“The most underestimated skill in science is imagination. Give yourself the freedom to fail — because every failure teaches you how to succeed.” As AI continues to reshape the future of medicine and scientific discovery, conversations about research integrity, transparency, and collaboration have never been more important. Against this backdrop, we had the honor of interviewing Dr Magdalena Skipper, Editor-in-Chief of Nature Magazine, during her visit to Tsinghua University, where she shared her insights on the evolving landscape of AI in medicine and research publishing.

Q1. What will high-quality AI medical research look like in the future?

Magdalena Skipper :

We often focus on the differences between AI-driven research, research that uses AI, and research that develops new tools. But fundamentally, the same principles of high-quality and high-impact research apply to both AI-based and traditional scientific approaches. Every study must be conducted carefully and address questions that genuinely need new solutions. What’s especially valuable in research is the ability to ask questions in innovative ways and to find creative approaches to long-standing problems. Above all, research quality remains the foundation that underpins all scientific work — and this holds just as true for AI-related research. Such work must be robust and transparent to demonstrate its credibility.

In the case of AI, this means clearly defining the data used for training, specifying how a model or tool was benchmarked, and disclosing all relevant parameters. In essence, the standards are the same; but because AI is still new and being applied so broadly, it naturally attracts greater scrutiny and attention.

Q2. As Editor-in-Chief of Nature, what responsibility do journals have in setting standards for medical AI research?

Magdalena Skipper :

I like to think of journals in the context of the scientific communities they serve — that’s really important. Your question might suggest that journals dictate standards and principles that the community then follows, but in reality, it’s much more collaborative. Editorial policies are often developed in response to evolving practices within the research community. Sometimes, we observe trends that require guidance or correction, but the core idea is collaboration. Journals, editors, and publishers are integral parts of the broader research ecosystem, and open communication is essential.

Through that dialogue, we can shape editorial policies that help researchers demonstrate the robustness and transparency of their work. In turn, this ensures that what we publish maintains high quality and contributes meaningfully to advancing future research. Journals certainly play a key role — but it’s a shared one. We’re most successful when journals and researchers work together to uphold these principles, allowing both rigorous science and impactful discoveries to thrive.

Q3. AI is entering scientific writing and peer review. How do you see its evolving role in publishing?

Magdalena Skipper :

AI is increasingly touching nearly every aspect of our lives, and as you mentioned, it’s being used widely across research. For many researchers, especially those for whom English is not a first language, AI tools can already help in writing, structuring, and polishing papers—and I can imagine this becoming even more common. Eventually, AI may even be able to take experimental results and construct coherent scientific narratives from them.

In peer review, I think AI’s near-term role will be in very technical areas. Take clinical trials, for example—these studies often involve specific statistical analyses, which are sometimes reviewed by specialist statisticians. In the future, AI could assist or even handle such narrowly defined aspects of review.

That said, at Nature, we deeply value the human perspective of our reviewers. Senior reviewers bring a broad understanding of the field and can assess a paper’s significance within the larger research landscape, while early-career reviewers often contribute deep technical insight. Their complementary perspectives are essential.

So, AI tools are indeed useful—for writing support, for technical checks, even for parts of peer review—but they remain tools. They cannot be listed as co-authors because they cannot take responsibility or accountability for the work.

And perhaps we should rethink a familiar phrase: rather than “human in the loop,” maybe we should say “AI in the loop.” After all, humans should remain at the center, with AI serving as a supportive instrument to help them achieve their goals.

Q4. Which areas of AI research do you see as most groundbreaking or impactful for young scholars to pursue?

Magdalena Skipper :

I’m always a little cautious with questions like this — predicting where the greatest gains in knowledge or benefit will come from is never easy, because no one can truly foresee the future. But I’ll answer it in two ways.

First, I think the medical applications of AI are among the most successful examples of AI implementation today. They’re not only advancing research but also providing real, practical solutions on the ground. For example, in radiology, the use of AI for diagnostic support has already achieved remarkable, tangible results. So almost any direction pursued within AI-driven medicine holds enormous potential and value.

