Could AI be the answer to widespread underdiagnosis of autism and ADHD in women?

On TikTok the hashtag #ADHDinwomen has 4.1 billion views and #autisminwomen has 337.2 million views. However, the recent rise in autism and ADHD diagnosis among women isn't a passing social media craze, but an appropriate response from a community that has been ignored by our health system.

For example, autism studies have been shown to exclude women and girls, leading to a gendered understanding of symptoms by doctors and biased diagnostic criteria, which in turn leads to underdiagnosis. Meanwhile, girls with ADHD often go undiagnosed until adulthood, and have to show more severe symptoms before they’re prescribed medication, compared to boys. 

The surge in social media interest is therefore a valid response from a community that has been underdiagnosed and misdiagnosed for years. It is a valid response from a community that has been socialised to “mask”, so they adhere to what is expected of them as “women”. It is a valid response from a community that is often excluded from scientific research and has received inadequate investment.

And this is just the tip of the iceberg of exclusion and discrimination faced by other marginalised neurodiverse groups. Similar sources of bias contribute to lower rates of autism and ADHD diagnoses among people of colour, adding an intersecting layer of discrimination and complexity for women of colour. Meanwhile, non-binary and gender diverse people also face being left out of data sets and neurodiversity research, despite being three to six times more likely to be autistic than cis people.

As a late-diagnosed autistic woman of colour, I believe this increase in awareness is overwhelmingly a good thing. This new level of knowledge neurodivergent people learn to understand themselves and the support they need to live happy and fulfilled lives, whilst neurotypical people have the opportunity to educate themselves on the issues facing the neurodivergent community.

Rising demand, limited access: navigating the neurodiversity service gap

However, although more people are now aware of neurodiversity, getting the necessary support remains a significant challenge. The growing awareness of autism and ADHD has led to an increased demand for diagnosis. But with NHS waiting times stretching up to seven years, many are forced to opt for expensive private care, wait for excessively long waiting periods, self-diagnose or receive no help at all.

This prolonged delay not only causes severe emotional distress but also underscores the pressing need for urgent and accessible healthcare solutions, especially in the context of mental health and suicide prevention which disproportionately affects the neurodivergent community.

To add to this, neurodiversity shows up in different ways in different people, as well as interacting with each person’s identity, personality and environment. This is particularly important to consider when building solutions for any underrepresented community. Gender, ethnicity and socioeconomics always need to be taken into account to provide effective and human-centred support. Combine this with the innate variation that characterises neurodiversity and it’s easy to see why the one-size-fits-all approach of current solutions is not working.

Responsible AI: diverse data creates inclusive support

Enter the age of artificial intelligence, which marks a new era of personalised support and tailored experiences for the neurodivergent community. Thanks to AI, a simple click can unlock a world of customised support based on a person’s goals, preferences, behaviours and data. This means that solutions can now be crafted to address specific challenges, cater to a wide range of unique needs and embrace diversity as a feature and not a bug.

However, as with any AI-powered solution, there are vast ethical considerations to be made to ensure AI is being used responsibly and fairly, with direct operational consequences. For example, an AI tool which aims to diagnose women, girls and people of colour, needs to be representative of their experiences, which historically have not been recognised or measured, and so will be missing from existing datasets. This means adding complexity to the AI development process, to make sure representative data is found, correctly labelled and added to the training set. Finding the data itself will be challenging and requires thinking imaginatively about how data is captured or what sources are used. This might mean not discriminating between neurodivergent individuals who self-ID and those with a formal diagnosis or incorporating non-traditional data sources like TikTok and Reddit.

Data is just one side of the responsible AI coin. If developed unthinkingly, the AI system itself may have human biases built into it and so great care must be taken during its creation. While no AI model is entirely bias-free, incorporating a wide range of viewpoints during its design helps anticipate and mitigate potential negative consequences. Involving individuals with firsthand experience of the problem is essential. Diversity within the team, spanning data science, engineering and product development, is also vital.

At COGS AI, the mental wellbeing startup for neurodiverse people that I co-founded, we have baked these methods into our team, design process and datasets. For example, we are currently creating the only dataset in the world that goes beyond diagnostic labels and categories when looking at neurodivergence, while our team of four all have either lived or learnt experience of neurodiversity and are ethnically and gender-diverse.

AI advantage in challenging times

In the midst of global economic uncertainty, startup funding has dwindled, causing investors to look for higher returns on investment. Fuelled by the rise of ChatGPT, startups are increasingly turning to AI as a means to make their products more enticing to funders. While AI may not hold the answer to all challenges and can be limited by the data it is trained on and the biases that may be built into it, its distinct strength lies in its capacity to personalise. When this personalisation is harnessed to foster positive social impact for a historically excluded and under-resourced community, the rewards are significant.

The surge in autism and ADHD diagnoses among women marks a profound societal shift, one that demands recognition and action. AI, if used responsibly, could emerge as a powerful ally in this journey, enabling personalised support that acknowledges the diversity of neurodivergent experiences.

Given the lack of research and knowledge concerning how neurodiversity impacts women, AI-powered tools with a low barrier to entry offer a beacon of hope. These AI solutions bypass unaffordable financial constraints, reduce the need for official diagnoses and sidestep seven year waiting lists. They also reject the stereotypes and misconceptions inherent in services that fail to target women specifically.

It's not just a technological leap; it's a leap towards inclusivity and empowerment. By harnessing the capabilities of AI, we can bridge the gender gap in diagnosis and support, creating a more equitable and understanding society for all.