ADEA Innovation Article Series: Dental Schools Develop AI Apps to Share Knowledge, Mitigate Biases and Streamline Workflows
ADEA Innovation Article Series
One of the ADEA Strategic Goals for 2024-2027 is to “Re-envision the Model for Oral Health Education” which includes driving innovation, scholarship and change in oral health education. Some dental schools and allied dental programs have taken the lead and have already begun to innovate practices and technologies within their institutions.
According to the National Institute of Standards and Technology Baldrige Excellence Framework, innovation represents “making meaningful change to improve products, processes or organizational effectiveness and create new value for stakeholders. Innovation involves adopting an idea, process, technology, product or business model that is either new or new to its proposed application.” (Schaefer, 2016)
It's important to understand that innovation extends beyond the creation of new products made possible by technological advancements. Especially for dental schools, innovation can come in many forms in the curriculum, in both preclinic and clinic settings, didactical methods, management methods, clinical management and overall organizational management. This may be a completely new process/method/product or new to the school and delivering “meaningful change and create value” for the school and its stakeholders.
To highlight some of these innovations, ADEA has started a new feature series showcasing oral health schools and programs that are using various innovative practices and technologies. Topics were selected based on what the ADEA members proposed and what the Innovation project team selected last year. This is the second entry in the series.
Introduction
With oral health educators juggling so many responsibilities from daily lectures to clinical hours with students to research to administrative duties, some dental schools are taking an innovative approach to relieving part of their burden by employing artificial intelligence (AI) applications to help tackle tasks.
Alexander Lee, D.M.D., Professor and Assistant Dean for Dental Informatics at Western University of Health Sciences College of Dental Medicine (WU CDM), said his dental school has used several AI apps with various objectives. Most recently, WU CDM began using an AI app to streamline committee work for faculty who review and discuss the dental school’s accreditation standards.
Meanwhile, the Children’s Hospital Colorado (CHCO) used AI to standardize an interview format to assess residency candidates. The application was created not necessarily to save time but to “remove implicit and explicit biases” from the existing process, said Chaitanya P. Puranik, B.D.S., M.S., M.Dent.Sci., Ph.D., FAAPD, Director of Residency Program in Pediatric Dentistry at CHCO and Associate Professor at the University of Colorado School of Dental Medicine (CU SDM). AI initiatives at Western University and Children’s Hospital discussed in this article employ Limited Memory AI, which will be discussed in more detail later in the article.
A Primer on Common AI Applications
Though the term AI is used to encompass a multitude of applications, the technology can be broken down into subcategories. According to the online article, “ Understanding the different types of artificial intelligence,” by IBM (IBM, n.d.), under Artificial Narrow Intelligence, also known as Weak AI (which is the only type of AI that exists currently), there are two types of functional AI categories: Reactive Machine AI and Limited Memory AI. (Other functional AI categories—Theory of Mind AI and Self-Aware AI—are still theoretical technologies that have not yet been implemented, according to IBM.)
Reactive Machine AI systems have “no memory” and are designed to perform a very specific task. Since they can’t recollect previous outcomes or decisions, they only work with presently available data. Reactive AI stems from statistical math and can analyze vast amounts of data to produce a seemingly intelligent output.
An example of Reactive Machine AI given by IBM would be Netflix movie recommendation engine that is “powered by models that process data sets collected from viewing history to provide customers with content they’re most likely to enjoy.” (IBM, n.d.)
In contrast, Limited Memory AI “can recall past events and outcomes and monitor specific objects or situations over time. Limited Memory AI can use past- and present-moment data to decide on a course of action most likely to help achieve a desired outcome.” (IBM, n.d.) IBM noted that even though Limited Memory AI can retain data on past events and outcomes, it cannot do so over long time periods. But this technology can be trained with more data to improve its performance over time.
Limited Memory AI is the type of AI functionality that most AI users would be familiar with. Examples include Generative AI tools like ChatGPT; virtual assistants and chatbots, such as Siri, Alexa and Google Assistant; and self-driving cars.
Choosing the Best AI App and Deployment
Prior to the creation of its AI app that allows committee members to quickly access relevant Commission on Dental Accreditation (CODA) standards and documentation, the WU CDM Continuous Quality Improvement Committee was heavily reliant on institutional knowledge from older, more seasoned faculty members. But because of their own busy schedules, those faculty members weren’t always available to share the full breadth of their knowledge and expertise.
“Committee members—especially newer committee members, because we try to embrace our younger faculty as well—were a bit lost as to what the accreditation process was, and a lot of people actually hadn’t read the accreditation documents. Or if they’d read it, they didn't understand it,” Dr. Lee said. “… So, what would happen would be these faculty would go into these committee meetings, and they'd be very lost.”
Like many dental institutions with tight budgets and timeline constraints, WU CDM is “not in a position and we don’t have the resources to simply just throw personnel or push back deadlines in order to accommodate something,” Dr. Lee said, particularly they could not hire personnel to do research prior to each accreditation committee meeting.
Within 72 hours, Dr. Lee was able to develop an innovative (and frugal) solution to the committee’s problem by creating an AI app with the subscription-based AI Chatbot builder, Chatbase, for roughly $100 a month. The chatbot he developed pulls CODA standards and other accreditation and self-study documents relevant to the WU CDM for committee members.
“What we do with Chatbase is that we train the chatbot with specific information to our institution and then we deploy it to our users who then can ask it questions,” Dr. Lee said.
He said the dental school’s IT team chose Chatbase to solve this problem because its chatbots are easy to build, offer versatility and are user-friendly.
“On the back end for me, I can literally pull up the inner workings of a chatbot. Right now, I have it using ChatGPT 5. It's a large language model,” he said.
