Making Technology Work for All Learners
How thoughtful AI implementation can transform education for neurodiverse students and those with disabilities
I’ve spent the past several months exploring the world of educational technology, and I’m excited about artificial intelligence’s potential to create more inclusive learning environments. The possibilities are genuinely transformative;but only if we approach AI implementation with care, intention, and a commitment to serving all learners.
The Promise of AI in Inclusive Education
Educational technology has long promised to personalise learning, but AI takes this potential to new heights. Well-designed AI systems can be helpful for neurodiverse students and those with disabilities. They can provide real-time support and offer multiple pathways to understanding;all without the stigma sometimes associated with traditional accommodations.
Universal Design for Learning as a Framework
The key, I’ve found, is aligning AI tools with Universal Design for Learning (UDL) principles from the outset. UDL provides a brilliant framework built on three core principles:
- Multiple means of representation: Presenting information in varied formats (text, audio, visuals) to accommodate different perceptual abilities and preferences.
- Multiple means of action and expression: Offering different ways for students to interact with content and demonstrate their knowledge.
- Multiple means of engagement: Providing varied approaches to motivate learners and maintain their interest.
AI can “supercharge” each of these principles in remarkable ways. For representation, AI-driven platforms can dynamically adapt how information is presented;converting text to speech, generating visual representations of concepts, or simplifying language. For expression, AI can enable students to respond through spoken answers, typed responses, or interactive simulations. For engagement, AI can tailor activities to student interests and provide immediate, personalised feedback that keeps them motivated.
As CAST (the organisation behind UDL) emphasises, the real power emerges when AI tools initially integrate these principles rather than as an afterthought. This approach enables educational experiences that support all learners, reducing the need for retrofitting or special accommodations later.
Building AI Systems That Work for Everyone
I find it fascinating how inclusive design principles improve AI for all users, not just those with identified needs. When developers design with the margins in mind, everyone benefits.
Unfortunately, the status quo falls short. A 2023 survey revealed that fewer than 7% of respondents with disabilities felt adequately represented in AI product development;despite 87% being willing to provide feedback. This stark disparity highlights a troubling oversight in how we’re building educational technology.
The Power of Co-Design
The solution starts with co-design. AI systems should be developed with input from diverse learners, including students with disabilities and those who are neurodiverse. By involving end-users throughout the process;from ideation to prototyping to testing;developers can identify and remove barriers before they become baked into the final product.
For example, if developers are creating an AI social skills coach, they should consult autistic students to ensure the programme is respectful and comfortable to use. If they’re building a reading app, they should test it with dyslexic students to confirm font options and read-aloud features meet actual needs. This approach isn’t just ethically sound;it results in better products.
By designing for the margins, we create AI systems that serve everyone more effectively. As CAST advises, inclusive design should be treated as a non-negotiable foundation;AI tools should be flexible and accessible to all from the start, rather than retrofitted later.
Making Learning Accessible, Personal, and Engaging
Well-designed AI tools can enhance education in three critical ways: accessibility, personalisation, and engagement.
Enhanced Accessibility
AI is opening new pathways for students with various disabilities to access educational content on equal footing with peers. These technologies align perfectly with UDL’s principle of multiple representation, ensuring information is accessible to all:
- Speech-to-text functionality allows students with physical impairments or writing difficulties to dictate their responses, converting spoken words into written text.
- Text-to-speech systems help those with dyslexia or visual impairments consume written material through audio channels.
- Automatic captioning and transcription has improved markedly for deaf or hard-of-hearing students. The Speech Accessibility Project has reduced speech recognition errors for people with non-standard speech patterns from 20% to 12%.
- AI vision tools assist students with low vision by recognizing and describing images or even physical environments. AI add-ins for screen readers can generate rich image descriptions of any website or interface.
These technologies provide multiple avenues to access information;auditory, visual, tactile;significantly reducing accessibility barriers. The result is that students with disabilities can gain more equitable access to educational resources, helping fulfill legal and ethical mandates for equal opportunity.
