Visual Pathways to AI Literacy: Challenges and Lessons Learned in the Medical Domain (Preprint)

Abstract

<sec> <title>UNSTRUCTURED</title> Artificial Intelligence (AI) is becoming increasingly popular in medical research for its perceived advantages in accuracy and dealing with large quantities of data. However, healthcare professionals are rarely experts in AI, which in turn may lead to methodological and reproducibility deficiencies in published research. These non-experts are often users of visual programming platforms, which allow for the creation of machine learning (ML) workflows without extensive programming knowledge. In this article, we will analyze these visual platforms and the challenges they face, with a particular interest in the educative point of view. In addition, to enhance the usability and educational value of visual ML platforms, several improvements are proposed based on user feedback and observed issues. These include segmenting the workflow into distinct phases for data management and ML processing, introducing an interactive tutorial library, providing a curated database library, implementing an icon glossary for better visual comprehension, and integrating a customized checklist to ensure methodological rigor. These enhancements are expected to address identified gaps in AI literacy among healthcare professionals, improving platform usability and research quality. Future evaluations will focus on the effectiveness of these changes in facilitating better user understanding and more reliable AI applications in medical research. </sec>

Ähnliche Arbeiten