With NIH Grant, Developers Focus on Cultural Attunement of GenAI Tools
A company called mpathic has received a National Institutes of Health Small Business Innovation Research (SBIR) grant to fund the expansion of its generative AI tools to facilitate cultural attunement between mental health providers and patients, bridge the gap in mental healthcare for underserved racial and ethnic populations, and potentially improve therapeutic outcomes and patient satisfaction.
Racial and ethnic minorities report higher rates of mental illness compared to white and privileged communities, but are less likely to find or access high-quality care, and less likely to finish treatment. These same communities often feel that mental health clinicians lack “cultural attunement,” where a clinician is understanding and responsive to the intersection of societal context, culture, and power of the patient’s lived experience, and are ill-equipped to respond to their mental health needs in an empathetic and affirming manner. Cultural attunement has been shown to be the driving factor that retains racial and ethnic minorities in mental healthcare.
The company notes that many digital health innovations are intended to improve patient-clinician relationships and the overall patient experience, including AI-powered technologies such as chatbots and ambient AI assistants. But male-dominant hetero-white language is the internet’s most prevalent language and is the foundation for widely used health technology AI models.
The company says it is the first to build a conversational analytics platform to detect and correct for cultural attunement.
In this project, mpathic will employ AI and natural language processing on conversational data from 300 30-minute health coaching sessions from a provider organization called Wave to improve cultural attunement in real-time provider-patient interactions.
Wave delivers personalized virtual mental health care through a combination of an app-based digital experience and certified mental health coaches. Wave’s inclusive design processes and equitable hiring practices have supported an 85% non-white, non-male and non-straight team of therapists, coaches and care navigators, lending itself to an equally diverse user population and coaching session transcript data.
In a recent interview, Alison Cerezo, Ph.D., mpathic’s senior vice president of research & health equity, discussed the project. “We mapped out what it could look like to develop an AI company that really had health equity at its core and was very attuned to the ways that AI could be biased, and trying to really address that at the root cause,” she said.
To do this work, mpathic is using Wave’s provider-patient appointment transcripts to build its AI model, she added.
“It is often the case that as we talk about AI, people talk about AI bias, but people don't always have very robust or clear ways of being able to address that, so a large portion of my job is really devoted to data partnerships,” Cerezo said. “That’s what we're doing with Wave — being able to collect rigorous data along with metadata so that we can track how the AI models work differently for different groups of people. When we talk about using AI in the healthcare space, there's oftentimes a focus on patient data, but what's really critical is provider data if you're going to develop co-pilots that can support providers to deliver higher quality care free of bias. That's why data partnerships are critical. Because your models are being trained on provider inputs and behaviors.”
“On the provider level, we are dedicated to an inclusive hiring funnel, where we are making sure that we have traditionally underrepresented populations on our treatment teams,” said Sarah Adler, Psy.D., founder and CEO of Wave.
“As a potential future customer of this product that we are working together to build, I would say that is a little bit of a chicken-and-egg situation," Adler added. "They're partnering with us because they understand the robust clinical quality of our service delivery model; we will then ultimately want to partner with them because their data is actually built on a robust clinical model. So we are investing in building something that will eventually help us leap forward into the future in terms of using generative AI, which is very much within our roadmap, but ensuring that we're using a product like an mpathic product, which has the robust clinical validity that underlies the modeling.”
“There are ways that people are developing apps off the cuff, which can be dangerous and scary,” Cerezo said. “That is why we were incredibly careful to choose Wave as our data partner, because Sarah's built a really wonderful company. Sarah's a licensed clinical psychologist. I am as well. We understand on the clinical front the barriers that prevent providers from being able to deliver quality care, but also barriers of the system itself. So this is one of our ways of being able to integrate into systems to ensure that AI can do good by providers and good by patients.”
