Number
808
Name
Implementing Large Language Models To Support Misconception-Based Collaborative Learning In Healthcare Education
Date & Time
Sunday, June 7, 2026, 5:30 PM - 7:00 PM
Location Name
Oglethorpe Ballroom
Speakers
Authors
Brandon C.J. Cheah, Department of Microbiology and Immunology, Yong Loo Lin School of Medicine, National University of Singapore
Shefaly Shorey, Alice Lee Centre of Nursing Studies, Yong Loo Lin School of Medicine, National University of Singapore
Jun Hong Ch\'ng, Department of Microbiology and Immunology, Yong Loo Lin School of Medicine, National University of Singapore
Chee Wah Tan, Department of Microbiology and Immunology, Yong Loo Lin School of Medicine, National University of Singapore
Presentation Topic(s)
Technology and Innovation
Description
PURPOSE
Healthcare misinformation can shape learners’ prior beliefs and contribute
to entrenched misconceptions that impair conceptual understanding and
subsequently clinical reasoning. Traditional approaches to identifying such
misconceptions, such as manual review of assessment responses, are
time-intensive, dependent on educator experience, and assume that
misconceptions are homogeneous across cohorts. Such approaches comprise
misconception-based collaborative learning (MBCL), where learners are
explicitly exposed to misconceptions and work collaboratively to refute them.
Large Language Models (LLMs) can offer a unique solution by assisting
educators in generating misconceptions and identifying idiosyncratic thought
patterns with instant feedback. This study proposes a framework for
integrating LLMs into MBCL to support scalable generation in healthcare
education.
METHODS
This study identifies areas in the conceptual acquisition process for
potential LLM augmentation, and how educator-led or student-derived
misconceptions can be improved with LLM assistance. Prompts to ChatGPT-4o
were standardized through stating the LLM’s expected role (i.e. university
lecturer) and specific learning objectives (i.e. SARS-CoV-2 viral evolution),
with the aim of generating common misconceptions and use cases across
healthcare education disciplines (Pharmacy, Nursing, Medicine, Molecular
Biology and Microbiology). Outputs were qualitatively assessed for thematic
relevance to concepts intended for instruction.
RESULTS
ChatGPT-4o generated misconceptions thematically relevant to learning
outcomes but had inherent assumptions about a student’s prior knowledge in
the subject, implying a metacognitive hindsight bias inherent to LLMs.
Clarification prompts are recommended to ensure proper scoping. A 10-step
guide including prompt setup, roleplay and suggestions for presentation is
proposed to implement the conceptual framework.
CONCLUSION
Misconceptions, when leveraged intentionally with LLMs, can improve student
learning with personalized feedback by supplementing misconceptions. However,
the educator’s role is crucial for defining context and cross-checking LLM
output. Methods developed in this study have the potential to reduce time
taken for educators to generate misconceptions, while encouraging greater
student agency in collaborative learning.
Healthcare misinformation can shape learners’ prior beliefs and contribute
to entrenched misconceptions that impair conceptual understanding and
subsequently clinical reasoning. Traditional approaches to identifying such
misconceptions, such as manual review of assessment responses, are
time-intensive, dependent on educator experience, and assume that
misconceptions are homogeneous across cohorts. Such approaches comprise
misconception-based collaborative learning (MBCL), where learners are
explicitly exposed to misconceptions and work collaboratively to refute them.
Large Language Models (LLMs) can offer a unique solution by assisting
educators in generating misconceptions and identifying idiosyncratic thought
patterns with instant feedback. This study proposes a framework for
integrating LLMs into MBCL to support scalable generation in healthcare
education.
METHODS
This study identifies areas in the conceptual acquisition process for
potential LLM augmentation, and how educator-led or student-derived
misconceptions can be improved with LLM assistance. Prompts to ChatGPT-4o
were standardized through stating the LLM’s expected role (i.e. university
lecturer) and specific learning objectives (i.e. SARS-CoV-2 viral evolution),
with the aim of generating common misconceptions and use cases across
healthcare education disciplines (Pharmacy, Nursing, Medicine, Molecular
Biology and Microbiology). Outputs were qualitatively assessed for thematic
relevance to concepts intended for instruction.
RESULTS
ChatGPT-4o generated misconceptions thematically relevant to learning
outcomes but had inherent assumptions about a student’s prior knowledge in
the subject, implying a metacognitive hindsight bias inherent to LLMs.
Clarification prompts are recommended to ensure proper scoping. A 10-step
guide including prompt setup, roleplay and suggestions for presentation is
proposed to implement the conceptual framework.
CONCLUSION
Misconceptions, when leveraged intentionally with LLMs, can improve student
learning with personalized feedback by supplementing misconceptions. However,
the educator’s role is crucial for defining context and cross-checking LLM
output. Methods developed in this study have the potential to reduce time
taken for educators to generate misconceptions, while encouraging greater
student agency in collaborative learning.
Presentation Tag(s)
International Presenter, Student Travel Award Winner, Student Presentation