That said, AI tools — like any technology — can have both positive and negative effects. One positive direction we should focus on is using AI to promote greater equity in healthcare. There is a real risk that access to advanced AI systems could deepen existing inequalities, but there’s also a tremendous opportunity to design these tools to identify and help close those gaps — ensuring fairer, more inclusive access to medical care.

Another fascinating opportunity lies beyond medical AI itself: can AI help solve its own sustainability problem? We don’t talk enough about the environmental cost of AI systems — their high energy and computational demands, and even the large amounts of water used for cooling hardware. These factors make AI development environmentally expensive. Exploring how AI could optimize its own efficiency and reduce its environmental footprint would not only be a major technological breakthrough but also tie directly back to health, since environmental sustainability and public health are deeply interconnected.

At Nature and across the Nature family of journals, we are now seeing a growing wave of submissions in the field of AI and medicine — reflecting both the excitement and the urgency of exploring these possibilities responsibly.

Q5. What key criteria do Nature editors prioritize when evaluating medical AI studies?

Magdalena Skipper :

The same principles apply as for any scientific paper: clarity, innovation, and transparency.

A strong paper clearly states the problem, why it matters, and how the study advances knowledge—whether conceptually, practically, or methodologically.

Transparency about data and methods is essential. While some medical data must remain private, we encourage maximum openness and clear explanation.

We’re particularly interested in explainable AI and studies that advance equity, reproducibility, and cross-context validation.

Q6. From your perspective, which areas of research integrity and publishing practice should Tsinghua Medicine prioritize as it builds an AI-driven academic health system?

Magdalena Skipper :

I’ll take this question from a broad perspective — and I think the second part of your question really captures the essence of it. When embarking on a major research endeavor, it’s important to think big — to consider not only publishing and sharing findings, which is of course essential for others to build upon, but also to remain grounded in why you’re doing the research and how it will truly advance the field.

Collaboration, particularly at the global level, is absolutely vital. Setting international standards and working across borders creates real synergy, as different partners bring distinct strengths and perspectives. There’s a tremendous opportunity here for Tsinghua to take a global leadership role, engaging diverse collaborators around the world. I was struck by a point made earlier at the conference — that while China has long focused on partnerships with the Global North, it’s now increasingly collaborating with its neighbors in East and Southeast Asia. That shift toward regional as well as global engagement is incredibly meaningful.

In medical AI, collaboration also means multidisciplinary integration. This field sits at the intersection of medicine, computer science, software engineering, and AI — but it must also draw from the social and behavioral sciences. After all, the adoption of AI tools depends not only on doctors but also on patients — on whether they trust and engage with the technology. Achieving truly impactful solutions requires weaving together these diverse threads.

Finally, collaboration must extend across sectors. Many of the most significant recent AI breakthroughs have emerged from the private sector, which often has greater access to computational resources. Rather than viewing this as a divide, I believe academia and industry should embrace collaboration — combining academic rigor with technological capability. True partnership means harnessing the best that each side has to offer, creating something far greater than any one group could achieve alone.

So, in short — collaboration, in the truest and broadest sense of the word — across disciplines, across sectors, and across borders — is the key to advancing medical AI and translating innovation into real-world impact.

Q7. What advice would you give young researchers hoping to make meaningful contributions at the intersection of innovation and healthcare?

Magdalena Skipper :

One of the key pieces of advice I would give is to the younger generation — those who are native to emerging technologies — to educate my generation about the value of these new approaches.

Sometimes new technologies don’t advance as quickly as they could within the proper ethical, legal, and policy guardrails, because the people in decision-making positions are often not the native adopters of these technologies.

For the younger generation, especially students now coming through university, who often understand these tools better than their professors or senior professionals, it’s worth thinking about your role as a teacher in this context — as an explainer.

As a student you have a lot to learn, but you also have a lot to offer. This kind of partnership can be powerful. You may open doors that you thought were closed, or perhaps didn’t even know existed.

