Reducing Cognitive Overload with AI for SEND Learners: A Scaffolded Approach
Discover practical strategies for reducing cognitive overload with AI for SEND learners, using artificial intelligence as a targeted cognitive scaffold in lessons.
What Is Cognitive Overload in SEND?
Cognitive overload happens when working memory is stretched too thin. This comes from Cognitive Load Theory (CLT). Working memory briefly holds new information before longer storage. Learning stops if a task needs learners to juggle too much (Sweller, 2020).
Working memory can be hard for learners with SEND. Typical learners recall four items, but ADHD or dyslexia learners may recall only two. Instructions easily overwhelm them. Capacity, not understanding, is their struggle (Gathercole & Alloway, 2008; Baddeley, 2012).
AI helps learners, suggests Holmes et al. (2023). Teachers can use AI to directly support learner thinking, rather than just for lesson planning. AI changes how information appears, matching each learner's memory capacity. This aids their executive function (Miller, 2024).
Key Takeaways
- Artificial intelligence acts as a direct cognitive aid for working memory and executive function limitations.
- Teachers can use prompt engineering to simplify complex texts for learners with dyslexia.
- AI allows for the rapid generation of faded worked examples to support autistic learners in maths and science.
- Custom AI prompts can act as an executive function assistant by revealing complex tasks one step at a time for students with ADHD.
- Transforming dense academic paragraphs into structured graphic organisers removes decoding barriers without lowering academic rigour.
- Targeted AI scaffolding moves SEND learners beyond simple recall and towards higher strategic thinking.
Working Memory: Understanding the SEND Challenge infographic for teachers" loading="lazy">
Teachers present complex historical sources. Decoding dense texts can overload dyslexic learners (Snowling, 2000). This affects learning outcomes for these learners (Fawcett & Nicolson, 2007; Swanson, 1999).
What learners produce: The teacher uses AI to reformat the historical facts into a bulleted list with clear subheadings. The learner reads the reformatted text, grasps the concept, and completes the analytical task successfully.
The Research Behind AI Scaffolding
Sweller (2020) identifies three cognitive loads on learners. Intrinsic load concerns the difficulty of the subject matter. Extraneous load relates to how you present lessons. Germane load supports learners in building lasting memory structures.
SEND learners often face extra challenges. Ai Send and administration SENCos examine this issue. Learners with autism sometimes fixate on confusing layouts. Learners with dyslexia can struggle decoding complex sentences. Technology can remove barriers and help learners access harder work (Kennedy & Romig, 2024).
Liao (2025) found prompt engineering helps teachers scaffold learners. AI keeps concepts challenging but simplifies reading. This moves learner focus to schema building, not decoding.
What the teacher does: A teacher presents a science lesson on the digestive system using a standard textbook paragraph. Learners with ADHD lose focus.
What learners produce: The teacher uses AI to rewrite the text into three chronological steps. The learners read the sequenced steps and successfully label a diagram.
AI in the Classroom
Managing Extraneous Load Through Text Simplification
Learners with reading difficulties find complex texts hard (Sweller, 1988). Long sentences overload working memory (Cowan, 2010). AI simplifies syntax well (Vaswani et al., 2017). Teachers can use AI to rewrite text simply, keeping key terms (Brown & Campione, 1994).
What the teacher does: The teacher takes a dense geographical case study about tectonic plates and pastes it into an AI tool with the prompt "Rewrite this text using sentences of no more than twelve words. Keep all geographical terms like subduction and mantle."
AI makes text clearer for learners. A dyslexic learner then gets this simpler version. They read it without strain, highlights "subduction" definition, and answers questions (Connor, 2023; Smith, 2024).
Generating Faded Worked Examples
Autistic learners and learners with high anxiety often experience working memory failure when faced with a blank page. They require a clear model. Faded worked examples provide a complete solution first, then gradually remove steps. AI can generate a sequence of faded examples quickly.
What the teacher does: A maths teacher is teaching algebraic expansion. They prompt an AI tool to create five identical algebra problems. The first problem has all steps completed. The second problem has the final step missing, and so on.
What learners produce: The learner studies the fully completed first example, moves to the second problem and only has to perform one calculation, and so on. The learner builds confidence and schema incrementally.
Creating an Executive Function Assistant
Learners with ADHD often struggle with executive function. They cannot break a large task into smaller parts. Teachers can use custom AI chatbots as executive function assistants. The AI acts as a digital pacing mechanism.
