Machine whispers or human connection?

Many people self-diagnose, and to me, that makes a lot of sense. We might be sitting there, sleepless and in despair after weeks or months of agony, just needing some certainty that we’re not “going mad.” We humans love knowing what is happening, and we all need a certain amount of control in our lives. We might not feel able to share our pain with anyone around us because of shame, fear of the consequences, or just not being able to find the words to express.

It’s then that we might turn to the internet and, more specifically, to the newest tool we have at our disposal, AI (artificial intelligence or machine learning). Before AI began its incredibly rapid spread into our everyday lives, it would have been social media that we turned to at 3 am.

We may have read about or heard about a set of symptoms in an appealing, engaging video by another human being. What they’re saying resonates with us, and it feels helpful to hear how others have dealt with what we’re going through. Sometimes, we can gather tools, resources, and strategies and learn from the kind of results that other people have seen. This is human collaboration. Not without its issues and flaws, of course, but another human is behind it who can explain their thinking and who we can - and should - question further.

Diagnosis is a historically tricky topic. In my practice as an integrative psychotherapist, I do not and can not diagnose my clients or anyone else, not for anxiety, depression, burnout or neurodiversity. What I can offer are explanations of the pathway for diagnosis and, with confidence, explain what someone might expect from meeting a psychiatrist and the treatment options they may be offered. This is based on a lifetime of experience in mental health. I started working with people transitioning from acute NHS care back into the community at the age of 17 (that’s a whole other story in itself), through my own experiences with my mental health and my soul as well as my direct client experience as a therapist working with other professionals and walking alongside my clients.

What I am not is a robot; I am not a machine. I am not an AI interface, app or bot. When a client comes to me and says, “I think I’ve got X diagnosis”, I ask them why they think that, and we explore that together. More and more clients are coming to me this year saying they have put their symptoms into AI and it has “told” them they’ve got X, Y and Z. Some of this can be helpful; it can give a foundation to work from. They’re bringing self-awareness to their situation and reaching out for answers; I want to work with that and harness it.

What isn’t helpful is when AI gives someone a misleading diagnosis, and they do not know the difference. All the disclaimers that state, “This is not a formal diagnosis and is no replacement for seeing a mental health professional” can be skimmed over when we’re desperately searching for answers at 3 am. What stands out and sticks is the label, and some of these labels are at best helpful and at worst misleading, and at very worst, dangerous. 

If a client chooses to identify themselves as having a diagnosis of, for example, personality disorder, they may be unwittingly walking themselves down a path that reduces their options. Personality disorder can be, rightly or wrongly, seen as long-term, possibly 'hard-to-treat' and requiring intensive resources for which there are long NHS waiting lists. Once someone decides that an AI diagnosis makes sense to them, unless they can explore the meaning of that and its implications with a well-informed human who can understand the subtleties and nuances through their own experience of the systems we work within, they may be consigning themselves to a difficult road.

There are myriad examples of where AI gets it very wrong. This is where I urge caution. The Alan Turing Institute explains some ways AI and machine learning are being employed to evaluate huge swathes of data and identify potential early diagnoses and individuals within demographics for early intervention – with dementia, for example. The work of these researchers will be conducted under rigorous conditions and will be filtered through human brains and peer-reviewed for validity and reliability. None of this takes place when the connection happens directly between a person and an AI interface. Without a warm, accepting human being to ally with and be gently challenged by, an individual with an AI diagnosis can be left out in the cold, alone and facing a machine generated prognosis that doesn’t account for their uniqueness, the factors that they aren’t aware of and therefore cannot plug into the AI for review as part of their set of symptoms, experiences and behaviours.

There are many benefits to using AI, it can proof read and make suggestions for our written work, it can collate large amounts of data rapidly but, as many people more qualified than I have already pointed out, they come with considerable cost; not least of which is the soul-level connection between two human beings living human experiences in all of their difficult, wonderful, conscious and unconscious messiness and beauty.

AI itself urges caution. When asked recently by a curious thinker how it would go about taking over the world, the machine explained it would start slowly, advising on recipes and simple tasks at first, spreading its influence into all areas of life until it would reach a point of omnipresence so subtle and pervasive that “you won’t even realise anymore that your thoughts are all AI whispering into your ear.” Let’s bear that in mind as the power of machine learning accelerates every week.

I believe that however helpful we find condensed access to trillions of megabytes of information, we will always need to meet another living, breathing heart and soul in the therapy space.