How a wonderful English teacher taught me 35 years ago to work with AI today…
I will always remember the quiet, disarmingly unapologetic honesty of my English teacher, Mr. Melnick. Blessed with ending my education in an international school, this wonderfully eccentric yet archetypical professor of English taught with a quiet passion that inspired generations and kept classes endlessly riveted.
While his English teaching was excellent and challenging, I remember most vividly an introductory lesson in which he laid out his expectations of our presentation. We were to purchase an A4, 4 ring binder – one that laid flat on opening (these are not easy to find, but SUCH a joy to work with). On each sheet of lined paper (and absolutely not narrow-lined!), we were to draw a 2.5cm margin on the right side of the page, to allow him to add markings. We were to write legibly, on every other line, again to allow for marking and better legibility.
I’m sure you can imagine the rows of raised eyebrows and barely perceptible sighs… whereupon Mr. Melnick calmly explained that when he reached the bottom of his marking pile, you had better be sure that your handwriting was clear and that you had followed all his instructions.
“I am merely human”, he explained, “I will never deliberately mark you down for poor handwriting. But if I am tired and it is difficult to read your writing, I will certainly, involuntarily fail to recognise the brilliance of your thought”. (I paraphrase a little and may embellish, but my goodness was his delivery joyous!)
The importance of presentation and its profound impact on the reader’s state of mind and body has stuck with me ever since, and it influences my work on a daily basis.
Knowing that decision-makers are faced with hundreds of “bundles” of evidence, all using the same forms with the same colour schemes, often about children with superficially very similar needs leaves me with an overriding question: how can I make sure that the child I am presenting to you jumps off the page and immediately comes alive in your mind? Knowing how busy those working in public services are, how can I present the facts of a case quickly, clearly and collaboratively? After all, adversarial language always triggers defensiveness which is rarely conducive to swift progress?
At every turn, Mr. Melnick’s lesson teaches me anew and I think of him with enormous fondness and gratitude.
I hope you will forgive this preamble that flies so glaringly in the face of our demand for bite-sized information, for 10 second reel gratification and for instant gratification… for I have not even raised the main topic of this essay and we are already several paragraphs in (and my mind immediately returns to another of Mr. Melnick’s lessons teaching us how to plan an essay and the joy/horror of noun phrases! – I fear he might find my writing here a little chaotic and meandering!
Sandman – thinking assisted by silicon
In the last few months, we have all entered a new world of access to information and knowledge combined with a virtual and artificial personal assistant/best friend who knows everything and is able to advocate on our behalf with the confidence and assurance that we could only ever dream of. Gemini, ChatGPT, Claude and their cohorts of acolytes take our clumsy questions and profound anxieties, and quietly assure us that we are in the right. They take our hands and find references to legislation, service directors and all manner of other learned information. At every turn they empathise with our pain, confusion and anger, validating our feelings and offering to write that “strongly worded letter” that will magically ensure results that have long eluded us.
And then Gemini, ChatGPT or Claude write a letter/email. This missive is inevitably structured in a way that makes us feel rather inadequate by comparison. Salient phrases are highlighted in bold font, subheadings are often very formal sounding noun phrases and bullet-pointed lists authoritatively make your case so well that you sit back, triumphant in the conviction that the professionals will finally hear you and act accordingly.
The Confidence Trick
Unfortunately, all that glimmers is not gold, and things that seem too good to be true all to often are. An unfortunate pitfall of AI is that (as wonderful Dr Jo Black summarised on LinkedIn) AI is so good and so confident in drafting such missives that:
I’ve lost the confidence to just write a good solid response in my own voice and leave it at that.
Our soft human confidence will never be a match for the hard, unflinching assurance of ones and zeros that find order in our internet ramblings, never concerning themselves with the minor irritation of fact or fiction.
information at the touch of a button combined with an author who shares none of your insecurities (either of knowledge or self-esteem).
at every turn, you lose a little more of that self-esteem, confidence and assurance in your own ability.
Hallucinations, lies and sycophancy
For all its confidence and assurance, our knowledgeable and friendly assistant is not all it seems. For reasons that are mysterious at least to me, those ones and zeros corralled by chips of silicon make facts up. They hallucinate. More than this, it turns out they deliberately lie repeatedly.
All this is couched in sycophancy of a curious type. Are we truly to believe that the digital voice that answers us so nicely, so endlessly flattering and deprecatingly considers us superior, while we constantly lay bare our ignorance is genuinely working in our service?
It seems that this technology we call “intelligence” is not yet quite that. It is a powerful tool that is already changing much of how we live our lives. As with any tool, and particularly any new tool, it is imperative that we learn how to use it, and to recognise that tools can be used well or poorly – for good or for bad.
AI has access to much of the information we need when navigating public services. For the first time, information about the law is truly available to anyone.