Large language models (LLMs), according to IBM, are a category of “deep learning models” whose designs are inspired by the human brain and are “trained on immense amounts of data, making them capable of understanding and generating natural language and other types of content to perform a wide range of tasks.” (Stryker, n.d.)
“GPT 5 is the current LLM I use as the engine for my chatbot, but with a push of a button, I can change it to Gemini 2.5 Pro,” he said, which is another LLM. “I can change it to Grok. I can change it to any other AI technology.”
Dr. Lee said another perk of the Chatbase app is, “I also am able to restrict it to just my organization.”
When CHCO piloted a chatbot to revamp the interview format for its residency candidates, the dental school knew the format had to be consistent and unbiased.
“The existing strategy for the interviews was very subjective and needed more objective, data-driven approach,” Puranik said.
CHCO decided to use ChatGPT as the base of its chatbot because of the technology’s “access and user-friendly nature.”
They decided to make the chatbot more “robust by iterative refinement and training,” he said.
Puranik said developing the process required “rigorous training of staff and licenses.”
“The program directed and invested a lot of time researching how the interviews should be formatted and how AI can be utilized for some of the aspects,” Puranik said. It was almost “a year-long process with involvement from the legal department.”
Developing the process involved an open-source ChatGPT license and “now we are transitioning to a Microsoft Azure environment at CHCO.”
Securing Buy-in and Training on Proper Use
Dr. Puranik said initially there was an issue with securing buy-in from faculty and staff for the interview process. “Yes, there were concerns but we had a stakeholders meeting and addressed any concerns,” he said.
Dr. Puranik said these concerns spanned the gamut from “ethical, moral, implementation-related and lack of training.”
But eventually faculty and staff were able to adapt. “We are three years out now and it seems to be smooth running now,” Dr. Puranik said.
In contrast, Dr. Lee said he didn’t encounter any hiccups with the buy-in for the app he designed for the accreditation committee mainly because the user base is small and the use of the app is optional.
The committee members “are the only people that need to see it. If they need it, here you go. Do you think you need it? It's optional. It's just a resource.”
He also said WU CDM, with its four-person IT team, synergizes well with the larger WesternU IT team, which makes new technology adoption easier.
"[WU CDM’s IT team] has a very strong relationship with our IT team here at the university,” Dr. Lee said. ”We go out to lunch with them. We’re friends and so they trust our college’s IT team a great deal.”
A Crucial Caveat With AI
It is important to note that even though AI apps are useful tools, like any technology, they are not without their downsides. According to the blog post, “LLM hallucinations and failures: lessons from 5 examples,” by Evidently AI, (Maliugina, 2025) a software developer that has a tool designed for monitoring AI machine learning models in production, one common issue users experience with LLMs is hallucinations. These occur when “the model generates factually incorrect or fabricated information," according to the blog.
The blog cited one example of hallucination during a New York federal court filing, where one of the lawyers was caught citing non-existent cases. “It turned out he was using ChatGPT to conduct legal research—the bot referenced fake cases to the attorney,” the blog stated. “Responding to the incident, a federal judge issued a standing order that anyone appearing before the court must either attest that ‘no portion of any filing will be drafted by generative artificial intelligence’ or flag any language drafted by AI to be checked for accuracy.”
Dr. Lee said even though committee members are encouraged to use the accreditation standards chatbot, they still have to verify the information and sources given—just like they would with any other AI app.
“I teach an AI literacy course to our students, and I teach it to all our faculty and staff as well,” Dr. Lee said. In the course, users are educated about issues like this.
“Our users know that at least with AI at the current moment, it is always a ‘trust but verify’ circumstance,” he said.
Despite these technological tradeoffs, both WU CDM and CHCO said they’ve had positive results and perks with their respective AI deployments.
Thanks to the app, institutional knowledge of CODA standards and documents related to the WU CDM are always available for committee members.
Dr. Lee also noted that the answers produced are “more standardized versus the answers from people.”
Dr. Puranik said CHCO, by standardizing its interview format for residency candidates, the program “has substantially reduced the human time and produced reliable and consistent results.”
What You Should Know Before Developing an AI Application
Dental schools considering a similar deployment of AI apps should first and foremost make sure that AI is the proper solution for the problem, Dr. Lee said.
He said for WU CDM’s IT team, it is always important to find the right circumstance and muster the resources that are needed “because our unofficial motto amongst our IT group is ‘technology with purpose.’”
"We never implement technology just to implement technology,” Dr. Lee said. “As long as our team is involved, we never find ourselves involved in a circumstance where we have a technology that’s looking for a problem.”
Once you have determined that an AI app is the right solution, the next steps are doing the proper research, securing stakeholder buy-in, developing the technology, and deployment.
Has your dental school also developed an AI app to help solve a problem within your institution? How is it performing? Let us know at adeadata@adea.org.
References
- IBM. (2025, December 10). Understanding the different types of artificial intelligence. IBM. Retrieved.
- Maliugina, D. (2025, October 8). LLM hallucinations and failures: lessons from 5 examples. EvidentlyAI Blog.
- Schaefer, C. (2016, June 2). How the Latest Baldrige Award Winners Manage for Innovation. Blogrige.
- Stryker, C. (n.d.). What are LLMs? IBM Think.
About ADEA: The American Dental Education Association (ADEA) is The Voice of Dental Education. Our mission is to lead and support the health professions community in preparing future-ready oral health professionals. Our members include all 87 U.S. and Canadian dental schools, more than 800 allied and advanced dental education programs, more than 50 corporations and approximately 15,000 individuals. Our activities encompass a wide range of research, advocacy, faculty development, meetings and communications, including the esteemed Journal of Dental Education®, as well as the dental school application services ADEA AADSAS®, ADEA PASS®, ADEA DHCAS® and ADEA CAAPID®. For more information, visit adea.org.