The number of AI-generated figures in scientific publications is increasing, unfortunately leading to high-profile retractions of papers featuring inaccurate visualizations. The lack of definitive guidelines for AI-generated scientific and educational visualizations results in several ethical issues and dilemmas. At the same time, we maintain that there should not be a double standard regarding the factual correctness of figures only due to AI involvement in their creation and argue in favor of measured responses. We present a framework considering the communicative purpose of a visualization, the type and function of the figure in a paper, the type of error, risks, and the appropriateness of the figure as a means to support decisions regarding the severity of issues of AI-generated images for scientific and educational aims. By outlining a more fine-grained analysis of error types and visualization characteristics, we provide orientation for the current controversy surrounding AI-generated figures. This framework can also serve as a starting point for considerations regarding AI use by students. In addition, we discuss more sophisticated ways of using AI systems to generate visualizations that avoid the pitfalls of general-purpose text-to-image tools.
No takes yet. Share an insight, caveat, or question.
Skulmowski et al. (2025) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: