive, socially-networked learning environments?

London and Hall (2011) addressed that question in their quest to employ adaptive learning strategies in real-life settings. They looked specifically at sales training, and included the use of virtual worlds as an adaptive learning environment framework.

They found that adaptive learning that utilizes social network or other networked and integrative / interactive approaches encouraged reflective observation, and led to abstract conceptualization. It also encouraged active exploration (London & Hall, 2011, 760).

Examples of Adaptive Learning Software

Perhaps the most aggressive player in the adaptive learning for college instruction is Knewton ( Their approach is based on a proprietary set of algorithms that perform calculations based on the students’ responses to align their instructional material with the needs of the learners. The approach is mastery-focused, which means that once a topic is mastered, questions are not repeated. However, as long as the learner is missing questions, additional content will be presented and questions asked until they are answered correctly.

Knewton is “continuously adaptive” which is to say that the system “responds in real-time to each individual’s performance and activity on the system” (Knewton, 2012, p. 3).
Knewton started with developmental mathematics content, and was evaluated in several universities, including Arizona State, University of Nevada – Las Vegas, Pennsylvania State, Mt. Saint Mary University, and Washington State University. The program was so successful that it was acquired by Pearson, which is incorporating it into its “MyLab” series.

MyMathLab for developmental math now contains adaptive learning modules.  MyWritingLab for developmental English is also currently being enhanced with adaptive learning.

In 2012-13, Pearson plans to roll out adaptive solutions for the MyLab products in the following subject matter areas: statistics, economics, first-year composition, and science.

For K-12 learners, Pearson has launched its SuccessMaker product at  It is designed for math, but is also being used in other subject matter areas.

Another popular adaptive learning program is Grockit ( which focuses on test preparation for entrance exams for college and graduate school. There are test preparation programs for the GMAT, SAT, LSAT, GRE, and ACT. This is a direct competitor with Sylvan Learning centers. 

Other Adaptive Learning Programs:
Testive --  Developed at MIT, and used for test preparation (SAT, in particular). 

Neurosky – Math Trainer - Uses the Neurosky Brainwave headset for adaptive learning that employs brainwaves.

NeuroCog – Focus Pocus - This uses the Neurosky brainwave measuring headset in conjunction with software to help develop one’s ability to self-regulate (relax with test anxiety), to focus, and to remember.

iKnow! --  -- adaptive, cloud-based adaptive e-learning: Japanese, Chinese, English. First five lessons free.  (I’m trying Japanese!)

Pimsleur -- - With the advent of the Nook-based learning platform, it appears that the Pimsleur method (which has always had elements of adaptive learning) may be adaptive elearning-enabled.


Brusilovsky, P. (2001). Adaptive hypermedia. User Modeling and User-Adapted Interaction. 11: 87-110.

Kenny, C., & Pahl, C. (2009). Intelligent and adaptive tutoring for active learning and training environments. Interactive Learning Environments, 17(2), 181-195. doi:10.1080/10494820802090277

Knewton.  (2012) Knewton Adaptive Learning. (whitepaper) (download)
London, M., & Hall, M. (2011). Unlocking the value of Web 2.0 technologies for training and development: The shift from instructor-controlled, adaptive learning to learner-driven, generative learning. Human Resource Management, 50(6), 757-775. doi:10.1002/hrm.20455

Merino, F. (2012). Resetting education: Adaptive learning to the rescue.

Mesoudi, A. (2011). An experimental comparison of human social learning strategies: Payoff-biased social learning is adaptive but underused. Evolution And Human Behavior, 32(5), 334-342. doi:10.1016/j.evolhumbehav.2010.12.001

Pedrazzoli, A. (2010). OPUS One: An Intelligent Adaptive Learning Environment Using Artificial Intelligence Support. AIP Conference Proceedings, 1247(1), 215-227. doi:10.1063/1.3460231

Shian-Shyong, T., Jun-Ming, S., Gwo-Jen, H., Gwo-Haur, H., Chin-Chung, T., & Chang-Jiun, T. (2008). An Object-Oriented Course Framework for Developing Adaptive Learning Systems. Journal Of Educational Technology & Society, 11(2), 171-191.

Wauters, K. K., Desmet, P. P., & Van den Noortgate, W. W. (2010). Adaptive item-based learning environments based on the item response theory: possibilities and challenges. Journal Of Computer Assisted Learning, 26(6), 549-562. doi:10.1111/j.1365-2729.2010.00368.x

Blog Archive