Experts are touting the important role AI can play in shedding light on women’s health issues, as studies warn of bias in the algorithms behind the technology.
While the data input into analytical systems that use AI remains a key concern in avoiding gender bias in health care, the biases that AI systems may reveal about the underlying data could also prove beneficial, speakers on a panel on addressing gender inequalities in health care said.
Discover effective B2B marketing
We combine business intelligence and editorial excellence to reach passionate professionals across 36 major media platforms.
learn more
Dr. Antonella Santuzzione Chada, CEO of the Swiss nonprofit Women’s Brain Foundation, highlighted a recent paper that revealed underlying biases within AI-driven systems. Mr. Chadha was speaking at a panel session at the Medica Congress, held in Düsseldorf, Germany, from November 17th to 20th.
This paper demonstrated that Apple’s automatic speech recognition (ASR) system, Siri, exhibits gender bias and limitations in providing guidance regarding women-specific health concerns. For example, if you requested guidance about menstrual pain, Siri was unable to provide meaningful guidance based on the data entered through the underlying deep learning model.
But in some ways, Chadha sees AI’s current shortcomings as beneficial. “AI is the reason we are becoming more aware of bias. Without AI, we would not be as aware of bias as we are today,” she explained during the panel discussion.
AI is pervasive in the digitalization of all industries and continues to evolve in the healthcare sector. GlobalData analysis predicts that the valuation of AI in healthcare will reach $19 billion by 2027.
Ambient AI tools for transcribing patient visits, such as Microsoft’s recently launched Dragon Copilot and AI triage systems, are all aimed at assisting doctors and optimizing their workflow.
However, research highlights how biases in the algorithms driving AI systems can reinforce discriminatory conclusions about women’s health and lead to incorrect care decisions.
For example, a recent study by the Center for Care Policy & Evaluation (CPEC) at the London School of Economics (LSE) found that Google’s ‘Gemma’ AI model downplayed women’s physical and mental problems compared to men when used to create case note summaries for social workers.
Chadha added that knowing what biases exist in data analysis can help assess what gender-specific medical issues need more focus and sensitivity.
Professor Helen Krimlisk, Associate Medical Director for Innovation, Research and Development at Sheffield Health and Social Care NHS Foundation Trust, who also took part in the panel discussion, highlighted that data collected and synthesized by AI tools determines how biases can be introduced into clinical practice, even inadvertently.
However, Krimlisk acknowledged that collecting diversity data, ranging from gender to other indicators such as ethnicity, can be a challenge, and that what data should be included and how to categorize it for useful application in healthcare are issues that require further consideration.
She said: “People may feel uncomfortable disclosing their diversity data.
“We want to get the maximum amount of data in that respect, but we have to recognize that for practical reasons we need to limit precisely the amount of data that we collect, perhaps in a structured way.”
