An interdisciplinary team at the University of Illinois Chicago is creating a customizable virtual assistant to help young children and their families establish good oral hygiene habits through nonjudgmental conversations.
Mohan Zalake, PhD
Research Assistant Professor, Department of Biomedical and Health Information Sciences
UIC College of Applied Health Sciences
Nikita Soni, PhD
Assistant Professor, Department of Computer Sciences
UIC College of Engineering
David Avenetti, DDS, MSD, MPH
Associate Professor and Head, Department of Pediatric Dentistry
UIC College of Dentistry
Center for Clinical and Translational Science 2024 Pilot Projects
Learn how you can get involved in translational research at ccts.uic.edu.
The University of Illinois Chicago Center for Clinical and Translational Science is supported by the National Center for Advancing Translational Sciences, National Institutes of Health, through Grant UL1TR002003. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
:00 Dr. Avenetti:
Is it possible to create a virtual agent that families can connect with? We always want patients to feel like they connect with their providers, so one of the big goals of the project is to ask the caregivers, how do you feel interacting with these virtual agents? Do you feel like it resonates with you?
0:22 Voice Over (VO), Lauren:
Welcome to Collaborative Endeavors, a podcast about how experts from different areas of research come together to tackle big health challenges, leading to better therapies and healthier communities.
In this episode, we meet the team behind an innovative technology aimed at improving children’s tooth brushing habits. IVORY is an acronym for ImproVing Oral health using Relatable virtual agents for parents of Young children. The application is designed to support caregivers in ensuring that children ages three to six are developing and adhering to good oral hygiene habits, such as brushing twice daily with fluoride toothpaste, to avoid dental caries- commonly known as cavities.
Funded by a CCTS pilot grant, the multidisciplinary team behind IVORY spoke with me about their long and short-term goals, and how the code behind the virtual agent could be used to address other health behaviors outside the clinic setting. Project PI and research assistant professor in UIC’s department of biomedical and health information sciences, Dr. Mohan Zalake, started by describing how this project came to be.
1:27 Mohan Zalake:
From a scientific standpoint, the project emerged from the need to address the high prevalence of early childhood caries. We wanted to develop this application which is engaging to caregivers, that can effectively be delivered at home, that's beyond the clinical setting. We use virtual humans to simulate these face-to-face interactions to make this scalable and address the need for engaging in non-judgmental communication around oral health. But from a personal standpoint, there's a very interesting story behind this. This all started with Nikita and her mother, Doctor Preeti Soni, who is a dentist in India. And Nikita one day asked me- can we use your research on virtual agents to address some of the dental issues that her mother is seeing in clinic among young children. And then that snowballed into both of us meeting with David and other team members.
2:29 VO:
Nikita is Dr. Nikita Soni, assistant professor in the department of computer science. She shared more about what sparked this project.
2:39 Nikita Soni:
I think it was all the experiences I had when I used to visit my mom's clinic in India. She's a general dentist in India. I used to see always her interacting with worried parents. She has had patients starting from children who are two days old, they sometimes have some teeth. And I've seen her interacting with parents from that young age to six years. And I see her using innovative approaches to tell parents how to engage kids in brushing and engaging not only parents, but also kids in the processes. So that I had in my mind. I can use my expertise, and collaborate with experts in other domains to help with this public health challenge.
3:30 VO:
Dr. David Avenetti is associate professor and head of the department of pediatric dentistry, the largest provider of Medicaid dental services to children in the State of Illinois. He spoke about some of the challenges linked to poor brushing habits in young children.
3:46 David Avenetti:
What we understand not only in Illinois, but also nationwide, is that pretty significant health disparities exist as it relates to oral health. Many children have extensive tooth decay. Tooth decay is related to lower, oral health related quality of life, increased missed days from school, pain, loss of function- so pretty significant adverse outcomes that come from that. We're always thinking about creative ways to engage in health promotion, educating the population as a whole, but particularly those health disparity populations. Our goal is always primary prevention, but if we can identify those that have increased risk factors and really push forward an intervention that can promote oral health, that would be great.
We are focused on young children. We believe that good habits start early in life. And what we see is that problems, oral health issues, early in life are strongly correlated with issues later in life. We also see, generationally over the life course, that parental attitudes, knowledge, behaviors translate into the behaviors of their children. So, if we can promote positive oral health knowledge and behaviors to parents who can then instill that in their children, not only will they hopefully adopt those behaviors, but also have improved oral health. We know that some people actually know what our goals are. For example, brushing a minimum of twice a day, using fluoride toothpaste, and helping their children with brushing. But one of the things that our intervention is trying to also address is, okay, if a parent has the knowledge and they know what they're supposed to be doing, how can we give them some of the skills and the confidence to follow through with those recommendations?
5:28 Dr. Zalake:
To add to what David said, the other two aspects that we're focusing on are child noncompliance and social stigma. So, young children often resist toothbrushing, making it difficult for caregivers and parents to establish consistent routines. And that's where we have one of our collaborators, who is expert in behavioral science, helping us develop these parenting strategies that could also address some of the child compliance related issues in, following best oral hygiene practices. And also the issue of social stigma. Nobody likes discussing or saying that they don't brush or they have cavities in the teeth. So, concerns about negative social evaluation can discourage some parents from seeking help or discussing their challenges openly outside their closed circle. Having this platform, or virtual agent, that can have this open, non-judgmental conversations with you could be a way towards addressing that specific issue.
6:27 VO:
In the simplest terms, virtual agents are software programs that follow certain rules to provide answers or directions based on a user driven interaction or question. These agents use natural language processing, machine learning, and artificial intelligence to participate in a conversation similar to another human. Only ten years ago, a technology like this would not have been accessible to all segments of the population. Now, however, more than 90% of people, regardless of age or economic status, have a smartphone or computer. But while virtual agents are an increasingly common technology, it was important to the team that they invest the time and resources into creating a virtual agent can engage with families like a health provider and meet them where they are.
7:14 Dr. Avenetti:
Something that was really novel to me when Dr. Mohan and Dr. Nikita reached out was this idea of a virtual agent. And the question that came to mind is, is it possible to create a virtual agent that families can connect with? We always want patients to feel like they connect with their providers, so one of the big goals of the project is to sort of have focus groups and ask the caregivers, how do you feel interacting with these virtual agents? Do you feel like it resonates with you? We're hoping that we will find the answer is “yes.” Then the next phase is, okay, how do we scale that? How do we measure the behavior change that comes out of working or interacting with these virtual agents?
Dr. Soni: Based on user-centered design process, if we just put a technology in hands of parents or children or whoever, it's not going to work automatically. You have to design it in a way that it effectively engages. You encounter people with different tech literacy. Some people know to use certain kind of technology, some don't. There are different family dynamics. When it comes to designing technologies for family context, there are different routines of the family. So you have to take all those contextual factors into mind, and then embed them when you're designing technology so that you're designing it in a way that parents can effectively take up rather than intruding their daily routines. It fits more organically into the way they operate. We are trying to kind of uncover these nuances of the family dynamics, how parents discuss these things with children, how things happen currently, so that we embed what we are designing into that existing process, rather than having parents change the entire process, which might make things quite difficult in long-term engagement. It's not a thing that you interact with the technology one time and it changes everything. It's going to be an iterative, long-term engagement. We are also wanting to understand how to design innovation for the long-term engagement so that we can have patients come, they can see the benefits of using this, and then it could help them with long-term behavior change.
Dr. Zalake: So virtual agents, if you see a human-like virtual character that you usually see in 3D movies or animated movies, you see they have a body, they have a face, and they have voice. All those things are customizable from a technical standpoint where you can change their appearance. Virtualization can look like an Indian middle-aged man like me and have a voice that might have little bit of accent. And then you can change what language they speak. All this ability to personalize or customize to different groups gives a lot of potential where this technology can be applied. For this specific pilot project, we are going to evaluate that to some of the focus groups by showing participants what kind of characters do they more relate to, what kind of voices and what kind of expectations do they have in just establishing that rapport with this virtual character. From a long-term perspective, see how we can apply it to different groups or different communities who might need this kind of application or this kind of virtual agent to be discussing some of the health topics in their own communities.
Dr Avenetti:
Where we're at currently in the project is developing the script and basically how people would respond to different questions and what the follow up prompts would be. Also, creating the virtual agents and kind of the personas that we think would represent many of the types of people we would interact with. We’re hoping that all of this goes as planned. Then next question is really: Okay, we've established the messaging. We know that that works. Now, what is the intensity of that intervention? How frequently should we ping these families and ask them to interact? So, what's the dose? What's the intensity? I think the next phase of this will be to sort of learn how much is enough interaction with these virtual agents to actually lead to behavior, change. And, are there any segments of the population that seemingly like to interact with these individuals more or less? Then we can focus these types of interventions on those people that it would seemingly be most effective for. We're testing a new technology, we're testing the messaging, we're testing this on a new population, we're testing the use of virtual agents. So, if at some point we do see behavior change, we just need to make sure that we fully understand what is most contributing to that, if that makes sense. We have to take this project step by step so that if we do see an effect, we know what led to that effect.
12:22 VO:
Dr. Zalake spoke about the various collaborators on this project and the critical insights they provide into making the technology both evidence based and relatable.
12:33 Dr. Zalake:
We are a team of experts in virtual agents. Nikita’s expertise is in human-computer interaction and designing technologies for families. David brings his expertise in clinical dentistry. And then other team members include Amy Slep. She’s from New York University, and she has expertise in behavioral science. She's the expert who's going to advise us on developing parenting strategies. Then we also have Sobia Bilal, who is from college of dentistry, too. And she brings her expertise in oral health promotion and prevention in public health.
Dr. Avenetti:
I want to give a shout out to a few other people that we have collaborated with. Molly Martin in pediatric medicine reached out to our department because she wanted to do a community-based intervention using community health workers in high risk populations based in Chicago. And, through working in that project, we really developed an interdisciplinary team. We worked with CCTS, we worked with Michael Berbaum. We felt our group worked together really well. So when Dr. Mohan reached out to me and said, “this is what I want to do,” I felt like we already had a team of individuals who were very supportive of oral health promotion. Also Helen Lee, in anesthesia, who is also interested in the oral health field.
Dr. Zalake:
I would like to highlight that all the knowledge that David mentioned that they gained from other projects has been really helpful in actually identifying the solutions, identifying different approaches in our own projects. So, Molly Martin, Helen Lee, they have been advising our team in past year as a senior researchers in this field, and they have helped us actually design or structure this whole project into something that could actually be useful in making the actual health impact that we're trying to do.
14:32 VO:
IVORY is drawing upon two existing approaches to oral health promotion: in-person and digital. In-person information sharing, whether from a clinician or a community health worker, is very effective. But scaling this intervention is expensive and logistically challenging- it would take a lot of people to reach every family in need of help. Digital approaches, like informational websites, can reach a lot of people, but they are less engaging. Virtual agents provide the best of both worlds. Dr. Zalake elaborates.
15:04 Dr. Zalake:
They simulate human face-to-face interaction while being scalable to reach a large number of families. To do this, we need to build upon all the best practices that people have learned from these two previous approaches. We need to understand what works best in face-to-face communication. What kind of conversations happen? How are those conversations converted into something that parents or caregivers can take into consideration when they are changing their daily routines? And from the digital standpoint, how are we going to use existing knowledge of developing interventions that make people change the behavior? For example, some of the things that we are considering in our application is do we need to include push notifications or do we need to nudge people into changing the behavior? And then I further customize these virtual agents. The facial animations that you see, the mouth animations, the body behaviors, the nonverbal behaviors like hand movements of the virtual agent, all those are driven by AI based application, not the actual content itself. We are taking much more of a hybrid approach rather than fully relying on AI based approaches because, as you know, these large language models that are currently popular, it can be really be a major concern when we are developing healthcare applications. We do not want to give misinformation through these platforms. My research assistant is working to identify what are some of the factual content in different papers, different resources shared by the team, and then use AI-based approaches to filter them out to, so that they can be crisp and provided in a small, digestible information. That way we keep the knowledge part accurate and verified by the experts, while also making sure we use the AI to have a virtual agent that can talk to people in a human-like way.
17:25 VO:
Dr. Soni talked about the aspects of the project that could be repurposed by other teams to speed the delivery of other virtual interventions.
17:35 Dr. Soni:
Currently the script is developed by multidisciplinary experts, being verified, content is verified. And this could actually also be used in the future for people who don't have resources to create these scripts. We can make this script public, so that if they are planning to design even a simple chatbot, they can use this to fine tune their existing base model. It could also be tailored for different age groups, different parental strategies, different type of routines parents have, so that parents can query what they want to ask specific to their context and then it could answer. So that could be next step in how the script could be used for designing more advanced applications in the future.
18:36 VO:
In addition to drawing upon the various support CCTS provides campus researchers, like access to biostatisticians and skill building seminars, Dr. Zalake and Dr. Soni talked about aspects of this project that have been helpful in speeding translation.
18:53 Dr. Zalake:
I think the top challenge for virtual agent application in healthcare is having a right team. As you can see in this project, it's a multidisciplinary effort. Each team member plays an important role. I can develop these virtual agents, but for these virtual agent conversations to be actually beneficial, you need to have right expertise, like from David, Nikita, Amy and Sobia. Without them, the conversations would not be hitting the right spot, for lack of better words. And that's where I think the top challenges is identifying the right team who can help you achieve the specific or the larger, broader impact that you want to have on the public health. When we are having these meetings with different collaborators, different interdisciplinary researchers, it's important to understand how different people bring different perspectives on the same topic. Using those different perspectives to actually understand how we can address this problem, that that was the most essential part in this team. And I think that was really helpful in achieving what we were trying to achieve in past one to two years.
Dr. Soni:
I think David was one of the catalysts because he connected us to other people so that they can guide us through the determination of what problem we have to face. He connected us with people like Michael Berbaum to kind of help with the statistical part of it and think through things. So as a junior research assistant professor, from that point having someone in the team who can catalyze and can connect you with right people would be really helpful to move the things forward.
Dr. Avenetti:
Although Dr. Nikita is saying I’m the catalyst, I didn't have to go very far to find some great people- CCTS, across the street to College of Medicine. I think something that we all have to be aware of is that there's a lot of really heavy hitters here on campus. And it's just a matter of bringing awareness to the work that people are doing so that we can connect individuals with the right other individuals in order to sort of synergize ideas and projects.
21:14 VO:
Dr. Soni mentioned another key tactic to help facilitate effective interventions.
21:20 Dr. Soni:
Another thing I would want to add is involving your users early and often. So instead of just taking the technology after you built it to them to evaluate and then assess how effective it is, consider incorporating and involving them as co-designers, with researchers and with teams, so that the people who would end up using it have their voice in the entire design process. They are there from the first day when you start designing to bring their thoughts. They can prototype things. They can use different ways to externalize what they are thinking this should look like so that designers would take that into account in very early stages before you have already invested a lot of cost and time in developing something. Because it's really hard to then go back after we have developed something to accommodate that.
22:12 VO:
While IVORY is still in the early development stage, the team has great aspirations for its future.
22:22 Dr. Zalake:
From a future aspirations perspective, specifically, I see, direct potential impact on other healthcare aspects where issues like child compliance or, lack of knowledge is one of the major issues. So addressing those major, issues like obesity, this is where virtual agents could be really helpful in addressing those problems. And as a long term project, we could, expand this knowledge or the, skills that we gain from this project to address some of the issues in those healthcare topics.
Dr. Soni:
To me, what is really exciting as one of the future goal is to bring this technology as a authoring tool to be in the hands of people who don't have us, don't have the team of programmers, don't have a team of people who can spend a lot of time in developing scripts. So bring the resources, compile a toolkit that they can use to quickly author these things for their own purpose, for the type of population they are specifically targeting. And then use some of the existing advanced technologies, especially with external oversight of healthcare providers to quickly develop something for themselves.
Dr. Avenetti:
Although this is kind of an untraditional way of thinking about increasing access, it does have a much broader reach than what we might be able to have access to within our dental clinic. We have a limited capacity. If we are able to scale this, really there's no limit to the capacity and the number of individuals that we reach.
23:58 VO:
And Dr. Soni had the opportunity to share their early work with her mother.
23:04 Dr. Soni:
We did show some of the initial prototypes to her, and something she's really excited about is she would want to use it and try it in one of our clinics once we have some samples. So that could be potential next step seeing how multicultural communities receive these projects, how they think about it, because there are different preferences and inclinations toward using technologies for different kind of things. So she is very excited.
24:34 Voice Over:
Collaborative Endeavors is produced by me, Lauren Rieger, on behalf of the Center for Clinical and Translational Science (AKA the CCTS) at the University of Illinois Chicago. To learn more about the team behind IVORY, visit the links in our show notes.
The CCTS is supported by the National Institutes of Health’s National Center for Advancing Translational Science through their Clinical and Translational Science Award. Opinions expressed by guests of the show are their own and do not necessarily represent the views of myself, the CCTS or our funding agencies.
You can find more episodes of Collaborative Endeavors on Apple Podcasts, Spotify, Pandora, Amazon Music & YouTube. While we do not monetize this podcast, we do love positive feedback! If you like what you hear, go ahead and give us a 5 star rating to help spread the word. To learn more about how you can work with the CCTS to make a positive impact on the health and wellbeing of our community, visit ccts.UIC.edu or follow us on X @UIC_CCTS.