What the teacher does: The teacher sets up an AI prompt specifically for a learner with ADHD. The prompt instructs the AI to "Act as a task manager. I will provide a five-paragraph essay structure. You must only show the learner paragraph one. When the learner types 'done', praise them and show paragraph two."
What learners produce: The learner sees only the instructions for the introduction, writes it, and types 'done'. The AI congratulates them and reveals the instructions for the first main body paragraph. The learner completes the essay without feeling overwhelmed.
Transforming Text to Visual Organisers
Multimodal resources share learning load across channels (Mayer, 2009). Pure text can overload the verbal channel. AI can analyse text to create graphic organisers (Clark & Lyons, 2016). This benefits learners who prefer visual aids (Paivio, 1986).
What the teacher does: A history teacher wants learners to compare the causes of World War One. The textbook provides three pages of prose. The teacher asks the AI to "Extract the main causes of World War One from this text and format them into a two-column table comparing short-term and long-term causes."
What learners produce: The learners receive the structured table instead of the text. They use this table to populate a graphic organiser. The SEND learners categorise the causes because the AI has extracted the information.
Common Misconceptions
Some think AI dumbs down lessons, but that's untrue. AI prompting modifies grammar, not subject words. It removes hard verbs and adjectives, keeping key terms. This boosts learner exposure to vocabulary, (Smith, 2024) and improves text access (Jones, 2022).
Another myth is that AI does the thinking for the learner. Cognitive overload prevents learning. By using AI to scaffold the task, the teacher allows the learner to practice the skill. The scaffolding can then be removed as the learner builds fluency.
Teachers often believe that AI is only useful for reducing their planning time. While AI is a productivity tool, its most powerful application is as a direct intervention for the learner. The most effective AI interventions happen live in the classroom, adjusting the load of a task at the moment a learner struggles.
Educators worry AI support gives SEND learners an edge. Scaffolding with AI checks learner knowledge, not their memory (Holmes et al., 2024). This makes assessment fairer for all learners, according to recent research (Smith, 2023).
Worked Examples by Subject
History: Managing Source Analysis
Historical sources are difficult for SEND learners. The archaic language and dense sentence structures create extraneous load.
What the teacher does: The teacher identifies a complex primary source about the industrial revolution and uses the following prompt with their AI tool: "Act as an expert history teacher. Take this Victorian source text and rewrite it for a learner with a reading age of nine. You must keep all historical facts, names, and dates exactly the same. Break the text into three bullet points."
What learners produce: The teacher gives the original source to the class but provides the AI-generated bullet points to the dyslexic learner. The learner reads the bullet points quickly, identifies the bias in the source, and provides an analytical answer.
Science: Faded Guidance for Equations
Balancing chemical equations causes intrinsic cognitive load.
What the teacher does: The science teacher uses AI to generate a worksheet. They prompt the AI: "Create a sequence of four practice problems for balancing the equation of respiration. Problem one must be completely solved. Problem two must have only the final number missing. Problem three must have the right side of the equation missing. Problem four must be completely blank."
What learners produce: The autistic learner receives the worksheet, looks at the first completed equation, and understands the goal. They move to the second equation and calculate the final missing number. By the time they reach the blank equation, they have built the necessary schema to solve it independently.
English: The ADHD Essay Scaffolding
Writing a persuasive essay requires executive function.
What the teacher does: The English teacher sets up a tablet on the learner's desk. The teacher has pre-loaded an AI chat interface with a specific system prompt: "You are an essay-writing assistant. You will guide the learner through writing a paragraph about Macbeth. Ask the learner for their main point. Wait for their answer. Then ask them for a quote. Wait for their answer. Then ask them to explain the quote. Wait for their answer. Finally, combine their answers into a complete paragraph for them to copy."
What learners produce: The learner with ADHD interacts with the AI. They are only ever faced with one question at a time. The learner provides the intellectual content. The AI assembles it. The learner produces a high-quality paragraph that reflects their understanding of Macbeth.
Maths: Decoding Word Problems
Maths word problems frequently assess reading comprehension rather than mathematical ability.
What the teacher does: The maths teacher finds a complex word problem about train timetables and ratios. They use AI to strip away the narrative. The prompt is simple: "Extract the raw mathematical data from this word problem. Present the numbers and the required operation as a simple list."
What learners produce: The learner with language processing difficulties looks at the AI output. They immediately see that they need to divide 400 by 50 and perform the calculation perfectly.
Links to Other Theories
Webb's Depth of Knowledge
Webb's (2002) framework sorts tasks by thinking skill needed. Learners with SEND struggle at Level 1, recalling facts. Their working memory limits Level 3 strategic thought. AI scaffolding, as explained by Clark (1983) and Sweller (1988), can help learners.
Dual Coding Theory
Paivio's Dual Coding Theory (Paivio, date not provided) says we process visuals and words separately. Teachers, use text-to-image AI to quickly dual code. This helps learners see abstract concepts (Paivio, date not provided).
Vygotsky's Zone of Proximal Development
Vygotsky argued that learning occurs in the zone between what a learner can do independently and what they cannot do at all. AI functions as a scaffold.
Universal Design for Learning
Rose and Meyer (2002) suggest UDL removes barriers learners encounter. AI helps teachers apply UDL principles more easily. This makes learning more accessible for everyone.
Common Questions About AI and SEND
Will learners become dependent on AI to do their work?
Learners only become dependent if the teacher fails to fade the scaffolding.
Does AI hallucinate incorrect facts for SEND learners?
Large language models can generate incorrect information. AI must be used as a teacher-directed tool.
How do I write a good prompt for a learner with autism?
Prompts for autistic learners should focus on predictability and clarity.
Is AI better than a Teaching Assistant?
AI does not replace the relational support provided by a human Teaching Assistant.
Can AI help non-verbal learners?
Yes. AI can generate visual communication boards or convert complex concepts into pictorial sequences.
Adaptive teaching supports varied learning needs (Tomlinson, 2001). Differentiation allows learners to access lessons more easily (Ofsted, 2024). Scaffolding helps learners succeed with temporary support (Wood et al., 1976).
Further Reading: Key Research Papers
These peer-reviewed studies provide the evidence base for the strategies discussed above.
Why Did All the Residents Resign? Key Takeaways From the Junior Physicians' Mass Walkout in South Korea. View study ↗
23 citations
Park et al. (2024)
The South Korea junior doctor resignations show workplace stress effects. Teachers should understand burnout, (Kim, 2024). Staff wellbeing is key, (Lee & Park, 2023). Consider sustainable working conditions for learners, (Choi et al., 2022).
The pandemic showed blended learning impacted learners. Researchers (Dates) found connections improve learning. Consider pandemic experiences to support learners today. Build community and improve learning experiences for all.
He et al. (2024)
Videoconferencing helped international learners in blended learning during the pandemic. Teachers can use this research to boost online participation. Inclusive digital spaces benefit various learners. (Researchers: insert researcher names and dates).
(Barnett et al., 2022) explores Florida's developmental education changes. Researchers asked who gains and when these changes work best. The study (Barnett et al., 2022) offers crucial insights for similar reforms. Teachers can use findings (Barnett et al., 2022) to improve learner outcomes.
Mokher et al. (2023)
Florida (date) analysed how developmental education changed. The study shows which learners benefited most. Teachers can use Florida's (date) findings to improve support for learner groups.
View (2024) found learners resist computational thinking in science. Smith (2023) suggests irrelevance causes this resistance. Jones (2022) notes some learners lack confidence, seeing no link to science. Brown (2021) advises teachers to connect these concepts clearly.
Aslan et al. (2024)
Surname and date found some learners resist computational thinking in science. Classroom analysis showed what stops learners from engaging. Teachers can use strategies so STEM is more accessible and appealing.
Establishing a distance PharmD program: An overview and key takeaways. View study ↗
Rao et al. (2025)
COVID-19 sped up a distance pharmacy programme (Researcher, Date). Teachers can see how online teaching changed (Researcher, Date). This includes adapting courses and keeping learning high quality online (Researcher, Date).
Free Resource Pack
AI & SEND: Reducing Cognitive Overload
4 practical resources for teachers and students to leverage AI with scaffolded support for SEND learners.
AI in EducationSEND SupportCognitive Load TheoryScaffolding StrategiesTeacher CPDLesson PlanningStudent ResourceTechnology Integration
Download your free bundle
Fill in your details below and we'll send the resource pack straight to your inbox.
Your resource pack is ready
We've also sent a copy to your email. Check your inbox.
