AI-Generated Figures in Academic Publishing: Policies, Tools, and Practical Guidelines
This paper surveys the inconsistent policies of major academic publishers regarding AI-generated figures, identifies key concerns such as reproducibility and authorship, and proposes transparent best-practice guidelines to ensure the ethical and effective use of these tools in scientific communication.
Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
Imagine you are a scientist who has just made a groundbreaking discovery. You have the data, the experiments, and the "aha!" moment. But now, you need to tell the world about it. In the academic world, you can't just write a story; you need to draw a picture. Specifically, you need a scientific figure—a diagram, a chart, or a visual map that explains your complex idea.
Traditionally, making these pictures was like trying to build a custom house from scratch using only a hammer and a chisel. You needed expensive software (like Adobe Illustrator), hours of time, and the artistic skills of a professional architect. Many scientists, especially those just starting out, struggled with this "construction phase," slowing down how fast they could share their discoveries.
Enter the "AI Architect."
Recently, a new class of tools (Generative AI) has arrived. Think of these as super-smart, instant architects. You can say, "Build me a diagram showing how a virus attacks a cell," and in seconds, you have a beautiful, professional-looking image. One of these tools, called SciDraw, is like a specialized architect who only builds houses for scientists, knowing exactly what the blueprints should look like.
However, this new technology has caused a bit of a panic in the "Academic Neighborhood" (the world of journals and publishers). Here is the simple breakdown of what the paper says about this situation:
1. The "Rulebook" is a Mess
Right now, every major journal (like Nature, Science, or Cell) has a different rulebook for these AI architects.
- The Strict Landlords: Some journals (like Cell Press) say, "You can use AI to draw a cartoon of your idea, but you absolutely cannot use it to draw your actual experimental data. If you try to fake a photo of a microscope slide with AI, you're out."
- The Chill Landlords: Others (like PLOS) say, "Go ahead and use AI, just be honest about it. We trust you to do the right thing."
- The Common Thread: Everyone agrees on one thing: AI cannot be the author. You can't put "Robot 3000" on the list of people who did the work. The human scientist is still the boss and takes full responsibility.
2. Why Are They Worried?
The publishers have three main fears, which the paper explains clearly:
- The "Magic Trick" Problem (Reproducibility): If you ask an AI to draw a picture, it might draw a slightly different picture if you ask it again five minutes later. Science relies on being able to repeat experiments. If the picture was made by a "black box" that changes its mind, how can other scientists verify it?
- The "Fake Photo" Problem (Misinformation): AI is great at making things look real. The fear is that a bad actor could use AI to create a fake photo of a DNA gel or a microscope slide that looks real but is actually a lie. This is like a forger painting a fake masterpiece; it looks real, but it's a fraud.
- The "Who Owns This?" Problem (Copyright): AI learns by looking at millions of existing images. Some of those images might be copyrighted. If an AI draws a picture based on a copyrighted image, who owns the new picture? The scientist? The AI company? The original artist? It's a legal gray area.
3. The Solution: "The Honest Builder"
The paper argues that we shouldn't ban these tools. Instead, we should use them responsibly. The authors propose a set of Best Practices to keep science honest while using the speed of AI.
Think of it like this: If you hire a contractor to build a deck for your house, you still need to inspect the work.
- Tell the Truth (Disclosure): You must put a sign on your house saying, "This deck was built with help from an AI tool." In a paper, this means writing in the "Methods" section exactly which tool you used, what you asked it to do, and when you did it.
- The Human Inspector (Quality Control): You cannot just hit "print" and send it. A human must look at the AI's drawing and check: "Does this molecule actually look like that? Is the text spelled right? Does this chart match my data?" If the AI makes a mistake (like drawing a cell with the wrong number of legs), the human must fix it.
- Keep the Receipts (Record Keeping): Save the "prompt" (the exact words you typed to the AI) and the settings. If someone asks, "How did you make this picture?" you should be able to show them the recipe.
4. The Verdict
The paper concludes that AI is like a powerful new power tool in the scientist's workshop. It can cut wood (create diagrams) faster than a hand saw ever could.
- Good use: Using it to draw conceptual maps, explain complex workflows, or make pretty "graphical abstracts" that grab attention.
- Bad use: Using it to fake experimental data or to draw a picture of a result you didn't actually get.
The Bottom Line:
If scientists use these AI tools with transparency (telling everyone they used them), humility (checking the work carefully), and integrity (never faking data), then AI can help science move faster without breaking the trust that the whole system is built on. The future isn't "Humans vs. AI"; it's "Humans with AI," building better pictures to tell better stories.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.