At the University of California Irvine Sue & Bill Gross School of Nursing, educators are developing ways artificial intelligence can contribute to better patient care and outcomes.
Amir Rhamani is a professor at the school of nursing – as well as professor of computer science at the Donald Bren School of Information and Computer Sciences, professor of electrical engineering at the Henry Samueli School of Engineering, and associate director of the Institute for Future Health.
His latest project is openCHA, a conversational health agent with a personalized framework based on large language models. He is developing it in collaboration with Mayyar Abbasian, Iman Aximi and Ramesh Jain, all from the UCI School of Information and Computer Sciences.
We interviewed Rhamani to learn more about openCHA – discussing LLMs’ need for expanded capabilities; how developers can integrate external data sources, knowledge bases and analysis models into their systems using openCHA; and the next steps to get openCHA into the healthcare market.
Q. Describe in general terms openCHA and your goals for the technology.
A. In healthcare, the abundance of misinformation can leave people feeling lost and insecure. Sorting through the sea of conflicting information online isn’t easy, and without the right guidance you can easily fall prey to inaccurate advice.
Language barriers further exacerbate this problem, making it difficult for individuals to access the help they need. But amid this confusion lies a possible solution: personalized health data. By utilizing biomarkers, genomics, images and other personal information can help individuals gain a clearer understanding of their health and make more informed decisions. This is where recent developments in AI come into play.
With the ability to analyze massive amounts of data and make personalized recommendations, AI is becoming an invaluable tool in navigating the complex healthcare landscape. However, the accessibility of these AI-driven solutions remains a challenge, akin to finding a needle in a haystack.
Enter the Age of Great Language Models, which is poised to revolutionize the way we access and interact with healthcare information, and will provide a beacon of hope in an otherwise dark sea of disinformation.
In recent years, LLM-based conversation systems have shaken things up. These systems are like the cool kids on the block: they give us access to tons of text information and provide conversations that are actually meaningful.
But the thing is, when it comes to managing health, we need more than just your run-of-the-mill LLM. We’re talking about conversational health agents (CHAs) – the superheroes of the health world. These guys need to be able to walk the talk, adapt to your ever-changing health needs, and analyze your personal data like a professional.
Guess what? They are powered by trusted LLMs, who ensure they understand you and can provide you with the personalized support you need, whether it’s answering your burning health questions or simply providing an empathetic ear.
Now let’s talk about openCHA. Right now, we are on the brink of creating frameworks that can disseminate information in the most friendly and culturally sensitive way possible. That’s where openCHA comes in – it’s a kind of toolkit for developers who want to build CHAs.
Our goal? To ensure that CHAs can truly connect with users and give them personalized, caring answers to their health questions. With openCHA we’re talking about enabling the integration of all kinds of data sources, knowledge bases, and analytical models to completely innovate the way CHAs interact with people.
This framework is a game-changer, giving CHAs the brains and tools they need to provide up-to-date health advice tailored just for you. Say hello to a whole new level of healthcare companionship – openCHA is here to ensure you get the information you need, when you need it.
Q. You say LLMs need broader capabilities, including critical thinking, knowledge acquisition and problem-solving skills. How do you take this into account in openCHA’s LLM-supported framework?
A. I’d like to introduce you to the orchestrator, the cornerstone of our framework, designed to mimic human behavior within the healthcare process. At its core, this orchestrator consists of two LLMs and one performer.
One LLM acts as a planner and coordinates with the executor to gather essential information and perform the necessary analyses. Using established prompting techniques, this primary LLM navigates the planning and problem-solving process and provides transparent reasoning behind the responses and decisions.
Within the openCHA framework, this capability allows user queries to be broken down into manageable sub-problems, facilitating the execution of tasks necessary to collect relevant information. Once all relevant data has been collected, the second LLM takes charge and uses the collected information to provide users with reliable answers.
This structured approach has proven to be a comprehensive and reliable response to user questions, increasing trust in the openCHA system.
Q. How can developers integrate external data sources, knowledge bases, and analytics models into their systems using openCHA?
A. We’ve rolled out an open-source codebase that gives developers all the tools they need to seamlessly integrate existing datasets, knowledge bases, and analytics models into CHAs.
We’ve kept the code flexible and modular, making it very easy to add new external resources with just a few lines of code. The orchestrator does the heavy lifting and manages the logic effortlessly, so developers can focus on what they do best.
Q. What are your next steps in bringing openCHA into the healthcare market?
A. Currently, openCHA is an open source solution to build a community. But with the vision of enabler technology to enable applications. Prototyping. It can be vague.
Our mission is to foster a thriving community around openCHA, driving innovation within the domain of CHAs. Our focus is on establishing an open architecture for openCHA, forging connections with other open health technologies, gaining access to open content resources, and shaping future standards for CHAs.
We are passionate about raising awareness about the future of healthcare the central role of LLMs. By integrating execution and planning methodologies, our goal is to deliver best-in-class health solutions that meet the ever-changing needs of users.
Our ultimate vision is to cultivate a collaborative environment where stakeholders can freely exchange ideas, share expertise, and collectively advance the field of conversational health technology.
Through collaborative efforts and shared insights, we are committed to guiding the development of CHAs toward greater effectiveness and relevance in addressing healthcare challenges.
Follow Bill’s HIT coverage on LinkedIn: Bill Siwicki
Email him: bsiwicki@himss.org
Healthcare IT News is a HIMSS Media publication.