Research reveals AI's double-edged impact on health information

Research reveals AI's double-edged impact on health information

Did you know that asking an AI about heart problems in Vietnamese could potentially produce advice about Parkinson's disease instead? This is just one of the eye-opening findings from a study led by RMIT Vietnam researchers, recently published in the British Medical Journal (BMJ).

The study, ‘Health Information for All: Do Large Language Models Bridge or Widen the Digital Divide?’ brings attention to a pressing concern in the AI-driven future of healthcare. While Large Language Models (LLMs) like ChatGPT offer the promise of 24/7 access to health information, the research team found that these models could potentially provide inaccurate or misleading information in languages with limited digital resources.

Over 80 per cent of all web content is in just eight languages, according to Statista. However, these languages represent only 21 per cent of the world's population – leaving the majority potentially vulnerable to AI-generated misinformation.

Dr Arthur Tang, the lead author from RMIT Vietnam said, "We discovered that LLMs, which perform well in high-resource languages like English, often struggle with linguistic fidelity and informational accuracy in low-resource languages like Vietnamese. Regions where low-resource languages are used tend to be economically disadvantaged. That means that technological advancements like LLMs, in the present state, could deepen social and economic disparities."

ChatGPT app on smartphone Large Language Models could potentially provide inaccurate or misleading information in languages with limited digital resources. (Photo: Pexels)

The study is the result of international cooperation, involving researchers from RMIT University Vietnam, Oxford University Clinical Research Unit (Vietnam), Chinese University of Hong Kong, National University of Singapore, and University of Melbourne. What makes this research particularly noteworthy is the inclusion of two then undergraduate students from RMIT Vietnam as co-authors.

The students received essential technical support and computational resources from the RMIT RACE Hub. The RMIT RACE Hub also enhanced the students’ learning experience by giving them exposure to cutting-edge research equipment.

Nguyen Quynh Giang, a recent graduate in Information Technology said: "Working on this research opened my eyes to the real-world implications of AI in healthcare. Ideally, AI will help provide equal access to healthcare and health information, but there are some barriers until we get to that stage.”

Bui Minh Nhat, who completed a bachelor’s degree in Software Engineering at RMIT Vietnam and is studying for a Master of Artificial Intelligence at RMIT in Melbourne, reflected on the practical impact of their work: "Our research showed that something as simple as asking about heart condition symptoms in Vietnamese could result in receiving information about an unrelated disease like Parkinson’s. This highlights the urgent need to develop AI systems that are inclusive and beneficial for all languages and cultures."

The research team emphasised that addressing this issue requires multisectoral initiatives from policymakers, research funding agencies, big tech companies, the research community, healthcare practitioners, and linguistically under-represented communities.

Funding agencies are pivotal in expanding support for AI language inclusivity. An example of this effort is the Bill & Melinda Gates Foundation's Grand Challenges 2023, which funded global development projects that will contribute to AI equitable access.

Big technology corporations also play a big role. Meta's No Language Left Behind project, for instance, aims to enhance AI translation for a diverse range of languages.

The study also states that the research community should lead initiatives in open-source linguistic data, models and tools that are essential for LLM training and research.

Dr Tang said, “At RMIT, we are deeply involved in various projects that apply LLMs in healthcare settings. Our efforts are focused not only on overcoming the technological challenges but also on enhancing the inclusivity of this technology for the Vietnamese language. This ensures that all Vietnamese people have the opportunity to benefit from recent technological advancements.”

Nguyen Quynh Giang (left) and Bui Minh Nhat (right) took part in the research project when they were undergraduate students at RMIT Vietnam. Nguyen Quynh Giang (left) and Bui Minh Nhat (right) took part in the research project when they were undergraduate students at RMIT Vietnam.

Dr Tang is particularly proud of the students’ involvement in the research, noting that the project originated as a spin-off from an undergraduate capstone project last year.

“The British Medical Journal is one of the oldest and most respected journals in general medicine. Its acceptance rate is about seven per cent, out of seven to eight thousand submissions per year. So, this is truly an outstanding achievement for our student co-authors,” he said.

"Our students are gaining hands-on experience with AI technologies that are reshaping industries worldwide. This prepares them to be leaders in applying AI ethically and effectively in their future careers.”

The research project discussed is supported by key contributors from the School of Science, Engineering & Technology at RMIT University Vietnam, including Dr Stanley Luong, Mr Tom Huynh, and Associate Professor Minh Dinh. This project was undertaken with funding from the RMIT Vietnam Strategic Innovation Challenge, and with the assistance and computing resources from RMIT AWS Cloud Supercomputing (RACE Hub).

Story: Ngoc Hoang

Thumbnail image: WavebreakMediaMicro – stock.adobe.com

15 October 2024
Engage

Related news

Team discussion

Cross-border hackathon attracts bright minds in AI and cyber security
Science and technology
Icon / Small / Calendar Created with Sketch. 18 Nov 2024

Students from Australia and Vietnam came together for RMIT University's first GenAI and Cyber Security Hackathon to tackle cutting-edge challenges.

Dao Hieu Phong portrait picture

Teen inventor of AI waste-sorting robot wins RMIT scholarship
Media Release
Icon / Small / Calendar Created with Sketch. 18 Oct 2024

A straight-A student, young inventor, and jazz music lover – Dao Hieu Phong’s talents and interests are as broad as his curiosity.

Woman holds stethoscope attached to a smartphone

Research reveals AI's double-edged impact on health information
Science and technology
Icon / Small / Calendar Created with Sketch. 15 Oct 2024

Did you know that asking an AI about heart problems in Vietnamese could potentially produce advice about Parkinson's disease instead?

news-thumb-healium.jpg

RMIT Vietnam students develop AI-powered medical information system
Science and technology
Icon / Small / Calendar Created with Sketch. 14 Oct 2024

HealthLight, an innovative AI-powered medical information system developed by team Healium from RMIT, is expected to address a critical gap in public health information accessibility in Vietnam.

Load more +