Introduction
In a traditional classroom setting the teacher differentiates the delivery of material, the activities that students participate in and the questions that are put to the students. The teacher is able to do this by referring to his or her student records when planning the session and in real time during the classroom by continuously reviewing how students respond to each element of the lesson. The signals that the teacher picks up can be quite subtle such as the expression on each student's face or a student's tone and response to a question. The teacher's knowledge and understanding of each student allows her to differentiate all aspects of the lesson in order to fulfill the goals and objectives of the lesson. The challenge facing colleagues who are developing adaptive learning environments is how to mimic or replicate this behaviour online by using the tools to hand; such as analytics and machine learning. It must be noted that adaptive learning environments are still in their infancy and whatever is developed will only be a simulacrum - an image which lacks the substance, quality and nuances of the original. That is not to say that adaptive learning environments have no value because if they are developed to support and enhance the role of the teacher; and if they offer a more personalised programme of learning to each student then they deserve further study and support. This article seeks to analyse the broader role of the adaptive learning environment and its potential impact on teaching, learning and assessment.

Background
The traditional medium for managing and distributing teaching, learning and assessment materials to students online has been the virtual learning environment. Whilst they have enabled distributive learning to take place they have been unable to offer a truly differentiated and personalised learning experience. So when teachers post learning and assessment materials to their virtual learning environments they do so in the knowledge that their students will be presented with the same teaching, learning and assessment materials regardless of each student's ability, learning preferences, academic history, needs and requirements. The advent of the adaptive learning environment provides teachers with the tools to remedy this shortfall in distributive learning.

The current landscape
The most detailed study of adaptive learning environments and their suppliers has been undertaken by Tyton Partners entitled Learning to Adapt - Understanding the Adaptive Learning Supplier Landscape. The paper provides a summary of each supplier's adaptive learning environment and offers invaluable insight for schools, colleges and universities before they approach the market place. Additional papers on adaptive learning environments are listed at the end of this article.

What constitutes an adaptive learning environment?
The growing maturity of web based services, content authoring tools, analytics and machine learning has fundamentally altered the nature of virtual learning environments. The advent of the adaptive learning environment represents the next stage of development in distributive learning; where data and evidence based teaching, learning and assessment becomes the norm.

The following section details the various elements that make up an adaptive learning environment.

1. The adaptive learning environment uses the student dataset to present differentiated content to the student. The use of the student dataset is enabling educational institutions to deliver content that is specific, differentiated and contextualised to meet the needs of each student. The breadth and depth of the student dataset means that teachers can use multiple student characteristics to design and deliver content and assessment materials that would have been impossible to construct in a normal classroom setting. Here is a list of variables and methods that are currently used by teachers at Bolton College when designing differentiated and adaptive content and assessment activities:

• student name, unique student ID;
• gender, learner support needs, age;
• course name, department or faculty name, campus location, time, date;
• entry qualifications, diagnostic test results, on course assessment grades, expected grades, course targets and goals;
• learning preferences, the learning styles that have been identified as needing further development;
• patterns of behaviour on the student home page and the wider personal learning environment;
• the student's personal profile which includes long term career goals, a student's risk register,
• current and historical assessment data; and
• current grades versus anticipated grades, predictive analytics; and gap analysis.

The use of these variables and others will be common to all adaptive learning environments regardless of their maturity. As the volume of adaptive content and assessment material grows, adaptive learning environments will increasingly make use of meta data to enhance and refine the delivery of differentiated and contextualised material to each student. As this service matures teachers will no longer create ad hoc tutorials and post them to the learning environment. Instead, they will create micro tutorials, activities and questions and add them to the content database along with the meta data or tags associated with each piece of content. Once these assets have been added to the database the adaptive learning environment can then shape the content and assessment activities to present to each student. Please refer to section 4 below for further information on this process.

2. Adaptive learning environments adjust the content and assessment activities that are presented to the student according to the actions and assessment results of the student within the adaptive learning environment and external to the environment. If a student performs well in an online tutorial and assessment activity within an adaptive learning environment he or she will be subsequently presented with a tutorial or assessment activity which is at a higher level of difficultly or the converse will happen if the student performs poorly in the first instance. The feedback that is presented to the student at each stage of the tutorial or assessment activity will reflect the progress being made by the student. In many instances the learning resources and assessment activities that are presented to each student are based on the targets that have been agreed between the student and teacher. In the following example the teacher and student agree on a learning and assessment target. In this scenario the choice of targets is very specific, they are quantifiable and measurable. The adaptive learning environment references the agreed target and it uses its library of resources to enable it to tailor teaching, learning and assessment materials that will support the student to achieve that target.