Q8. What is the most underestimated yet essential skill for early-career researchers today?

Magdalena Skipper :

I would say the most underestimated skill — and one that I would recommend young students and researchers really tap into — is their own imagination. Focusing on what could be or what might be possible is incredibly important.

This also connects to something I often think about, which isn’t discussed enough in science or in other walks of life — the importance of failure. We all have a natural aversion to failure. We want to avoid it at all costs, and when it happens, we prefer not to talk about it. But in reality, there is no success without failure.

In research especially, exploration often means moving in one direction, stopping, retracing your steps, and trying another approach. Many of these attempts could be seen as failures, but they’re immensely valuable because they teach you something new and prepare you to find the right path forward.

This ties back to imagination — the more freedom you give yourself to experiment, to try different ideas, directions, or projects, the more you’ll learn from each experience. You will fail at many of them, but those lessons will ultimately guide you to real success and help you understand how to sustain it.

Q9. Tomorrow, you’ll deliver a lecture titled “Nature: A Mirror of Research and the Scientific Community” at Tsinghua University. Why did you choose this theme?

Magdalena Skipper :

I’m very much looking forward to my visit tomorrow. This is a theme I began working on about six years ago, when Nature celebrated its 150th anniversary. It was a wonderful opportunity to look back at the journal’s history.

As I reflected on that history, it became clear that Nature itself has evolved enormously — in its formats, scope, and what we publish. Although it has always been a broad, general-scope journal, that breadth has expanded over time. Today, we’re especially focused on making the most of our multidisciplinarity. The format of our papers, the kind of journalism and opinion content we produce — all of this has evolved hand in hand with the evolution of science itself.

The way we conduct research today — whether clinical, applied, or fundamental — was largely shaped in the 20th century, particularly after World War II. Before that, research looked very different. It was funded differently, and the idea of publicly funded research as a common good didn’t really exist. The structures to support it simply weren’t there. From that point onward, a new set of expectations emerged about how science should be conducted.

By examining Nature’s own editorial practices and policies over time, we can actually trace how science itself has evolved. For example, only relatively recently did we begin to fully recognize the importance of diversity in research — particularly in human studies.

As a geneticist, I’ve seen this firsthand. While humans are closely related evolutionarily, there are meaningful differences among populations — in disease susceptibility, drug responses, and adverse effects. There are also important gender differences. Yet for a long time, this wasn’t acknowledged. Clinical trials, for instance, were conducted almost exclusively on white men — an approach we now understand was far from optimal.

Today, our editorial policies explicitly address these issues. We require authors to explain, where appropriate, why data were collected or analyzed in certain ways, whether analyses were disaggregated by relevant variables such as sex or population, and if not, why not.

We also know that diverse teams produce better research. In collaborative projects, especially those involving partners from both well-resourced and less-resourced settings, we have policies to help prevent so-called “helicopter research” and promote true partnership.

This is why I find this theme so compelling — it perfectly illustrates the vital partnership between scientific journals and the broader scientific community. The real strength lies in that collaboration and in working together to advance science responsibly and inclusively.

Q10. What advice would you give to students considering whether to pursue university studies or a PhD in science?

Magdalena Skipper :

At any stage of life, it’s important to do what gives you satisfaction and allows you to make your best contribution.

That contribution can take many forms. It might be something on a small, local scale — a commitment to your community that brings real value and meaning. Or it might be on a much larger scale — leading international scientific projects, for example — which is equally significant in a different way.

Sometimes it’s difficult to know exactly what you want to do, especially at an early stage. But if you have an appetite and curiosity for knowledge, and a desire to contribute to extending that knowledge and making the world a better place, then I would strongly encourage you to pursue it.

If you have the opportunity to attend university and especially to undertake a PhD, to do research — take it. It’s an extraordinary privilege to be able to engage in research, whatever the field or setting.

At the very least, if you have that curiosity and are considering it, it would be a missed opportunity not to give it a try. You can always change paths later if you discover it isn’t right for you. But not taking that opportunity at all would truly be a missed chance in itself.