Navigating the Ethics of AI in Scientific Visual Communication

For over three decades, a dedicated science photographer has played a crucial role in assisting academic professionals, researchers, and students at MIT in visually conveying their groundbreaking work. During this time, a variety of tools have emerged to aid in the creation of striking images; some have proven beneficial, while others may hinder the accurate and comprehensive representation of research findings.

Visual Communication Challenges in Science

In a recent opinion piece featured in Nature magazine, the photographer addresses the rising trend of using generative artificial intelligence (GenAI) in image creation, exploring both the challenges and implications this technology poses for effective research communication. On a personal note, she reflects on the potential future of science photography within the research community.

When discussing the nuances of image manipulation, she asserts that any photo taken can be considered ‘manipulated’ from the moment it is captured. The choices made regarding framing and content structure, as well as the tools employed, already alter reality to some degree. It’s vital to recognize that an image serves merely as a representation, not an exact duplicate of the original subject.

Ethics and Transparency in Image Manipulation

She emphasizes that while some digital alterations can enhance the communication of a specific message—like removing distractions to highlight key features—such changes must not distort the underlying data itself. Transparency is key; she always notes any modifications made to images in her texts.

To ensure research is communicated accurately and ethically, she identifies three significant issues concerning visual representation: the distinction between illustration and documentation, the ethics of digital manipulation, and the ongoing need for researchers to gain training in visual communication. While MIT’s current communication curriculum focuses primarily on writing, it is crucial to integrate visual literacy as well.

“Any photo taken can be considered ‘manipulated’ from the moment it is captured — it’s vital to recognize an image as a representation, not an exact duplicate.”

Training and Guidelines for Researchers

She advocates for researchers to develop critical skills in analyzing published graphs and images to identify any anomalies. Furthermore, discussions about ethical considerations regarding image alterations are essential. In her article, she recounts an experience where a student modified one of her images without permission, highlighting the importance of considering ethics in such actions.

As generative AI continues to evolve, she posits that clear guidelines are necessary for visual communication within the scientific community. For example, during her exploration of AI-generated imagery for the Nature article, she prompted an AI model to create an image of nano crystals, revealing that while some outputs resembled cartoons rather than reality, there will come a time when AI visuals may be indistinguishable from authentic documentation.

While acknowledging that AI-generated visuals can serve illustrative purposes, she firmly believes that they should never replace authentic documentation in scholarly work. Thus, if such visuals are submitted for journal publication or included in presentations, it is imperative that researchers disclose their use appropriately.