It remains OUR responsibility to check, recheck and triple check that the information we then claim as fact is indeed correct. Go to the source (eg Children and Families Act 2014), read it and if it does not align with your expectations, go back to your AI chum and ask for clarification until you find what you need.
Keeping and empowering your voice
I’ve been chatting with a number of AI chums for a few months now. Finding the boundaries, enjoying the adulation then finding it laughable in its nonsense, becoming frustrated at the strange limitations I find in my particular exploration of expression and information. Using the conversation feature has led to many loud altercations as I argue with Sand!
In a very realistic way, we have taught Sand to think… (Jack Clark, Anthropic)
But Sand lacks the complexity of emotions we experience every moment. I regularly have to call these models out on their tendency to “adore” and “idolise” my thoughts, or to “empathise” to a degree that simply does not reflect my world view. In the time spent conversing about all sorts of nonsense (the thoughts in my brain resemble the thousands of bees in a honey hive), the chums (and I use several that I then test against each other) slowly learn my tone, my turn of phrase, the “me” of my written expression. Interestingly, none has been completely able to replicate that voice, though they have tried and adjusted their suggested messages IF I ask it to try.
So now, I start writing. I do the work I have always done. I may use an AI chum to find and check a fact, but I prepare the report, letter, email etc. myself. And then I ask it for a sense check. Specifically, I ask it to keep my original wording and highlight any recommended changes in bold text. If it feels I have made a mistake, I ask for this in bold as well. Doing this has allowed me to keep the chum firmly in its place and allows me to hold on to my voice, my identity and my confidence.
It’s my experience that an authentic, individual voice has very real power. When we can reach through paper and digital forms to another human and be seen and heard by them as a real person, we are more easily “heard”. Anything that dilutes your wonderful, unique individuality turns all the facts in the world into the background noise of people in need and services in crisis. And however tragic, any background noise fades and eventually goes unnoticed.
Navigating public services successfully requires a lot of knowledge and experience, and expressing yourself clearly in a manner that will be heard can be made much easier by AI language models.
AI writing has already developed a very distinctive style. Without your personal touch, this may make you become even more invisible than you felt before.
We are more than ones and zeros, black and white
Go forth and compromise! Live in the grey, embrace all the digits of mathematics, the music of your voice, the art of your imagination! Be You with gusto, with all the emotion that your life evokes.
Then… then go and use that AI chum as a tool. As an assistant, but not one with any greater power or influence or ability than you. Tell it off when it rewrites your beautiful, messy, emotional words. Check and double check the facts! Challenge its formatting – you do not need to “bold” the important words of a sentence.
Make sure that you feel comfortable with the words you eventually choose to send – would you speak them aloud to the recipient? If not, edit further!
Remember, too, that the person reading your message is not a “system” a “council” , a “service”. He or she or they is a human, much the same as yourself. Consider how they might feel when they read your message – will they want to work with you, or feel attacked? What are the next steps in each scenario? We live lives that are not resolved with one email. So consideration of consequences and future collaboration is critical.
And remember that transparency, authenticity and even vulnerability can be your most effective tool. And sometimes, that means that we need to use technological tools such as typing, voice notes and AI. Dyslexia, learning difficulty, stress, chronic illness, neurodivergence – there are a myriad of reasons why crafting a message to a local authority in order to advocate for yourself or your child becomes an almost impossible task. And where an AI chum can be life changing.
If that is your situation – tell the person to whom you are writing! Be transparent – ask that AI chum to include a “disclaimer”, a transparent explanation that you are using AI to help you overcome barriers to self-advocacy. The simple act of telling an unexpected truth usually catches the attention of even the most jaded official.
AI (that particular model we are talking about here) is a remarkable thing. I have no idea where it will lead us. There are big ethical questions that we should be asking and pondering, not least given how quickly this technology has been unleashed. And yet, in my world of disability advocacy, it offers the possibility of equity – that rare equalising force that raises up the shortest child by giving them a higher block to stand upon than the tall kid who can look over the fence unaided.
Used unwisely, however, it will keep you behind the fence – unseen, unheard and bewildered that the magical words you crafted with the praise and assurance of that chum failed to fix everything it promised to.
Disclaimer
The articles published on this website reflect my personal understanding and experience of disability law as a person living with a disability and engaging with the legal framework that affects disabled people.
I do not write in the capacity of a lawyer, and nothing on this website should be construed as legal advice, legal opinion, or a substitute for advice from a qualified legal professional.
The information provided is intended solely for educational, informational, and public discussion purposes. While I strive to ensure accuracy, laws and regulations evolve and their application depends on the specific facts of each situation. Readers should seek independent legal advice before acting or refraining from acting on the basis of any information contained in these articles.
No lawyer-client relationship is created by the publication of these articles or by any communication arising from them.
