Профессор Пенн-Стейта рассказал, как научить ученых писать убедительно и почему ИИ не заменит автора

UNIVERSITY PARK, Pa. — Scott Milner, William H. Joyce Chair Professor of Chemical Engineering at Penn State, has been writing and publishing across careers in both industry and academia for decades. In his new book, “Writing and Presenting for Scientists and Engineers,” publishing later this year through MIT Press, Milner provided insights into methods for writing about scientific research in a more compelling way. In this Q&A, Milner discussed his motivations for writing the book, tangible advice for those looking to be more impactful writers and how he believes artificial intelligence can — or cannot — replace the human act of writing. Q: What motivated you to write this book? Milner: I've been interested in writing since I was in high school and college, where I worked on newspapers quite a bit. So, I have a somewhat different background than a typical scientist, with a little bit more appreciation of what makes tight prose. I think many scientists and engineers tend to regard writing as this awful chore that you must do at the end of some interesting research. They know they need to effectively communicate their ideas, persuade funders and so on, but they don’t always love that process. Throughout my career, I’ve focused on writing effective papers, and have become known as someone who writes clear papers and gives clear talks. I was in industry for 20 years before coming to Penn State. When I came to Penn State and began teaching graduate and undergraduate students to write research papers, I found myself explaining the same concepts frequently. I often would encourage my students to start with a detailed outline, focus on an engaging introduction and draft a strong title. Another thing I would emphasize is that their audience is not me, but their former selves. Tell your former self what you did and why. For me, each new student is a reminder of common writing mistakes and how to address them. Over time, I developed a list of common writing errors and the desire to collectively address them, which led to a class. I was hesitant to teach something outside of my discipline, but I had considerable energy for the subject. The class became the basis for the book. Q: What can readers expect from your book? Milner: One of the main ideas is that a good paper should preserve the spirit of a good talk, and that writing and presenting are a lot closer than people tend to think. I’ve observed that many experienced scientists have good presenting skills, both in terms of what they put on their slides and how they present them, but that sometimes goes out the window when they sit down and write a paper. Papers tend to be very formulaic. There’s even an acronym — IMRDC: Introduction, Methods, Results, Discussion and Conclusion — that separates methods from results and results from discussion. But this separation is artificial. When you give a talk, you start with a question, and then you describe a first exploration or experiment, which leads to a result that you interpret, which leads to another question and so on. There’s no reason a paper can’t be written in that way. Many people assume IMRDC is a hard rule. But when you come across a paper that really explains how the author was thinking, it feels like a life raft in a sea of confusion. It helps you understand what people are doing in a field. Why not write more like we present? That is an important theme of the book and is why the book discusses both writing and presenting. The book also traces the trajectory of a graduate student or a young researcher; it starts with effective figures and graphs, then slides for a talk and from there discusses presentations. Presenting is performing, which is a reality that many researchers resist. They might think their data should speak for itself; but no, you must speak for your data. Remember that your audience consists of humans, who respond to stories, and you’d like to capture their attention. Q: Does your book encourage scientists and researchers to infuse more narrative and rhetoric into their scientific writing? Milner: I do encourage scientists to write effective prose. You don’t want to get bogged down immediately in details and numbers. Write more like a journalist, less like an academic. Also, scientists use analogies a lot in their thinking. The list of famous scientific analogies is long: the electron cloud, the tree of evolution, the planetary system as clockwork. Analogizing is often what you do when you figure out something new. It behooves you to share those mental images and analogies with your audience. You’re still showing them your evidence, but you’re also giving them an image that helps them structure their own view of the research. Q: What skills are discussed in this book that you had to learn for yourself in the process of transitioning to Penn State and academia? Did you have stumbling blocks during that transition? Milner: Teaching people how to do stuff makes you better at those same skills. In academia, it’s a truism that teaching and research reinforce each other. When I teach introductory courses, I often find that my understanding of foundational topics becomes clearer, which then stimulates new ideas relevant to my research. In a similar way, teaching about writing and presenting has led me to reflect on my own writing process, which has improved as a result. Q: Does your book include anything on artificial intelligence (AI) with regard to writing? Milner: I gave an early draft to a colleague, who pointed out that it didn’t include anything about chatbots. In response, I wrote a chapter about chatbots. I wondered how ChatGPT might handle some of the exercises in the book, so I gave it some specific assignments: summarize a paper, describe a theoretical concept, draft something from an outline, make a piece of writing more or less formal. Its performance was a mixed bag. It’s surprisingly good at summarizing papers. If you ask it about a subject that isn’t widely understood, it can hallucinate a lot. It’s not bad at expanding a very detailed outline; of course, writing such an outline requires a lot of careful thinking. But it has a tin ear for style, and the prose it generates is bloated and bland. I also think AI-assisted writing can be dangerous for a beginning writer. To improve as a writer, you need to know what good writing is, which you learn by working at it yourself. Furthermore, writing clarifies your thinking. Whenever you write, you’re organizing your thoughts: As you “audition” sentences, you often discover that what you thought you wanted to say changes. If you have something original to say, and you outsource to a chatbot the hard work of putting it in the right words, you lose that important benefit. In addition, students and researchers should exercise caution if employing AI in their writing and planning to submit their writing as part of coursework or for publication. Schools and many publishing outlets, including scientific journals, have strict AI guidelines governing appropriate use.