Adaptive learning

Adaptive learning is an approach within educational technology that features the use of computer software, algorithms, and analytics to identify changes to a computer-based educational product based on an individual learner's performance. Thus, the system will adapt the activities and content provided to each individual learner's participation and mastery of material, giving learners a personalized experience.

Definition

An adaptive learning system is one that will change the digital content and activities that are provided to individual learners based on the levels of participation and achievement that they demonstrate. Adaptive learning software will identify evidence of students' learning and interactions within the system and will provide customized, adapted content and activities based on this performance.

Additional Information

Adaptive computing systems in education are a popular area of research and development in recent years. With the improvements made in artificial intelligence (AI) and machine learning (ML) technologies, computers can be used to identify patterns of learner behavior and mastery of skills and content. With this in mind, a computer can make decisions to give learners more or less challenging activities to perform based on the learner's skill level, provide additional content to them as needed, or offer scaffolds and supports.

The power in adaptive technologies is that the computer identifies evidence in formative evaluation data on a learner's competence and changes in their skill (i.e., learning). The system can also detect patterns of behavior within a digital system, such as how a learner participates and interacts within a digital learning environment. Using this evidence, the adaptive software can ideally be programmed to display custom content based on each learner's individual skill levels, needs, and interests.

The use of adaptive learning in practice still remains highly limited, however. AI and ML approaches are still not very common in educational settings, which often require a different set of skills among designers and educators than is present in typical classroom contexts (i.e., these are most commonly seen in computer science and corporate settings). AI and ML systems also require large datasets of actual interactions of learners so that the computer may "learn" the patterns that they are looking for by working through the data case by case. Additionally, AI systems are often "trained" by experts to how classifications of behavior should occur and what categories should even exist in the system in the first place. These trainers of computer systems also program the AI software to determine appropriate outputs based on how the system interprets the data, which in this case includes how to change the system to provide new content, deliver supportive help or scaffolding, or to change the activity to better meet the needs of the learner.

Adaptive systems typically do the following tasks:

Identify patterns in how a learner is participating in a system
Identify evidence of a participant's learning from a system (often through formative assessments or through examining the participant's work products)
Make decisions on how to change a system based on the learner's participation in the system OR their level of demonstrated learning/competence.

AI computer systems in real-world contexts have seen limited success that can both (1) accurately identify and predict students' skill and participation levels within a system and (2) subsequently make changes to the interface that align with students' needs and interests. Further research and development are needed in these areas.

There are also equity, ethics, and accessibility issues associated with adaptive learning technologies. Philosophically, it can be easy to rely on a computer algorithm to assign whether a learner succeeds or not, and an algorithm may not pick up on all of a learner's individual needs, background experiences, and affect. This, in turn, can give learners an unfair learning experience if the system misidentifies or incorrectly categorizes a student based on the expected levels of behavior. Designers of adaptive systems for learning should thus be cautious about how computer systems categorize and subsequently present information and activities to students to ensure that each and every student is equitably treated within such systems.

Tips and Tricks

Adaptive learning systems can be difficult to create and implement. If you are considering such a system, a large investment of time and financial resources is likely required. You may also need to assemble a team of experts with a broad range of skills to implement such a system.
It is important to consider what kinds of behaviors and evidence of learning that you want the AI/adaptive system to look for. How often should participants be assessed? This will help you program adaptive systems to spot the kinds of behaviors and mastery that will help the system make subsequent changes for the learner to personalize the system.
It is also important to consider the types of "outputs" that you want the system to make. What kinds of changes to the system should occur, based on the system's evaluation of a learner's performance? How should the system change content, make activities more difficult or easier for learners, or how should the system present supports and scaffolds to help students when they need help? How should the system identify when a learner needs help in the first place - what does that look like?
Sometimes adaptive systems and AI/ML systems in general can help you spot patterns of behavior and categorize new behaviors before you even see them as a researcher or designer. Lean on the statistical analyses that your system might provide to identify clusters of new behavior or patterns of how people are demonstrating mastery to help you consider how and why people use your product to learn.

Adaptive learning

From The Learning Engineer's Knowledgebase

Contents

Definition

Additional Information

Tips and Tricks

Related Concepts

Examples

External Resources