Written by Betty Johnson
Betty Johnson is Director of Rinnovate Recruitment & Consultancy and founder of B3Inspire, a women's empowerment coaching brand. She is a School Improvement Partner, AI in Education advocate, and contributor to an upcoming Routledge publication exploring AI and primary school leadership.
Artificial intelligence is reshaping classrooms at speed. But if we are not careful, we risk automating the very inequalities we have spent decades trying to dismantle.
There is a moment I return to often. I was sitting with a group of teachers in South London, brilliant and committed educators, watching them use an AI writing assistant for the first time. The excitement was palpable. Then one teacher paused and asked quietly: “But does it know our children?”
That question has stayed with me. Because behind it lies the central challenge of our time in education: how do we harness the extraordinary power of artificial intelligence while ensuring it serves every child? Not just the ones whose experiences, languages and cultures it was trained on.
The promise is real
Let us start with what AI can genuinely do. Adaptive learning platforms can tailor content, pacing and assessment to individual learners in ways that even the most devoted teacher simply cannot replicate in real time when managing a class of thirty. Intelligent tutoring systems can offer immediate, personalised feedback. For students with SEND, for English as an Additional Language learners, and for children in under-resourced schools, this is not a trivial benefit. It is potentially transformative.
Research published in AI and Ethics (Fitas et al., 2025) confirms that AI can meaningfully support special needs provision and address language barriers, provided it is deployed thoughtfully and with equity at the centre of design.
The peril we are not talking about loudly enough
Here is the uncomfortable truth. AI systems learn from data. And the data that exists in the world reflects the world as it has been, not as it should be. When a model is trained predominantly on content that centres Western, English-speaking, middle-class experience, it does not arrive at the classroom as a neutral tool. It arrives carrying biases.
Research published in the International Journal of Artificial Intelligence in Education (Baker & Hawn, 2022) found that algorithmic bias in educational AI can produce systematically unfair outcomes across assessment and instruction, disproportionately affecting the very students who most need support. A 2025 review in Frontiers in Education confirmed that unchecked AI models can continue to amplify underlying social and cultural biases, impacting equity and fairness in both assessment and instruction.
“The environment the machine operates in is social in nature. It functions within the socio-technical system that encompasses the cultural values and beliefs of the people using it.”
Put simply: AI is not neutral. It is a mirror. And if we have not yet built a system that reflects the full richness of our children’s identities, we should not be surprised when the mirror distorts.
The digital divide: a growing fault line
There is a second, more immediate danger. Access to AI tools is not equally distributed. Schools in disadvantaged communities face significant challenges in investing in technology and establishing fair access policies. A 2025 study in Frontiers in Education warned explicitly that equitable access to AI tools is crucial to prevent educational inequalities from widening further.
Attainment outcomes continue to vary significantly between ethnic groups in England, with some communities continuing to experience substantial educational disadvantage despite decades of policy attention and intervention. Education Policy Institute data indicates that Gypsy, Roma and Traveller pupils experience some of the largest attainment gaps in England, equivalent to approximately 30 months of learning behind their peers by the end of secondary school. Meanwhile, Universities UK continues to report degree awarding gaps between White students and several minority ethnic groups across higher education.
If AI becomes the new frontier of educational provision and some children cannot access it, or access a version of it that does not truly see them, we will not close those gaps. We will cement them.
What can we actually do? Seven practical steps
- Ask the equity question first.
Before deploying any AI tool in your school, ask who it was trained on. Ask who it serves well and who it might underserve. Demand transparency from providers about their training data and bias testing. This is not a technical question. It is a values question.
- Use AI to reduce teacher workload, not to replace teacher relationship.
The most powerful protective factor for disadvantaged students is a trusted adult who knows them. Use AI to free up that time, for planning, marking and administration, so teachers can invest more of themselves in the human connections that matter most.
- Curate and contextualise AI outputs.
Do not present AI-generated content as given. Teach children and staff to interrogate it. Whose perspective is missing? What has been assumed? This is media literacy for the AI age and it is a diversity lesson in itself.
- Pair AI with inclusive curriculum design.
AI is a tool, not a curriculum. Ensure that what children are doing with AI reflects a genuinely diverse and representative body of knowledge, including authors, histories and contributions from across the global majority.
- Invest in staff training with diversity at the centre.
AI professional development must include bias recognition and culturally responsive pedagogy, not just technical skills. Teachers need to understand not only how to use these tools but when to challenge them.
- Amplify student voice in AI decisions.
Which students are at the table when your school decides how to use AI? If the answer is mainly those who already have power, change the table. Invite your most marginalised students into the conversation. They will see what others miss.
- Advocate beyond your school.
The decisions that shape AI in education are being made at government, policy and corporate levels. Join or create the conversations that push for representative datasets, diverse AI development teams and equity-first regulation. Your voice belongs in those rooms.
A word to women leaders in education
I want to speak directly to the women reading this, particularly those who, like me, came into this profession from communities that have historically been spoken about in education rather than listened to within it.
We are not incidental to this conversation. We are central to it. The research is clear that cultural and social context shapes how AI interacts with learners. We carry that contextual knowledge. We have spent careers navigating systems that were not built with us in mind and we have built extraordinary things within them regardless.
The question is not whether AI will change education. It already is. The question is who gets to shape it. The answer should include every teacher in every underfunded classroom, every headteacher in a school where the children look like the world, and every leader who has ever been the first, the only, or the one they did not expect.
“We do not get a fairer future by accident. We build it, deliberately, strategically and together.”
The next time someone asks whether AI is good or bad for education, resist the binary. The better question is simply this: good or bad for whom? Ask that question consistently and loudly, at every level, and AI in education may yet become the equaliser it promises to be.
References
Baker, R.S. & Hawn, A. (2022). Algorithmic bias in education. International Journal of Artificial Intelligence in Education, 32(4), 1052-1092.
Fitas, R. (2025). Inclusive education with AI: supporting special needs and tackling language barriers. AI and Ethics, 8(2), 115-129.
Education Policy Institute (2024). Annual Report: Ethnicity Gaps.
Universities UK (2019). Black, Asian and Minority Ethnic Student Attainment at UK Universities: Closing the Gap.
Frontiers in Education (2024). Promoting equity and addressing concerns in teaching and learning with artificial intelligence.