Statistically, boys are four times as likely as girls to receive a diagnosis of autism spectrum disorder (ASD). But that’s not because boys are four times as likely to have ASD. According to Autism Speaks, a research and advocacy organization, many girls living with ASD simply “do not fit the stereotypical picture of autism seen in boys.”
This gender bias often leads parents and clinicians to miss signs of autism in young girls, resulting in later diagnosis and intervention. Researchers at The Rhode Island Consortium for Autism Research and Treatment (RI-CART) found that on average, girls are diagnosed with autism 1.5 years later than boys.
This is extremely problematic, as the earlier autism is diagnosed the earlier treatment can begin. The timing of treatment can have a profound lifelong impact on the child. Indeed, early intervention has been shown to enable over 75% of children living with autism to participate in mainstream education and up to 25% of high-functioning autism patients to progress beyond their original diagnosis altogether.
The good news is that the medical community can make great strides in overcoming this bias and fundamentally improving the standard of care and quality of life for all children and families living with autism by turning to increasingly valuable technologies for medicine: artificial intelligence and machine learning.
Behind the gender bias
What makes this a problem uniquely suited to AI? Simply put, autism’s gender bias is a data problem – and AI excels at deriving insights from massive datasets.
Anyone familiar with the ongoing conversation surrounding bias in AI itself is well-acquainted with the “garbage in, garbage out” mantra: Inputs of incorrect, incomplete, or otherwise biased data lead AI systems to produce incorrect, incomplete, and biased outcomes. The same is true outside the realm of AI.
Research and outcomes have long been skewed toward how ASD manifests itself in males, coming at the expense of females. This is in part because boys have accounted for a disproportionate share of those diagnosed with autism, girls have been underrepresented historically in research, and because ASD can manifest itself very differently in girls.
For example, studies have revealed that girls living with autism are more deft at picking up social cues than are boys living with autism. Girls also tend to exhibit fewer repetitive behaviors like hand flapping, and they utilize more “cognitive process” words like “think” and “know.” Girls diagnosed with autism tend to be more passive than boys at a significantly higher rate. Throw in the effects of gender-based expectations surrounding socialization and prevailing thoughts on how boys and girls behave and it all adds up to a diagnostic process that continues to ill-serve girls living with autism.
How AI can help
Since AI itself has been criticized for gender bias, might it not simply perpetuate the problem? Not if AI developers do what a growing number of researchers are doing: building gender-conscious datasets that control for the influence of human biases.
AI is able to assess thousands of human traits and features – including a variety of verbal and movement indicators – to identify the most predictive traits for current or future state autism. AI is also able to ensure that the validation sets include the features that have been commonly linked to females so that any child is being evaluated against “girl” features and “boy” features, which reduces bias.
AI systems are also able to integrate extensive amounts of external data on how ASD manifests in girls, building upon the findings and learnings from hundreds of thousands of other children to assist in delivering an early, accurate diagnosis.
Given the importance of early intervention to a child’s lifelong outcomes, the seemingly incremental improvements AI can enable can have a truly transformative impact.
What’s more, greater use of AI will translate into increased efficiency in the healthcare system. When an AI-based diagnostic empowers a pediatrician to quickly and accurately diagnose a child, most children will not need to be referred to a specialist. This approach will greatly relieve the burden on specialists, alleviating bottlenecks and allowing them to focus on harder-to-diagnose cases. This, in turn, will dramatically improve the standard of care for all children – boys and girls alike.
Toward a better future
As a spectrum disorder, autism affects no two people the same way. But researchers and advocates have convincingly shown that the wide gap in diagnoses between boys and girls isn’t simply a matter of boys being significantly more likely to be on the autism spectrum.
It’s imperative that we gain greater understanding of the gender bias in our system, with AI being the ultimate game-changer. By harnessing AI and Machine Learning tools we will bring about a more effective medical system when it comes to diagnosing and treating autism that leaves no child, regardless of their gender, behind.
Photo: MF3d, Getty Images
