Are students harnessing AI for genuine learning or just easy answers?

The challenge for educators is to shift students' AI use from simply completing academic tasks to supporting the processes through which students actually learn, says educational psychologist Professor Andrew Martin.

Generative AI is rapidly becoming part of how students study, prepare assignments, revise for tests, and work through academic tasks. But much of this use is ad hoc, ill-informed, unstructured, unguided, and poorly managed.

Some students try a prompt, see what comes back, change a few words, try again, and hope they eventually get something useful. Some students use AI to complete academic tasks as quickly as they possibly can. Other students delegate major parts of their thinking and learning to AI.

Adding to the problem, educators lack coherent organising frameworks to guide how AI can be used to support students’ learning.

Psychologically-informed use of AI

For decades, educational psychology has been examining how students learn, what helps them through this process, what gets in the way, and what educators can do about it. This research has shed a lot of light on instruction, motivation, engagement, learning, and achievement.

The insights educational psychology has unearthed over this time can be a basis for psychologically-informed AI use by students.

By psychologically-informed AI use, I mean using AI in ways that are explicitly guided by established educational psychology theory and research on how students learn.

Rather than asking, “How can AI do this academic task for the student?”, psychologically-informed AI use would ask, “How can AI help the student become better able to do this academic task?”

In the case of the former question, a student preparing for an exam would ask AI to generate all the answers they need to know. In the case of the psychologically-informed second question, they would use AI to identify what they already understand, break unfamiliar material into manageable parts, practice important knowledge and skills, receive feedback on their attempts, and use this as a basis to attempt sample exam questions independently.

Both involve AI, but represent very different ways of using it for learning. One is unguided and expedient – while the other is grounded in what we know is evidence-based best practice in learning.

Load reduction instruction – one approach to psychologically-informed AI use

As a case in point, I will unpack an example of how AI can be used to support students’ learning by harnessing the principles of load reduction instruction (LRI).

LRI is an approach to explicit instruction aimed at reducing unnecessary cognitive load while students are learning. It identifies five principles that move students towards increasingly independent learning.

The five principles are:

1.     Difficulty reduction during initial learning, matched to their existing knowledge and prior learning

2.     Scaffolding

3.     Practice

4.     Feedback and feedforward

5.     Independent problem solving

LRI was developed primarily as an instructional framework for educators. But its five principles can also provide students with a useful guide for working with AI to support their own learning.

What this looks like

Here is what that might look like – and further below I introduce the GenAI Motivation and Learning Buddy that provides the specific AI prompt scripts for doing this.

Principle 1. Difficulty reduction

When students first encounter new material, the task can contain too much unfamiliar information at once. LRI begins by reducing difficulty to a level that matches the learner’s current knowledge and skill. AI can help here. For example, a student might tell AI what topic they are studying and ask it to check their existing knowledge. Based on their responses, AI could then help break the topic into smaller parts and begin with the material the student is ready to learn.

Principle 2. Scaffolding

Students benefit from scaffolding as they work through an academic task. Rather than asking AI for the finished product, students can ask for prompts, hints, explanations, or step by step guidance. AI would walk them through a problem one step at a time. It might even ask the student to suggest what they should do next before AI gives them a hint. Here, AI gives assistance, while the student still has to think, decide, and act for themselves.

Principle 3. Practice

AI can generate many opportunities for practice. A student might ask AI to give them three practice questions on a topic, starting with the easier one, and instruct AI that they must answer each question before AI shows them any explanation. Or they might ask AI to show one worked example of a solution and then provide a similar problem for the student to complete themselves.

Principle 4. Feedback and feedforward

Feedback provides corrective information about what students have just done – and feedforward involves guidance about what they can do next to improve. After completing a practice task, a student might ask AI to check their response, tell them what they have understood correctly, identify one or two things needed to improve, and suggest what the student can work on next. This is aimed at the student improving their own work, not AI progressively doing the work for them.

Principle 5. Independence

The aims of principles #1 to #4 are to build students’ mastery towards independent learning and problem-solving. Indeed, imposing highly structured support when students no longer need it can be unhelpful to learning. At this point in the learning process, students can inform AI that they understand the basics and would like a new problem to complete independently, without any scaffolding unless the student asks for it, and for a reflection activity once the task is completed. Thus, successful AI-supported learning should not necessarily result in greater dependence on AI. In many cases, it should result in the opposite.

The GenAI motivation and learning buddy

LRI can also be the basis for developing detailed AI prompt scripts that can be used to support learning. The open access GenAI Motivation and Learning Buddy was recently developed to do this.

The GenAI Buddy provides a library of prompt scripts that students can use with widely available generative AI tools such as Claude, ChatGPT, Copilot, and Gemini. It treats AI as a collaborative partner in learning, helping students work through an academic task, but not to do that task for them.

The GenAI Buddy contains five prompt scripts corresponding to the five principles of LRI: difficulty reduction, scaffolding, practice, feedback and feedforward, and independence. Students work through structured, self-paced steps for each of these principles.

The prompts include definitions and explanations of each principle, practical learning strategies, examples and exercises, encouragement when difficulties arise, and reflection for future learning. They also include explicit guardrails intended to stop the AI simply providing answers.

Of course, none of this means AI can replace teachers. It is well known that AI can give inaccurate and inappropriate responses, and its suggestions need to be judged critically. For school students in particular, educator guidance remains vital. The point is not to hand learning over to AI, but to make AI one more tool students and educators can use together to support learning.

We know a lot about how students learn – so let’s use it

If AI use remains ad hoc, unstructured, and focused on producing answers and completing tasks, students risk missing much of its educational potential.

Psychologically-informed use of AI can help here – and we have a tremendous body of theory and research on which to draw.

LRI is one example – but there are many opportunities to develop structured prompt scripts around key tenets of psychological theory and research to teach students how to work with AI to support their learning – not do their learning for them.

Andrew Martin, PhD, is Scientia Professor, Professor of Educational Psychology, and Co-Chair of the Educational Psychology Research Group in the School of Education at UNSW. He specialises in student motivation, engagement, learning, and quantitative research methods.

This article was originally published on EduResearch Matters. Read the original article.