Generative AI and explicit instruction: a guide to optimising students’ learning

There has been rapid uptake of genAI in schools, universities and colleges. However, genAI use in education can be ad hoc and often ineffectively implemented.

Poor use of genAI can impose unhelpful cognitive burden on students and impede their learning.

Load reduction instruction (LRI) is an approach to explicit teaching that eases cognitive load on students as they learn. In recent research, we explained how LRI provides important guidance to teachers and students as they use genAI.

By applying the principles of LRI to genAI implementation, educators can individualise and optimise learning among diverse students.

What is cognitive burden?

Before we look closely at how to apply LRI principles to genAI, we present a quick refresher on how students learn and the importance of effectively managing cognitive load as they do so.

Cognitive psychology has identified two forms of cognitive load that get in the way of students’ learning: intrinsic load and extraneous load.

Intrinsic load refers to the difficulty or complexity of learning content and instructional material, given a learner’s current knowledge.

Thus, a lesson that is very complex for a ‘novice’ (and therefore high in intrinsic load) may be very low in intrinsic load for an ‘expert’, as their more advanced knowledge helps them easily make sense of the lesson materials.

Extraneous load refers to cognitive burden that is a result of complicated and confusing instructional practices.

It is important for educators to reduce these two forms of load so that what they teach can effectively pass through students’ working memory and be stored in their long-term memory.

Working memory receives and processes information in real-time, including new information. It is very limited in duration and capacity.

Long-term memory is a space for storing a vast amount of information. Learning occurs when information is moved from working memory and stored in long-term memory for later retrieval and problem-solving.

Cognitive burden imposed by genAI

In several ways, there is significant potential for genAI to impose cognitive burden on students. Ineffective use of genAI can present fragmented information and introduce task-irrelevant distractions.

Indeed, genAI provides so much access to so much information that it risks the information being presented too rapidly and overwhelm the learner’s cognitive resources.

Another burden can be the constant decision-making needed by students to determine if and how to use genAI in a given learning task, and whether to trust its responses.

How can we harness LRI to reduce cognitive burden?

Harnessing the principles of LRI provides a way to effectively manage the cognitive burden on students as they engage with genAI to learn.

LRI is an approach to explicit instruction that comprises five key principles:

1. Difficulty reduction as appropriate to prior learning;

2. Support and scaffolding;

3. Structured practice;

4. Feedback-feedforward, and;

5. Independent practice and problem-solving.

Applying LRI to genAI

These five principles can be harnessed as a guide for implementing genAI in learning. Here we provide some brief ideas, with full details provided in our recent research article.  

Principle #1 (Difficulty Reduction) is to reduce task or content difficulty as appropriate to students’ prior learning and existing knowledge. GenAI can adapt tasks and content to learners’ existing knowledge.

There are genAI learning tools that link to prior learning and automatically adjust difficulty based on the learner’s progress.

For example, there are language learning apps with grammar exercises, speech recognition, and vocabulary recall tasks that draw on the learners’ prior knowledge and adapt difficulty to learners’ needs.

Principle #2 (Support and Scaffolding) involves sequenced and structured real-time support for learners to help them through a task. Intelligent teaching and tutoring systems are good for this.

They develop learning paths personalized to learners, with supportive resources to guide them through these paths. There are also scaffolded tutoring routes that can be generated and adapted to individual differences among learners.

Principle #3 (Structured Practice) is relevant to the many ways that genAI provides tools to support students as they practice.

For example, genAI can be used for flashcard creation adapted to individual learners as they memorise content, provide spaced repetition and then further adapt practice tasks based on the learner’s performance.

GenAI tools are also effective for rehearsing and testing learners’ recognition (e.g., generating multiple choice questions) and learners’ recall (e.g., generating quizzes) with incorrect answers revisited for revision in a proximal timeframe.

Principle #4 (Feedback-Feedforward) involves corrective information to the learner (feedback) and improvement-oriented guidance (feedforward).

A well-trained genAI agent can offer targeted and personalised feedback and feedforward at scale (e.g., to every student in a classroom), which is difficult for a teacher in a typical classroom of diverse learners.

GenAI applications also offer diverse feedback-feedforward pathways towards independent self-assessment and learning.

Principle #5 (Independent Practice and Problem-Solving) is vital for students’ engagement in independent practice and problem-solving once they have mastered the necessary knowledge and skill.

GenAI is helpful for providing structured prompts to guide students through complex learning tasks and teach them how to independently formulate effective queries and prompts - for example, by way of the LRI prompting template we present below.

GenAI also provides opportunities for personalised suggestions, immediate feedback, timely guidance, suggestions for goal adjustment, and structured reflection during problem-solving.

LRI-informed GenAI Prompting

LRI can be harnessed by students (and educators) to develop genAI prompts as they learn. Through structured prompting, students can leverage genAI to function in an LRI-informed way.

In the figure below, we provide an example of our LRI-informed prompting that students can use with genAI to guide their learning:

In summary

LRI is a framework for explicit instruction that offers guidance for how to use genAI so it does not overly burden the learner.

The future of genAI-related learning will benefit from attention to LRI principles in order to better cater for individual differences among diverse students and optimse their learning.