In my previous post on the topic of AI and inclusion I highlighted Hull College’s use of Microsoft Translate as a standout example of socially beneficial AI. Without such innovations, the educational progress of displaced students coming to the UK would have to wait until their command of English caught up. With AI, this bottleneck is eased significantly. Since publishing that piece, I have been quietly pondering whether we should have in our minds a broader understanding of AI’s translation capabilities.
As a monoglot I tend to consult friends and family members to gauge AI’s multilingual credentials. And I have it on good authority that translations between English and Farsi, Italian and Spanish are stunningly accurate. At the same time, I am in the habit of using AI to help me rewrite my work so that it adheres to certain style guides, for instance. In other words, AI helps me to translate between one mode of English to another.
Recently, a discussion with Kelly Webb-Davies, Lead Business Technologist at University of Oxford’s AI & Machine Learning Competency Centre and an expert in linguistic inequality, helped crystallise my thinking around AI translation; and, moreover, brought to light its relevance to social inclusion.
Here’s a flavour of her thinking on the subject:
“I think it’s important to recognise that being able to express your ideas clearly in standardised written English is a form of privilege. Not everyone has equal access to the linguistic norms that are often expected in professional and academic spaces. For many people, like those who speak English as an additional language, varieties of stigmatised English, or who are neurodivergent, tools like generative AI can be used for linguistic accessibility.” (Post on LinkedIn)
Having begun my career teaching science and maths at a school in a very deprived area, this insight resonated with me. I’ve known many brilliant secondary students, brimming with ideas and razor-sharp in their reasoning, whose confidences were frequently dented by challenges in wielding ‘standard’ English.
As per Kelly’s arguments, AI offers the promise of linguistic accessibility for my former students. I can imagine them using Chat GPT to help rephrase parts of their personal statements. Or asking Gemini to help them draft an important email to someone senior within their college, university or place of employment. I am confident that in most cases these tools will not be used to substitute for individuals’ intelligence but to adhere to expectations of how intelligence should be communicated: expectations that are often arbitrary.
So perhaps the takeaway is that generative AI should be used to help us game a linguistically rigged system – allowing us to translate from ‘bad’ English to ‘good’. That’s certainly part of the picture. But there’s more to consider.
Adapting how you communicate is a vital and resoundingly human skill. As a teacher, for instance, I needed to use language differently depending on whether I was interacting with younger students, older students, parents or colleagues. Similarly, academics will frequently have to communicate the same ideas to their peers and to the general public – audiences with starkly different levels of prior knowledge and motivations for engaging.
Failing to equip students to modify how they use language would be a dereliction of duty. But that doesn’t preclude ChatGPT’s role here. I’ll try to demonstrate why with an example.
As a former science teacher and long-time uncle—now also a new father—I’ve always enjoyed sharing my enthusiasm for science with the youngest members of my family. However, I often find that my explanations, while accurate, go over their heads. To bridge this gap, I’ve started using ChatGPT in a couple of ways. Sometimes, I input my explanation of a concept and ask ChatGPT to assess its suitability for, say, a 7-year-old. Other times, I write out my own version (without sharing it) and then ask ChatGPT to generate its own explanation from scratch. By comparing the two, I can identify differences and adjust my approach accordingly. Through this iterative process, I’ve improved my ability to communicate scientific concepts to children, reducing my reliance on external tools over time. In this way, generative AI doesn’t just translate my words into a more accessible form—it helps me develop a new way of communicating.
The same logic can be applied to a student who is the first in their family to attend university, and to whom an academic style of writing feels intimidating. Such a student will benefit from learning how to write in a way that balances command of a topic with critical flair – attribution with independence of thought. That said, they needn’t be thrown in the deep end, struggling to articulate well-thought-through ideas while grappling with unfamiliar conventions.
For students like these, generative AI can provide both short-term guardrails, and long-term solutions. But the onus for addressing linguistic inequalities mustn’t fall solely on students or technical innovations. Norms and expectation about how English should be used are sometimes arbitrary.
To help you examine underlying biases, try this: use generative AI to create a well-argued essay in non-standard English. Then have it generate a weakly reasoned piece in eloquent prose. Compare your reactions. Are you judging based on content or form? If appropriate, share these outputs with peers to spark further discussion.
Just as regional accents can disadvantage job interviewees, uses of English that don’t conform to prescribed conventions (using ‘less’ instead of ‘fewer’, ‘I’ instead of ‘me’ etc.) has the potential to trigger unconscious biases in academic contexts, perhaps resulting in brilliant students being marked down – or not even being accepted in the first instance. A key corollary here is that institutions and individuals in a position of power should be mindful of whether their expectations around how English is used are creating unfair barriers for less privileged learners.
In summary
I am becoming increasingly convinced that AI can help level the playing field by reducing linguistic inequality. This is not just because generative AI tools can translate great ideas into the version of English that is considered proper – AI can also support people to diversify and adapt how they communicate.
That said, it is important to reflect on whether our expectations of how English should be written are unduly disadvantaging different groups of learners. Generative AI can support us in undergoing such reflections.
Find out more by visiting our Artificial Intelligence page to view publications and resources, join us for events and discover what AI has to offer through our range of interactive online demos.
For regular updates from the team sign up to our mailing list.
Get in touch with the team directly at AI@jisc.ac.uk