Personalised Learning
Perhaps AI’s greatest strength is its ability to tailor learning experiences in real-time. Adaptive learning software uses AI algorithms to analyse a student’s performance, pace, and preferences, then adjusts content accordingly:
- Students who excel in an area receive more advanced challenges
- Students who struggle get additional practice or simplified explanations
- Content delivery adapts to learning style preferences (e.g., more visual supports for visual learners)
This personalisation can be transformative for neurodiverse learners with unique patterns of strengths and challenges. Research shows that such customisation helps fill individual learning gaps and improves outcomes by addressing specific learning differences.
AI tutors provide one-on-one guidance that teachers simply don’t have time to offer to every student. A teacher might configure an AI writing “sidekick” to assist students in a narrative writing assignment by giving hints about figurative language without providing answers. This scaffolded support allows students to develop skills at their own pace while still doing the cognitive work themselves.
The flexibility of AI personalisation aligns with UDL’s call for multiple means of action and expression;students can express knowledge in different ways and follow varied pathways. Every learner’s interaction with an AI tool might look different; in a class using an AI chatbot for research, one student may engage in a lengthy Q&A while another uses it to get examples before jumping into hands-on work.
Increased Engagement
Keeping students motivated is a perpetual challenge, particularly for neurodiverse learners who might disengage due to past struggles or lack of interest in standardised content. AI offers new engagement strategies through interactivity, relevance, and responsiveness:
- Interest-based customisation: If a student is passionate about music, an AI maths tutor might frame problems in musical contexts to capture their attention.
- Game-like elements and rewards tuned to individual preferences provide immediate positive reinforcement.
- Agency and control: Learners can choose how to interact with material;watch a video, read an article, ask questions;increasing their ownership of the learning process.
Studies show that when students feel actively involved and supported by immediate feedback, their overall engagement increases. In one study, students reported that AI feedback and the ability to make choices in their learning path enhanced their interest and engagement with the material.
Mitigating Bias in Educational AI
Despite its promise, AI comes with risks;particularly bias that could harm marginalised students. Mitigating bias requires vigilance and strategic approaches:
Inclusive Development Processes
When diverse stakeholders help shape an AI system, they can uncover biases that developers might miss. During testing, neurodiverse students might point out that an AI tutor’s feedback feels discouraging or that questions are confusingly phrased;allowing adjustments before deployment.
Diverse Training Data
AI models learn from data. If that data lacks representation or contains biased patterns, the AI will reflect those biases. Developers should use training datasets that represent the diverse student populations the AI will serve:
- Speech recognition AI should be trained on diverse voices, including varying accents and neurodiverse speech patterns
- Essay analysis systems should see writing samples from English language learners and students with different expression styles
- Testing AI on protected groups is critical to check if accuracy is consistent across populations
Schools evaluating AI products should ask vendors whether and how they tested for fairness with diverse student data. Demanding transparency about training data and fairness audits can incentivize companies to prioritize bias mitigation.
Transparency and Explainability
Bias can hide in the complexity of AI algorithms. Educational AI systems should not be “black boxes” to educators and administrators:
- Providers should disclose what data was used to train the AI and how the algorithm makes decisions
- Explainability features should allow users to understand why certain recommendations or predictions were made
- School leaders should ask specific questions: “What student data was the AI trained on? Were students with disabilities represented? Can we review how it works?”
Clear documentation and communication about an AI’s design help identify bias early. If a vendor cannot satisfactorily answer questions about fairness and representation, that’s a red flag.
Human Oversight and Intervention
No AI system in education should operate without human oversight, especially given the high stakes of student learning and wellbeing:
- AI tools should assist, not replace, teacher judgment
- Teachers should review and approve AI-generated content before presenting it to students
- Educators should have access to logs or transcripts of AI interactions with students
- Regular review of AI outputs can help catch and correct biased content
This approach augments human decision-making rather than automating it, reducing the risk that biased AI outputs go unchecked.
Addressing the Digital Divide
If advanced AI tools are only available to well-resourced schools, existing inequities will worsen:
- Schools should advocate for affordable pricing or licensing models for AI educational software
- Districts might need to invest in infrastructure (expanding Wi-Fi or providing devices)
- Policies should support equal access to AI tools for all students
- Offline or low-bandwidth modes of AI apps can help bridge access gaps
Without attention to these issues, the benefits of AI for personalised learning will disproportionately help those already advantaged.
Continuous Bias Monitoring and Improvement
Even after deployment, bias mitigation is an ongoing process:
- Schools should continuously monitor outcomes for signs of disparity across student groups
- Performance data should be analysed by subgroup to ensure all students are benefiting
- Regular product testing with diverse student groups should be iterative
- Student feedback about fairness and helpfulness can highlight biases that data might miss
Schools can maintain an equitable learning environment by creating feedback loops and being willing to adjust or withdraw problematic AI tools.
Practical Guidelines for Teachers and Schools
How can educators harness AI’s potential while avoiding pitfalls? Here are practical strategies:
Best Practices for Integration
- Align AI tools with UDL principles, ensuring they offer content in different formats and allow various interaction methods. If a tool lacks flexibility, supplement it with additional resources.
- Start with a clear purpose rather than using AI as a gimmick. Identify specific needs that AI could address, such as providing reading support for dyslexic learners or enrichment for gifted students.
- Build AI literacy through professional development for teachers and explicit instruction for students. Teachers need to understand the capabilities and limitations of AI tools, while students should learn to use AI as a “research assistant” or coach, not a shortcut. Teach students to ask good questions and fact-check AI responses.
- Introduce AI gradually through pilot programmes before scaling. Start with one class or subject to work out issues and gather feedback. For neurodiverse students who rely on routine, introduce AI as part of the daily structure.
- Monitor student engagement with regular check-ins. Some students may need encouragement or scaffolding;for example, a student with attention difficulties might need help setting small goals when using an AI tutor. Balance AI activities with human interaction, perhaps pairing students to discuss what they learned from the AI.
School Policies for Ethical AI Use
Schools should establish clear policies covering:
- Student privacy and informed consent: Be transparent with families about what data is collected, how it’s used, and who has access. Offer opt-out alternatives if families are uncomfortable.
- Academic integrity guidelines: Define acceptable use of AI for schoolwork. Update honor codes to address AI assistance while encouraging its appropriate use as a learning aid.
- Fairness and non-discrimination: Ensure AI-based recommendations are never the sole determinant of tracking, placement, or grading decisions. Guard against “digital redlining” where certain students are steered toward different content.
- Accountability and oversight: Establish a committee or designate administrators responsible for monitoring AI integration. Periodically audit the AI’s impact on achievement gaps and address any incidents of bias.
- Accessibility compliance: Mandate that all educational technology, including AI, meets accessibility standards in accordance with legal requirements.
- Professional development: Provide training whenever a new AI tool is introduced, covering not just how to use it but how to use it inclusively and fairly.
- Continuous review and community involvement: Stipulate regular reviews of AI tools’ performance and equity impact, involving student and parent feedback. Be ready to course-correct if a tool isn’t meeting inclusive ideals.
The Path Forward
As AI becomes increasingly prevalent in education, we face a critical choice: will these tools widen or narrow existing gaps? The answer depends on our commitment to inclusive design, bias mitigation, and thoughtful implementation.
By integrating UDL principles into AI systems, involving diverse learners in design, and maintaining human oversight, we can harness technology to create more equitable classrooms. The goal isn’t to replace teachers but to augment their ability to reach every student.
I’ve come to believe that AI’s true potential lies not in efficiency or automation, but in its capacity to meet each learner where they are. When implemented with care and intention, AI can help create classrooms where differences are not just accommodated but valued;where technology and inclusive teaching work hand in hand to empower every student to thrive.