3 Questions: Visualizing Research in the Age of AI – Proaitools
Science photographer Felice Frankel, with over 30 years of experience, has been instrumental in helping MIT professors, researchers, and students effectively communicate their work visually. In a recent Nature magazine opinion piece, Frankel delves into the increasing role of generative artificial intelligence (GenAI) in imaging, highlighting both the opportunities and challenges it poses for research communication. She also ponders the future of science photographers in the research landscape.
Q: You’ve mentioned that as soon as a photo is taken, the image can be considered “manipulated.” There are ways you’ve manipulated your own images to create a visual that more successfully communicates the desired message. Where is the line between acceptable and unacceptable manipulation?
A: Frankel explains that the decisions made on how to frame and structure the content of an image, along with which tools are used, inherently manipulate reality. She emphasizes that an image is a representation, not the actual thing. The critical point is to avoid manipulating the underlying data, which in many images is the structure. She provides an example of digitally removing a petri dish to highlight a yeast colony’s morphology, clarifying that the colony’s form was not altered. Frankel always discloses any image modifications in the accompanying text, a practice she discusses in her handbook, “The Visual Elements, Photography.”
Q: What can researchers do to make sure their research is communicated correctly and ethically?
A: Frankel identifies three major issues with visual representation in the age of AI: the distinction between illustration and documentation, the ethical considerations surrounding digital manipulation, and the ongoing need for visual communication training for researchers. She advocates for visual literacy programs, noting that visuals are no longer secondary to journal submissions. Most readers, she suggests, review figures after reading the abstract.
She stresses the importance of teaching students to critically assess published graphs and images for anomalies and to discuss the ethics of influencing an image to fit a predetermined outcome. Frankel recounts an instance where a student altered one of her images without permission, underscoring the need for ethical awareness in visual communication. She calls for campus discussions and a visual literacy requirement alongside writing requirements.
Q: Generative AI is not going away. What do you see as the future for communicating science visually?
A: For her Nature article, Frankel experimented with diffusion models to create an image of Moungi Bawendi’s nanocrystals, using the prompt:
“Create a photo of Moungi Bawendi’s nano crystals in vials against a black background, fluorescing at different wavelengths, depending on their size, when excited with UV light.”
The AI-generated results were often cartoonish and unrealistic. Although AI-generated images will eventually improve, Frankel believes clear standards are needed, especially emphasizing that GenAI visuals should never serve as documentation. She suggests that AI-generated visuals can be valuable for illustration, provided that researchers clearly label the image as AI-created, specify the model used, include the prompt, and provide any source images used to inform the prompt.
</n
