Chatbots in Healthcare: Why Personalize a Diabetes Chatbot?

Authors: Aaditya Malhotra (1), Bera Ankaya (2), Cindy Kuete (3), Gulsima Karsli (4), Marwa Hemat (5)

Editor: Fatma Zehra Tunca, George Washington University

Author Affiliations: (1) Thomas Jefferson High School for Science and Technology, (2) Deep Run High School, (3) CMIT North High School, (4) George Mason University, (5) Alexandria City High School

Introduction

Chatbots, which are computer programs designed to simulate human conversation, are rapidly becoming a transformative tool in healthcare. Chatbots began as basic tools for answering FAQs and website navigation, but have evolved into highly intelligent, adaptive, and responsive systems. Healthcare organizations use virtual assistants through chatbots to provide patients with accessible, efficient, around-the-clock services. The healthcare industry uses AI-powered systems to help patients track symptoms and understand medical conditions, schedule appointments , and even receive mental health support.

How Chatbots Support Patient Care

Their usefulness goes beyond convenience. The AI-powered symptom check and medical guidance system of Babylon Health operates as a platform. The virtual nurse functionality of Florence helps patients remember their medication schedule and monitor their treatment progress , while Ada Health provides symptom interpretation through individual health assessments.

Figure 1.0. Growth of the Healthcare Chatbot Market: This graphic highlights the rapid expansion of the healthcare chatbot market, which surpassed 430 million dollars in 2023. The growing adoption reflects how digital health tools improve care accessibility while reducing the burden on clinical staff

Streamlining Healthcare Systems

The widespread adoption of chatbots stems from their capability to enhance operational efficiency without diminishing patient interaction. The automation of basic administrative work through chatbots enables healthcare providers to dedicate their time to treating complex and urgent medical situations. Through their functionality, patients can determine if their symptoms need emergency room visits while receiving pre-procedure instructions and post-discharge feedback. Healthcare chatbots demonstrate the ability to resolve eighty percent of typical patient inquiries. The implementation of chatbots reduces both time consumption and expenses because it transforms multiple-dollar customer service interactions into minimal three-cent transactions.

A Brief History of Chatbots

Healthcare chatbots have experienced rapid development since their inception, although their origins extend back to 1965. The first chatbot, named ELIZA, emerged in 1965 to mimic psychiatric conversations with patients. ELIZA operated through keyword matching to provide pre-programmed responses, which established the groundwork for future human-computer dialogue research. The development of chatbots continued with ALICE and Parry, focusing on mental health and entertainment before Siri, Google Now, and Alexa introduced voice assistance to everyday life.

The Impact of the COVID-19 Pandemic

The COVID-19 pandemic established itself as a defining moment in history. The sudden increase in worldwide need for accessible contact-free medical advice led chatbots to become essential frontline tools. The healthcare system implemented chatbots for patient triage , symptom assessment , test appointment booking, and health guideline distribution. Since their introduction, chatbots have established themselves as fundamental components for managing communication and support across industries. Healthcare organizations utilize chatbots to provide patient education while managing chronic diseases, planning proactive care, and delivering mental health therapy. The business world has also adopted them. Chatbots serve as essential tools for retail businesses, banking institutions, insurance companies, social media platforms , and food service operations because they reduce costs and enhance user satisfaction while saving time. The research conducted by chatbot.com shows that 84 percent of businesses predict chatbots will play a vital role in customer communication in the future.

What the Future Holds for Chatbots

The technology's future development makes this advancement particularly thrilling. Future chatbots will possess the capability to detect emotions through voice tones , written words , and facial expressions. Advanced natural language processing will enable these systems to handle sophisticated dialogues. The technology will advance to become hyper-personalized by offering support for text, audio, and video interactions and maintaining memory of previous conversations spanning weeks or months. The combination of multi-agent reinforcement learning enables these systems to work with other AI systems for complex task solutions, such as an HR chatbot working with finance and IT bots to handle multi-part employee requests. Such advanced intelligence, along with team-based capabilities and personalized features, will transform our relationship with technology.

Limitations and Risks of Chatbots in Healthcare

Healthcare chatbots present risks due to the extensive innovation happening in this field. The main issue with healthcare chatbots is their potential to distribute false information. A patient's health becomes endangered when a chatbot delivers wrong guidance through accidental or intentional mistakes. The risk to patient health becomes most significant when users depend on chatbots for critical medical situations or urgent care needs. Another concern is overreliance. Patients who heavily depend on chatbots might postpone getting medical help and neglect their scheduled appointments. The ability of chatbots to handle complex emotional discussions remains limited. The system performs well with structured inquiries, yet fails to detect emotional signals and handle mental health emergencies effectively.

