Sarah: Changing the Narrative

12/08/2026

Communication in the Age of AI

Part 3/5 – AI Isn’t Neutral

The more I explore AI, the more I realise that two things can be true at once.

AI can be an incredibly useful tool. For some disabled people, it can remove barriers, conserve energy and increase independence.

But calling something a tool doesn’t make it neutral. A knife and hammer are tools but they can be used to harm.

How we use AI matters. So does understanding something about what sits behind it: the information it learns from, whose experiences are represented, whose are missing, and what assumptions might already be embedded there.

For me, that means using AI with knowledge and intentionality. Not accepting everything it produces simply because it sounds convincing, but questioning it, challenging it and ultimately taking responsibility for what I choose to use.

Because AI learns from enormous amounts of human-created information. And that information comes from a world where racism, ableism, sexism and other inequalities already exist.

So what happens when we teach machines using the records of an unequal society?

We shouldn’t be surprised when some of those inequalities come back out.

This was something that really stayed with me after hearing Dr Abeba Birhane speak at an AI event organised by Greater Manchester Coalition of Disabled People.

Her work on AI accountability made me think much more deeply about the systems and power structures surrounding AI, and how existing inequalities can be amplified rather than somehow disappearing when technology becomes involved.

It led me to explore the work of Dr Joy Buolamwini and the Algorithmic Justice League too.

Her experience is a striking example of why representation matters. Facial analysis software struggled to detect her dark-skinned face, while detecting a white mask she wore. She went on to research commercial facial-analysis systems, and her Gender Shades work with Dr Timnit Gebru demonstrated significant disparities according to gender and skin type.

That’s not some hypothetical future danger.

It’s an example of inequality becoming encoded into technology.

And it made me think about disability.

If disabled people are missing or poorly represented in the information used to develop and test these systems, our lives can become the exceptions the technology doesn’t understand. This is exactly what Dr Birhane was warning us about.

But there’s another kind of inequality too.

Who actually gets to benefit from AI?

That question has been on my mind particularly since Reesha Armstead
shared an article about autistic people and people with learning disabilities potentially being left behind as AI develops.

For some disabled people, including people with communication difficulties, AI might increase independence or make something that previously took hours considerably easier.

But those benefits aren’t automatically distributed equally.

Someone may not be able to afford the technology. They may not have reliable internet access. The training may not be accessible to them. Their language, disability or way of communicating may never have been considered when the system was designed.

And this is where intersectionality really matters.

Nobody experiences the world as simply disabled, or Black, or female, or poor. People can experience several forms of marginalisation at the same time, and the barriers created by those inequalities can interact and compound one another.

Technology doesn’t make those structures disappear.

Sometimes it can reproduce them.

That’s why I don’t think the answer is simply to embrace AI uncritically.

But I don’t think excluding disabled people from it is the answer either.

We need disabled people, Black and Brown people and other marginalised communities involved in designing, testing, challenging and regulating these systems from the beginning.

Because if we’re not in the room when the technology is being shaped, we shouldn’t be surprised when it doesn’t understand our lives.

AI can remove barriers.

It can also reproduce them.

Holding both truths matters.



Image description: A square Changing the Narrative graphic on a white background using black, blue and orange text. At the top it reads: “Communication in the Age of AI. Part 3/5 – AI Isn’t Neutral.” The main message reads: “AI can remove barriers. It can also reproduce them.” “Remove” is highlighted in blue and “reproduce” in orange. At the bottom it says: “Holding both truths matters.” Changing the Narrative branding and a small Trexy character appear at the top.

08/08/2026

07/08/2026

Communication in the Age of AI

Part 2/5 – Whose Words Are They?

One of the comments I expected when I started talking about using AI was:

“They’re not your words.”

I’ve been thinking about that.

Because on the surface it sounds obvious. If AI helped shape a sentence, surely the words don’t really belong to the person anymore?

But then I started wondering where we draw the line.

If I ask a friend to read an important letter before I send it, are they still my words? If I ask a colleague to check my report about a child, If an editor suggests a better way of expressing something, whose words are they then? What about predictive text, spellcheck or speech recognition?

At what point do we decide that a communication tool has become the author?

As a Speech and Language Therapist, I’ve spent years helping people find ways to communicate. Sometimes that meant changing the words, sometimes changing the environment, and sometimes changing the expectations of the person listening.

The goal was never perfect speech. It was meaningful communication.

That’s why I find the discussion about AI so interesting.

When I use it, it doesn’t know what happened to me. It doesn’t know my values, my experiences or my relationships unless I choose to provide that context. And even then, it only knows what I’ve told it.

Quite often, it challenges me.

It might point out that I’m making an assumption, ask whether there could be another explanation, or help me recognise when emotion has filled in gaps that the evidence can’t.

Sometimes I disagree with it completely.

Because the thinking still has to be mine, and so does the responsibility.

Every sentence has to pass one test before I use it:

Does this genuinely reflect what I think, believe and experience?

If it doesn’t, I change it, delete it or ignore it.

AI can suggest different words.

Only I can decide which ones become mine.

Perhaps authenticity isn’t about whether we used a tool to help us communicate.

