To think AI should be used to diagnose autism & ADHD

Do you mean AI? How do you envision that working exactly?

Annual leave?

I’m sorry you’re struggling by the way. What sort of adjustments do you need?

I think it would end up being over-diagnosed

Don't be so ridiculous

This may happen in the future but we're nowhere near this yet.

Russell Barkley and some others are on a mission to get ADHD recognised as a public health issue so that it can be diagnosed and managed by GPs in much the same way that GPs can now prescribe antidepressants and anxiety medication.

I think this would be great as would really help the waiting lists. Complex cases could still be referred on.

Sounds like it might be possible in the future OP.
https://www.verywellhealth.com/artificial-intelligence-to-treat-autism-4706533

We’ll get there in the next few years.

Diagnoses are already based on scores on complicated questionnaires so AI could do most of the process already, with human oversight/checking.

There are massive waiting lists for diagnosis already, and a shortage of suitably qualified experts. So there’s a lot of money to be made - I’d be shocked if nobody is already working on this.

You don't need a diagnosis to get reasonable adjustments

It'll happen eventually I'm sure, but it's astounding how much people don't value the knowledge and skills doctors have and think they just sit and tick boxes and that a machine could do the same.

That is exactly why machines are not there yet. Because if it was a tick box exercise then you're right, machines could do it already. This is exactly how the private clinics screen for whether or not you should proceed to diagnosis, and it's pretty good in terms of accuracy - over 90%.

For more depth, there are already computer tests that you can do for ADHD, testing things such as working memory, reaction times, split attention and so on, I had a doctor who was very keen on me doing them and I did them about 3-4 times on different medications which he claimed was the gold standard way to test the medication. However, I did not agree (based on my experience at that time and also later, titrating medication in the usual way), and according to experts, these tests are very unreliable in terms of predicting the level of impairment in the real world, which is also what I found. Apparently the tests are more likely to be testing something similar to IQ - great for data nerds, totally useless on a day to day scale unless you are comparing two people with vastly different IQs. You'll see a difference in functioning between someone with IQ 70 vs IQ 90, but not very much difference at all between someone with an IQ of 90 vs 100, and even if you have an IQ of 130+ (classed as gifted) this is no use in the real world if you have severe social anxiety and cannot cope with normal social interaction.

I do think that in the future (not yet) bots similar to Chat GPT but much more advanced in language analysis would be able to interpret different patterns in how people talk about their struggles in the case of different issues whuch can look similar or cause similar issues e.g. ADHD, ASD, OCD, depression, anxiety, etc. And there is potential for machines to become even better than humans at this - humans tend to get a bit stuck in "hammer and nail" thinking where we are quick to recognise things which mirror our own experience, expertise, etc. So if you have an ADHD specialist they may be predisposed to see ADHD. Being able to "speak" to a machine would in essence be like having a whole room of experts on different conditions who can all see the signs of their own condition and come to a collaborative conclusion, which would be absolutely incredible medical care but not physically possible using humans - there are too many patients and not enough specialists (or time!) for this to be realistic. But as said, this is far away. Don't think of the current language models when you think about this.

You could then have one more general clinician who can help the person understand what the machine is saying and guide the person in deciding on therapy, treatments, management etc.

Mental health is currently so incredibly undersupported/poorly supported that it would be brilliant if there was a way to make mental health support available on demand to every single person at all times.

Just to echo a pp, you really don't need a diagnosis for reasonable adjustments to get a job and for work or school. I'm dyslexic and have a range of reasonable adjustments at work and in job interviews and noone has ever asked to see 'proof'. But yes AI may be able to diagnose it one day, just like most conditions. There's still a lot of benefit from a human doing it at the moment. Its like the difference between googling your symptoms and get a huge list of possible conditions and asking a doctor who will test for the most plausible conditions first rather than leap to something rare and obscure.

I don't see how they could, because it isn't about how someone ticks boxes but what impact it has on their functioning.

An example is that DD2 said that she would 'go and live with husky dogs' when she is an adult because 'they don't have a big list of demands like people do'. Lots of people might say that they find animals less complicated than people, but the difference is that for DD2 she was saying that people are overwhelming and demanding, and that she can't cope with it. I don't think that AI could pick up that nuance.

Yes I agree! And it's quite possible - the main chunk of any diagnostic appointment with a mental health professional is completing questionnaires which form part of the diagnosis. For this to be done via AI and a report based n the results then sent to clinician will free up their time to enable them to spend more time with other patients, and thus reduce waiting lists. It's best done in conjunction with clinician for diagnosis though, not AI alone.

It's coming, I promise. Baby steps a little, especially in the NHS, but is coming. It's being used for depression and anxiety in services at the moment.

Why is it ridiculous? Maybe not in the next year or two, but eventually.

😂😂😂😂😂😂

I don't think that AI could pick up that nuance.

At the moment, no, it can't. But there's no reason that it couldn't in the future. It would probably be better at this kind of thing than humans are, because we tend to ascribe our own meaning/stories to things whereas if an AI model has input of hundreds of thousands of people with autism, ADHD, depression, BPD, as well as neurotypical people then it can likely pick out patterns and nuances better than humans can, and likely in different ways.

In the time frame we are talking about then it might not even be doing this via language but more transmitting thoughts/memories/experiences directly, that is something that is being looked into (but a long way off)

Because for me to have an experience, then put that into words with my cultural bias and my life experience and my understanding of the concepts and words, I then say them to you and you hear and interpret them based on your cultural bias, your life experience, your understanding of concepts and words we will each think that we have communicated but have a subtly different understanding. And as we go through life our understanding will change too so I could describe something that happened to me at 18, and have a completely different perspective on it now compared to what I thought my experience was back then.

If in the future we can communicate directly with the feeling/experience without putting it through the very lossy filter of words, that may be something that could be used in diagnosing mental health conditions.

Why the quotation marks?

I hovered over the voting a bit.

I don’t think this is a job AI could do - drs have to be specialist to diagnose and even gps can’t do it - so not sure how AI could. If it was a case of just taking a quiz surely they’d do that now?

I do think however something needs to be done as the long waiting times are affecting children’s life chances. If you’re fortunate enough to pay privately you get a huge advantage for your child. If parents are pushy enough to make sure things are looked at you get an advantage.

I think all children should be screened for it as it’s well known that girls in particular mask, and aren’t picked up.

Unless you’ve got more to say, I find this shockingly ignorant.

Children’s neurodiversity actually being picked up so that their life changes can better match that of their peers? Oh no that would be awful!

Autism (ASC) cannot be diagnosed through simple questioning - when I ask a question I am then asking extended questions off the back of the first question

Very apt user name. Maybe you should take your own advice..

I would think genetic testing could be more used.
If we tested everyone on the waitlist surely we would identify most of the genes
Also it will become easier in next gen as more parents diagnosed.
Tbh i would prioritise siblings for assessments
And surely with such long wait lists non speaking etc could assumed?
The criteria the SW used to prioritise was more like 'in care' or 'in trouble'
Generally nothing which showed more signs of asd except speech.
2yr plus to wait and we havent even had any sort of screening its all going to be 2yrs.
Surely say sound sensitivity over x age could get a 'working diagnosis'
Unfortunately schools are not helping as they dont know what they are looking for. I raised adhd to school nurses when dd was 5 and they couldnt have cared less.it honestly put me off trying for a referral. And yet we had been having huge issues.