Ethical Challenges and Data Protection

The protection of user data, along with privacy issues, represents vital obstacles to overcome. The secure storage of user data, along with HIPAA compliance, becomes vital because health chatbots handle sensitive information. The deployment of tools without medical oversight and the rush to deploy tools led some chatbot developers to neglect best practices. The training data contains biases that create another significant problem. The training data used for chatbot development contains only specific population groups , which leads to unfair and inaccurate responses for minority groups and underserved populations. The legal and ethical responsibility to bear liability remains unclear. The responsibility for harmful advice from a chatbot remains unclear between the developer , the healthcare provider , and the AI system. The development of chatbots requires careful medical supervision and transparent methods to address ongoing debates about their use.

Inside Diabot’s Architecture

What makes Diabot effective is the thoughtful technology stack behind it. At the core is Azure OpenAI’s GPT-3.5 Turbo, a powerful language model that enables Diabot to comprehend responses in clear, natural language. This is important because it allows for human-feeling conversations for people managing diabetes.

To support this architecture, as a feature or function of the application, is a simple and well-designed data preparation and training process relative to tools like Google Colab and Pandas. Data is organized in a JSONL format, facilitating the development of patient scenarios and common questions with user dialogue flows. The chatbot is fine-tuned to only be trained with previously researched information. This organizational effort allows Diabot to respond accurately and consistently in a real-world, real-time environment.

Figure 2.0. Fine-Tuning a Pre-Trained Model for Targeted Tasks: This diagram shows how a pre-trained model is adapted through fine-tuning to improve performance on a specific downstream task, using additional data while retaining the foundational structure of the original model.

From the user's perspective, Diabot is a mobile application developed in Xcode with Swift and SwiftUI. Diabot designates a secure method for requesting APIs with a backend using Swift's URLSession to communicate and retrieve results quickly and assuredly. The demo uses a mock log-in function to demonstrate how a user might sign in and access a user's standard but personally identifiable information in a "real" setup.

Diabot runs entirely on Microsoft Azure, providing solid mechanisms for keeping it reliable, private, scalable, and ultimately flexible. This combination retains the simplicity and effortlessness of using Diabot and the smoothness and dependability of Diabot whenever a user needs it.

The Future of Diabot

Diabot should have the ability to do much more than chat in the future. If it is connected to glucose monitors or electronic health records, it could generate real-time alerts when you are in hyperglycemic or hypoglycemic conditions and would lessen the likelihood of episodes by recommending action such as taking insulin or a quick snack. The Diabot could also remind a patient to check their blood sugar after eating or following exercises. Additionally, Diabot might have the capability of monitoring longer-term trends to demonstrate how various food items, stressors, and activities affect a patient's blood glucose levels.

Diabot could be useful to doctors, too. Diabot could recognize episode patterns, such as frequently having low blood sugar or frequently having high blood sugar episodes, and communicate these to the healthcare providers so they could modify treatment to reduce the likelihood of recurrent episodes. Diabot could also include support for multiple languages or the use of voice replies for those who find reading or typing difficult, which would make diabetic management easier for everyone.

In the longer term, the Diabot could become a consistent companion that understands your daily routine, including typical meal timings, typical workouts, speaks your language, works with your monitoring devices, and communicates with your care team! The Diabot could take advantage of advances made in artificial intelligence and mobile technology to evolve into a trusted friend that empowers others to live a healthier and more fulfilling life as someone with diabetes.

Glossary

1. AI (Artificial Intelligence): A field of computer science focused on developing machines and systems capable of performing tasks that typically require human intelligence, such as learning, reasoning, decision-making, and conversation.

2. Chatbot: A software program that uses AI to simulate human conversation. It can interact with users through text or voice to answer questions, provide services, or complete tasks.

3. ChromaDB: A vector database used in AI applications to store and retrieve embeddings from past interactions or documents. It helps chatbots like Diabot respond in a context-aware, memory-efficient manner.

4. Diabot: An AI-powered chatbot built to support people living with diabetes. It provides personalized guidance, medication and glucose monitoring reminders, symptom interpretation, and emotional support.

5. EHR (Electronic Health Record): A digital record of a patient's comprehensive medical history, including diagnoses, test results, treatment plans, prescriptions, and provider notes. It is maintained by healthcare professionals.