Perhaps it’s about whether we’re prepared to take ownership of what we ultimately choose to say.

And maybe, before dismissing someone’s words because of the tool they used, we should ask whether we’ve actually listened to what they were trying to tell us. 💙🧡



Look out for part 3 next week.

Image description:
A square Changing the Narrative graphic on a white background using black, orange and blue text. At the top it reads “Communication in the Age of AI. Part 2/5 – Whose Words Are They?” The main message reads: “AI can suggest the words. You still own the voice.” Underneath, smaller text reads: “The tool can help shape the message. You decide what you mean.” A small Trexy character and Changing the Narrative branding appear at the bottom.

06/08/2026

This is how inclusion should be done.

I’ve been incredibly lucky to meet some genuinely supportive Pilates instructors: Nisha, Laura Pettener, and especially Vic Considine, who even attended a Thalidomider seminar to better understand our needs and help develop safe, accessible Pilates exercises.

I don’t generally enjoy exercise. It can make me anxious, and the after-effects are often unpleasant and sometimes very painful. Even so, I’m so grateful to all three of them for encouraging me to keep going with my 1:1 Pilates sessions.

Looking back, I know I would have struggled so much more with things like bridging, rolling, and transferring from bed to chair after my ankle injury if I hadn’t already been building those skills. It’s a reminder that accessible exercise isn’t just about fitness. It can make a real difference when life suddenly changes.

05/08/2026

More evidence of why political donations need much tighter rules. Accepting hundreds of thousands of pounds from someone with commercial interests in healthcare (recruitment for the NHS and private healthcare providers) creates an obvious conflict of interest, or at the very least the appearance of one.

At the same time, proposals to reduce disability benefits risk pushing more people into poverty, and poverty is consistently associated with poorer health.

That’s hardly the direction the NHS or the country needs.

Is it time to ban all private political donations?

Whether they’re from healthcare investors, billionaires, trade unions or corporations, large donations create real or perceived conflicts of interest.

Democracy should be decided by voters, not by whoever can write the biggest cheque.

Legality and integrity aren’t the same thing. The question isn’t just, “Did they break the rules?” It’s also, “Did they uphold the public trust?” Those are very different standards!

Newly appointed Health secretary Yvette Cooper has accepted almost £500,000 in the last decade from Peter Hearn, who held a significant stake in NHS & private healthcare recruitment firm Odgers.

Odgers was involved in the appointment of former TalkTalk executive Dido Harding as the lead for the heavily criticised NHS Test and Trace programme during the covid-19 pandemic.

A quick search found many other senior NHS jobs for which Odgers was listed as the “recruitment partner.”

Despite doctors’ concern over Cooper’s acceptance of these donations, she has declined to rule out taking further donations from donors or companies with commercial ties to the NHS or those linked to private healthcare while in post as health secretary.

Between 2015 and 2024 the MP, who previously served as foreign secretary and home secretary, took £248 619 from Hearn directly as well as £220 623 from two of his companies, MPM Connect5 and OPD Group.

Neither of these companies have a public website but both are registered with Companies House at the address of an accountancy firm in Hertfordshire. In 2023.

Source: The British Medical Association

😭
🔥🌹
😤
⬅️

04/08/2026

How scary this must be!

03/08/2026

Part 1/5 – Accessibility Isn’t Just About Ramps

As a Speech and Language Therapist, communication has always fascinated me.

Not just speech, but communication.

How we express ourselves. How we’re understood.

And sometimes… how we aren’t.

When people talk about accessibility, they usually think about getting into a building.

That’s important.

But what if the biggest barrier isn’t the doorway?

What if it’s being heard?

I’ve been thinking a lot about AI recently.

Not because I think it’s the future of everything.

Not because I think it has all the answers.

Quite the opposite.

It’s raised questions I never expected to be asking.

For some people, AI is becoming another accessibility tool for communication.

That doesn’t mean it replaces their voice.

It can’t.

It only works with what the person chooses to share, and every word still has to feel authentic before it’s ever used.

Yet I’ve noticed something interesting.

As soon as AI enters the conversation, many people stop asking:

“What is this person trying to say?”

and start asking:

“Did AI help them say it?”

The method suddenly becomes more important than the message.

That makes me pause.

Because communication has never been about everyone speaking in exactly the same way.

It’s about helping people find the way that works for them.

Maybe accessibility isn’t about making everyone communicate the same way.

Maybe it’s about making sure everyone has the opportunity to be heard.

What do you think?

Can AI ever be an accessibility tool for communication, or does it create more barriers than it removes?

This is the first post in a series exploring communication, disability, accessibility, ethics and AI. In later posts I’ll also look at authenticity, systemic bias, environmental impact and what ethical AI use might look like.



Image description (alt text):
A clean white CTN graphic titled “Communication in the Age of AI – Part 1: Accessibility Isn’t Just About Ramps.” Five diverse characters, including a wheelchair user, a person signing, Trexy the dinosaur mascot, and others stand side by side beneath identical empty speech bubbles, symbolising that different ways of communicating deserve equal value and equal opportunity to be heard. Orange and blue CTN branding appears throughout.