Trusting AI with your mental health

Every time someone tells a counsellor they used ChatGTP for advice with their mental health, a little part of us dies. OK, I’m exaggerating but people using AI to try to figure something out or to help them cope in a mental health crisis seems to be increasing. At this point, 37% of adults in the UK have used AI in this way. More than a third!

While the immediacy (and accessibility) of AI ‘mental health support’ can be appealing, it comes with it some pretty big negatives that can be easy not to think about in the moment.

We know that AI is terrible for the environment. Datacentres use enormous amounts of water and fuel. At an already critical time, this technology, if it continues on the path it’s going, will greatly increase global warming with huge implications for our future on this planet.

So, like I say, a pretty big negative.

From a service perspective there are also a few things to consider.

An AI service like ChatGTP is designed to tell you what you want to hear to keep using it, potentially rewarding and validating thoughts and opinions to keep you using it.

If someone is vulnerable and not thinking clearly or doing things that are causing them harm, an AI ‘counsellor’ can easily end up validating and reinforcing unhelpful thoughts or behaviours. We’ve all heard the stories in the news about people trying to get clean being advised to take drugs or people struggling with self-harm encouraged to hurt themselves because the AI model is prioritising validation and engagement over care.

In contrast a human counsellor will recognise and challenge unhelpful behaviours and patterns of thinking instead of just telling you what they think you want to hear. The relationship between a counsellor and client is also by agreement time limited with clear boundaries. Reaching into a pocket for immediate validation might feel that you’re getting support, but is that actually helping you move forward or encouraging dependancy?

The therapeutic alliance that exists between a human counsellor and a client can be a pretty good predictor of positive outcomes and one that counsellors work hard to develop. That just doesn’t exist in an AI relationship and AI can’t form a genuine therapeutic alliance. While someone can -feel- connected with an AI, that relationship is entirely one way.

Qualified counsellors are members of professional bodies and bound by a strict code of ethics and behaviour. This helps ensure quality of service, safety and confidentiality.

ChatGTP gives you none of that. Data hallucinations and bad advice aside, we all saw in the headlines recently that whatever we put into ChatGTP is not private. Yikes! Anything that goes into ChatGTP is treated in law as the same as a google search, with no attorney-client or doctor-patient privilege. Those ‘private’ conversations about things that are upsetting you are essentially public data, can be used to help train updated models, and wheeled out in an un-redacted form when legally requested.

There is headway being made on more specific, limited software that is intended to be used alongside other sources of support, but right now a digital counsellor in your pocket doesn’t exist and there are some serious dangers in buying into the AI hype.

If you’d like to work with a human on your mental health, get the occasional movie recommendation and potential guest appearance from a cute cat, you can get in touch via the website or contact me directly on mckellarcbt@gmail.com