Understanding the Gap Between Academic and Public Perceptions of Medical AI
With smartphones and social media now integral to daily life, discussions about medical artificial intelligence (AI) are no longer limited to experts. From hospital decision support systems to wellness apps and drug development, medical AI is rapidly transforming our everyday lives. But how does the academic community view medical AI? And how do ordinary people feel about it? The study introduced today takes on these two questions directly. By analyzing scholarly articles and social media comments using BERTopic and SnowNLP, it compares how different groups perceive and emotionally respond to the same issue: medical AI.
Backgroud
How academics and the public think about medical AI
Over the last decade, medical AI has grown at an extraordinary pace. As algorithms improve diagnostic accuracy, enable personalized treatment plans, and accelerate drug discovery, the societal impact of AI continues to expand. At the same time, the medical domain brings inevitable concerns around data privacy, ethics, misdiagnosis risks, and accountability.
In this landscape, academics tend to focus on structural and long-term issues, while the public responds more directly to the immediate benefits and risks they experience in daily life.
Understanding both viewpoints is essential for discussing the future direction of medical AI—and this study fills that gap.
Core Idea
Comparing agendas and emotional differences between groups
Academics and the general public naturally interact with medical AI from different perspectives. While academia emphasizes technology, policy, and clinical applications, the public reacts more strongly to personal usefulness, uncertainty, and privacy concerns. For this reason, the researchers used two distinct datasets—scholarly articles and social media posts—to capture how each group thinks and feels about medical AI.
Solution
1. Comparing topic landscapes using BERTopic
The researchers applied BERTopic to 8,110 scholarly documents and 57,537 social media posts, extracting eight major topics for each group. BERTopic uses Google’s BERT language model, then performs document embedding → UMAP dimensionality reduction → HDBSCAN clustering → c-TF-IDF keyword extraction to identify coherent topic structures.
2. Quantifying sentiment with text mining
SnowNLP was used to analyze sentiment for each topic, classifying texts into positive, negative, and neutral categories.
3. Tracking changes over time
The scholarly literature was divided into three phases—emergence (P1), growth (P2), and rapid expansion (P3)—to examine how topics and sentiments evolved.
Result
Different interests and different temperatures
Academic literature revealed themes focused on systems and technology—medical education, clinical diagnosis, medical ethics, hospital informatization, medical imaging, disease prediction, assistive robotics, and interdisciplinary research.
Public discussions, however, reflected daily experience and emotional concerns—humanistic care, health management, drug development, medical education, clinical diagnosis, assistive robotics, the pharmaceutical market, and data privacy.
Only three topics overlapped between the two groups: medical education, clinical diagnosis, and assistive robotics. This shows that the two groups perceive medical AI from fundamentally different vantage points. Even within shared topics, emotional responses diverged sharply. For instance, academics viewed medical robots as a promising technological advancement, while the public expressed anxiety—“Is it safe to have machines replace humans in care?”
Academics tended to show neutral or positive sentiment toward topics like medical ethics or hospital informatization. The public, on the other hand, showed strong negative sentiment toward humanistic care, data privacy, and clinical diagnosis—especially concerns about whether medical data is being safely protected. In attitudes toward new technologies, the contrast deepened. Academics generally respond with optimism and focus on research possibilities, whereas the public reacts first to risks and uncertainty. In the case of disease-prediction AI, academics highlight improved diagnostic accuracy, while the public worries about misdiagnosis.
This study clearly demonstrates how differently academia and the public think about medical AI—quantitatively and thematically. Academics focus on technical progress and institutional improvement, while the public interprets medical AI through emotional and everyday lenses. In the end, the sustainable growth of medical AI depends not only on scientific innovation, but also on building trust, ensuring transparency, and strengthening ethical safeguards.
NetMiner Team
netminer@cyram.com
References
[1] Li, C., & Hu, X. (2025). Medical artificial intelligence in scholarly and public perspective: BERTopic-based analysis of topic-sentiment collaborative mining. Data Science and Informetrics, 5, 33–42. https://doi.org/10.1016/j.dsim.2025.05.001