3. The adaptive learning environment uses historical data to draw up a student profile which in turns informs the content and assessment materials to present to the student. At the present moment in time the education sector's use of historical data to inform the delivery of real-time teaching, learning and assessment materials to the student is not widespread. Academic teams currently take advantage of student profiling to inform multiple services such as academic and pastoral support, library services, the mode of delivery for courses and with regard to adaptive learning environments they will start to use historical data to inform what teaching, learning and assessment materials to present to each student on a course. For example, a course may have students who match various student profiles. The student profile is composed of multiple variables such as a student's entrance qualifications when joining a course, age, gender, learning preferences, learning needs, start of course diagnostic test results, on course assessment grades and so on. The institution's historical dataset is also used to inform the student profile. As students join and undertake their courses they are matched against a given student profile. The profile associated with each student may change as the student progresses through the course. A student who requires more support and guidance on a particular element of a course will be matched against a particular student profile which in turns informs the teaching, learning and assessment materials to present to that student to successfully take them through that part of their course. A student who is progressing well will be assigned to another profile; and the adaptive learning environment will deliver content and assessment activities that stretch and challenge the student further. As the current cohort progresses through their studies their anonymised data will be used to inform and update the wider student profile for a course. Please note that different courses and departments will manage the student profiles that are applicable to them. The use of student profiling in this manner will help course teams to improve student support; and it will help course teams to raise student retention and course completion rates. Data mining will also reveal valuable information to course teams which will improve their ability to manage the progress of each student through their studies.

4. Adaptive learning environments query the wider dataset before determining the highest ranking tutorials to present to students. Adaptive learning environments can be used to determine the teaching, learning and assessment materials to present to the student with complete autonomy from the teacher. Tutorials at Bolton College are currently ranked according to two variables. The first of these variables is the average assessment score that is recorded by students when taking formal tests on their online courses. Tests are undertaken following the completion of one or more tutorials. Tutorials that raise test scores have a high weighting. The second variable uses gap analysis to determine the difference between anticipated grades and actual grades. If tutorials can demonstrate that students achieve or beat anticipated grades they get a higher weighting. The adaptive learning environments that work well will be the ones that have a large and extensive repository of tutorial slides for every element of a course. The choice of content to be presented to each student may not be based on whole tutorials; because the adaptive learning environment may identify, rank, order and present a series of individual slides that make up a larger tutorial to the student. The adaptive learning environment will be dynamic in nature; after all, students are unique, their learning preferences may change from day-to-day and their needs and requirements will vary according to the point on their course. The dynamic nature of adaptive learning environments may well mean that teachers and student support teams will not be able to ascertain the make up of each tutorial that has been presented to their students. This is compounded further if the algorithms or hypotheses used by the adaptive learning environment adjust and adapt independently without the intervention of the course team.

Adaptive learning environments will start to behave more autonomously because they will make decisions which are made independently from the teacher, data specialists and the content developers. The adaptive learning environment will learn and adapt its behaviour based on its acquired memory and it will increasingly make use of statistical techniques such as Bayesian probability to test its own hypotheses. For instance the adaptive learning environment may assign a prior probability for a given hypothesis. Thereafter it will collect and analyse data that corresponds to that initial hypothesis before assigning a revised and updated posterior probability against the original hypothesis. An example of this method is described in the following hypothesis: Students who need to gain a distinction on their next piece of coursework will need to complete tutorials A, B and C because these tutorials have previously helped students to gain a distinction grade. In this example, the prior probability of students doing well could be high, but subsequent assessment data could show a significant proportion of students falling short of the top grade. If this happens the higher prior probability is updated to a lower posterior probability to reflect the actual lower average student grade. In this example, the original hypothesis is either revised by using backjumping techniques or another hypothesis is tested until all students acquire the top grade on their next piece of coursework.

Suppliers
If you are interested in learning more about adaptive learning environments here is a list of companies that offer products and services associated with the filed.

• Acrobatiq
• Aleks
• Brightspace LeaP
• Carnegie Learning
• Carego
• CogBooks
• Difference Engine
• Drillster
• Fishtree
• Flat World
• Fulcrum Labs
• Knewton
• LearnSmart
• LoudCloud
• Open Learning Initiative
• Realizeit
• Smart Sparrow
• Snapwiz
  

Summary
The advent of the adaptive learning environment is a welcomed addition to the distributive learning landscape because it provides teachers with additional tools to deliver personalised and contextualised teaching, learning and assessment activities to each of their students. The use of machine learning in adaptive learning environments is the most significant development in distributive learning because it marks the time when a new agent is introduced into the classroom. That new agent is the adaptive learning environment which quietly queries and analyses vast quantities of data before it autonomously determines the tutorials and assessment activities to present to a given student. Further progress has yet to be made before adaptive learning environments become commonplace in our schools, colleges and universities; but the progress that is currently being made with analytics, machine learning, content creation, machine marking and natural language bodes well for the future.

Footnotes:

1. Learning to Adapt: A Case for Accelerating Adaptive Learning in Higher Education (2012)
2. Learning to Adapt: Understanding the Adaptive Learning Supplier Landscape (2012)
3. Learning to Adapt: The Evolution of Adaptive Learning in Higher Education (2015)