6. ELIZA: The first known chatbot, created in 1965, which mimicked a psychotherapist using simple keyword-matching techniques. ELIZA laid the foundation for future developments in human-computer dialogue.

7. GPT (Generative Pre-trained Transformer): An advanced AI model developed by OpenAI that can understand and generate human-like text. Diabot uses GPT-3.5 Turbo to power its natural, responsive conversations.

8. HIPAA (Health Insurance Portability and Accountability Act): A United States law that ensures the confidentiality, integrity, and security of protected health information (PHI), setting standards for data privacy in healthcare.

9. LangChain: A framework designed to manage how AI systems reason and interact. In tools like Diabot, LangChain helps structure multi-step conversations and improve contextual understanding.

10.  Multi-Agent Reinforcement Learning (MARL): A machine learning approach where multiple AI agents learn through interaction and collaboration to complete tasks, such as coordinating between different services or bots.

11. Natural Language Processing (NLP): A branch of AI that focuses on enabling machines to read, understand, and respond to human language in a meaningful way. NLP is key to how chatbots interpret user input.

12. Symptom Checker: An AI tool that guides users through identifying potential health conditions based on their symptoms. It offers suggestions or recommendations for next steps, such as seeking medical care.

References

Alhejaily, Abdul‑Mohsen G. “Artificial Intelligence in Healthcare (Review).” Biomedical Reports, vol. 22, no. 1, 12 Nov. 2024, doi:10.3892/br.2024.1889.

Baig, M. M., H. GholamHosseini, and M. J. Connolly. “Mobile Healthcare Applications: System Design Review, Critical Issues and Challenges.” Australasian Physical & Engineering Sciences in Medicine, vol. 38, no. 1, Mar. 2015, pp. 23–38. Epub 5 Dec. 2014. doi:10.1007/s13246-014-0315-4. PubMed, PMID: 25476753.

Bocian, Zuzanna. “Key Chatbot Statistics You Should Follow in 2023.” ChatBot Blog, Chatbot.com, Accessed 16 July 2025, www.chatbot.com/blog/chatbot-statistics/.

Coombes, Emily. “Chatbots: From the 1950s to Now.” Japeto AI, 9 Jan. 2025, www.japeto.ai/chatbots-from-the-1950s-to-now/.

Eldridge, William L., and Uma Gupta. “Four Examples of Healthcare Chatbots.” Scholar Commons – University of South Carolina Undergraduate Research Symposium, 2024. Accessed 16 July 2025, scholarcommons.sc.edu/scurs/2024symposium/2024presentations/18/.

“Five Exciting Ways AI Technology Is Transforming Healthcare.” Touro University News, 6 Mar. 2024, www.touro.edu/news--events/stories/five-exciting-ways-ai-technology-is-transforming-healthcare.php.

Floatbot.ai. “Chatbot History and Use Cases of Chatbot – Conversational AI 2025.” Floatbot.ai Blog, Accessed 16 July 2025, floatbot.ai/blog/the-evolution-of-chatbots-from-origin-to-conversational-ai.

“The Future of AI Chatbots: 9 Key Trends Ahead.” CHI Software Blog, 6 June 2025, chisw.com/blog/future-trends-in-ai-chatbots/.

Nandigrami, Prithviraj, et al. “Computational Assessment of Protein‑Protein Binding Specificity within a Family of Synaptic Surface Receptors.” The Journal of Physical Chemistry B, vol. 126, no. 39, 2022, pp. 7510–7527. doi:10.1021/acs.jpcb.2c02173.

Priede, David L. “Are Healthcare AI Models Trained with My Health Data?” Biolife Health Center, 23 Apr. 2025, www.biolifehealthcenter.com/post/are-healthcare-ai-models-trained-with-my-health-data.

Sciences, Chinese Academy of. “AI for Diabetes Management: Personalized Care and Early Detection.” Newswise, 25 Feb. 2025, www.newswise.com/articles/ai-for-diabetes-management-personalized-care-and-early-detection.

Thomas, L. Brannon, et al. “Artificial Intelligence: Review of Current and Future Applications in Medicine.” Federal Practitioner, vol. 38, no. 11, 2021, pp. 527–38. doi:10.12788/fp.0174.

Wang, Yingtan, et al. “Pathological Gambling in a Patient on Piribedil: A Case Report.” Medicine, vol. 100XX, 12 Feb. 2021. U.S. National Library of Medicine, pmc.ncbi.nlm.nih.gov/articles/PMC7886474/