Dr Leon Furze

Over the past few weeks I have introduced six new strategies for using Generative AI in education. These posts have focused in particular on text-based LLMs like GPT and Claude 3.5 Sonnet, and have explored the following six areas:

• Designing
• Differentiating
• Engaging
• Imagining
• Editing
• Evaluating

If you haven’t already, make sure you check out those two posts:

This post continues off a series where I’ll explore each Strategy in much more depth, looking at the ideas which underpin each area and how the technology has developed in the past 18 months. These posts are intended to deepen your understanding of GenAI in education whether you’re a beginner or you’ve been using these tools for a long time. For the previous post on ‘Designing’, click here.

Differentiation and the Pitfalls of ‘Personalised Learning’

In the original Practical Strategies for ChatGPT in Education I used the term “personalising” to introduce prompts which could be used to adjust and modify teaching materials. Unfortunately, the term “personalised learning” has been thoroughly co-opted by tech companies and “AI experts” who have little understanding of the realities of teaching and learning.

“Personalised learning” has become a catch phrase in edtech, and influential technology industry people like Bill Gates, Sam Altman, and Andrej Karpathy have all weighed in. But their understanding of personalised learning often lacks nuance: it’s “personal” because it involves individual learners working on individual devices with AI tutors and chatbots. To me, that sounds more like “isolated learning”.

So I’m moving to the term differentiation, to acknowledge that adjusting educational resources for students doesn’t always need to be at the individual level: you can make adjustments for groups, differentiate through collaborative learning, and make use of the fact that many students work best in communities rather than in isolation. Of course, some students will learn best in the ways promoted by Gates and co: I’m one of them myself. But there is no one size fits all solution for differentiation.

More Practical AI Strategies: Designing

Differentiation with GenAI

The concept of differentiation in education is not new, but it remains a huge challenge for educators striving to meet the diverse needs of their students. Rooted in the understanding that learners differ in their readiness, interests, and learning profiles (as opposed to the educational myth of learning styles, which often crops up when using AI), differentiation asks teachers to modify content, process, product, and learning environment to ensure all students can access and engage with the curriculum.

In conversations about differentiation with GenAI, a lot of attention has been focused on “low hanging fruit” such as adjusting the length, complexity, or sophistication of texts. This is something which any AI platform can do reasonably well, but it isn’t always what students need. It’s helpful to use clear frameworks and best practice for differentiation, and to look beyond those simplistic modifications.

In these examples, I’m drawing on a range of resources from my own teaching experience but in particular this list from the NSW Department of Education.

The first example uses the concept of ‘tiered instruction’ to establish a classroom environment where differentiation can happen more naturally. Rather than differentiating for individual students – an almost impossible task if you have over twenty students with individual needs – tiered instruction aims to provide a starting point for students wherever they are. It’s a fairly lengthy prompt because I’m not taking for granted that the model “knows” what I mean by tiered instruction.

Using the concept of tiered instruction, create a three-tiered lesson plan for teaching the water cycle to a Year 7 Science class. For each tier, provide activities that address the same fundamental concepts but vary in complexity, scaffolding, and modes of expression. Ensure that higher tiers involve more abstract thinking and independent work, while lower tiers offer more concrete experiences and structured support.

Use the following tiering approach:

Tier 3 (top level): Critical thinking, where students evaluate and analyse. For each tier, provide activities that address the same fundamental concepts but vary in complexity, scaffolding, and modes of expression. Ensure higher tiers involve more abstract thinking and independent work, while lower tiers offer more concrete experiences and structured support.

Tier 1 (base level): Basic learning and skills all students must know.

Tier 2 (middle level): Application of knowledge, where students manipulate information.

In my experience with GenAI and lesson planning, LLMs will often produce problematic output with assessments targeted to ‘kinaesthetic learners’ and ‘visual learners’. Gardner’s theory of multiple intelligence has been a popular but ultimately misguided way of viewing intelligence that likely features very heavily in the internet-scraped dataset used to train GenAI.

To avoid the AI model “defaulting” to multiple intelligences when producing differentiated resources, this prompt uses an internet connected model, e.g., ChatGPT, Copilot, Gemini, to browse a specific resource (in this case the NSW page on differentiation mentioned above) to develop alternate assessments from a different perspective on intelligence.

Based on Sternberg’s intelligence preferences (analytical, practical, creative) outlined in the NSW differentiation guide at https://education.nsw.gov.au/teaching-and-learning/professional-learning/teacher-quality-and-accreditation/strong-start-great-teachers/refining-practice/differentiating-learning/strategies-for-differentiation, generate a set of three different project options for a Year 10 History unit on World War II. Each option should cater to one of Sternberg’s intelligence types while still meeting the core learning objectives. Include how each project option allows students to demonstrate their understanding in ways that align with their preferred learning style.

Differentiation includes modifying work to be suitably complex, as well as reducing the complexity. “Compacting” is a strategy which involves “the process of eliminating teaching or student practice if students have already mastered a concept or skill.” The following prompt incorporates the process of compacting to reduce the unnecessary aspects of a Year 9 Mathematics unit.

Using the compacting strategy described below, create a plan to compress the attached Year 9 Algebra unit for advanced learners who have already mastered basic concepts. Outline a pre-assessment to identify students’ current knowledge, suggest advanced topics these students could explore instead, and propose meaningful extension activities that deepen their understanding of algebraic concepts in real-world contexts.
Steps for compacting:

1. identifying the learning objectives or standards that all students must learn
2. offering a pre-test opportunity or planning an alternate path through the content for those students who can learn the required material in less time than their age peers
3. planning and offering meaningful curriculum extensions for students who qualify
4. eliminating all drill, practice, review or preparation for students who have already mastered such things
5. keeping accurate records of students’ compacting activities

Another way of differentiating curriculum materials might be to use a “layered curriculum” approach (Nunley, 2014). Like the example above, it is useful to use an internet connected model and begin with a simple search prompt like “What is Kathie Nunley’s layered curriculum model?”, and then use a more complex prompt like the following:

Design a layered curriculum approach for a Year 8 English unit on Shakespeare’s ‘A Midsummer Night’s Dream’. Following Kathie Nunley’s model, create activities for layers C, B, and A that progressively move from basic understanding to application and critical thinking. Ensure that each layer offers student choice and addresses different learning preferences.

The final example uses a specific model – Claude 3.5 Sonnet – to create an interactive “artifact” which can be shared with students, and which offers differentiated tasks suitable for a range of learners:

Develop a cubing activity for a Year 11 Geography class studying climate change. Create prompts for each face of the cube (describe, compare, associate, analyse, apply, argue for/against) that encourage students to examine climate change from multiple angles. Ensure that the prompts cater to different levels of complexity and allow for diverse modes of expression to accommodate various learning needs and interests. Create an interactive artifact.

These prompts demonstrate how GenAI can be used to support differentiation strategies that go beyond simple adjustments to reading levels, focusing instead on creating varied learning experiences that address students’ individual needs, interests, and learning profiles.

Key ideas for Designing with GenAI

To wrap up this exploration of differentiating with generative AI in education, here’s a summary of key ideas that educators can keep in mind when approaching the technology:

1. Understand AI limitations and provide context: Recognise that GenAI models may default to outdated or oversimplified educational theories. Always provide clear context about grade level, subject, and specific learning objectives to ensure relevant and appropriate content generation.
2. Use established frameworks and strategies: When designing differentiated activities, reference established educational frameworks and best practices. Explore various differentiation strategies like tiered instruction, compacting, layered curriculum, and cubing to create diverse learning experiences.
3. Balance individual and group learning: Remember that differentiation doesn’t always mean isolated, individual work. Design activities that incorporate both personalised learning and collaborative approaches, recognising that many students thrive in community-based learning environments.
4. Incorporate student choice and real-world applications: Design activities that allow students to demonstrate their learning in ways that suit their strengths and interests. Encourage the AI to generate tasks that connect learning to real-world contexts, especially for advanced learners or extension activities.
5. Iterate and refine with human insight: Use GenAI as a starting point for ideas, but always refine and adjust the output based on your professional judgment and knowledge of your students. Remember that while GenAI can assist in creating differentiated resources, the teacher’s role in understanding individual student needs and providing personalised support remains crucial.

Like anything else when working with GenAI, the technology isn’t a magic bullet for differentiation. “Personalised learning”, as sold by tech companies, is mostly snake oil. But that doesn’t mean the technology can’t be used to create relevant, engaging resources that suit many learners: you just need to know how to apply your own expertise as an educator.

Want to learn more about GenAI professional development and advisory services, or just have questions or comments? Get in touch: