The International Science Council and its Member, the China Association for Science and Technology (CAST), in partnership with Nature, have launched a new six-part podcast series exploring the evolving landscape of research careers. Across the series, early- and mid-career researchers will be in conversation with senior scientists, sharing experiences of growth, collaboration, and resilience in the face of rapid change.
In this third episode, Mercè Crosas, Director of Computational Social Science and Humanities at the Barcelona Supercomputing Center, and Mohammad Hosseini, Assistant Professor of Ethics at Northwestern University, discuss how AI and digitalization are transforming scientific careers and the research process.
The conversation highlights both the opportunities and challenges AI brings for early- and mid-career researchers. While new technologies enable breakthroughs and open up entirely new avenues of inquiry, they also raise concerns around equity of access, over-reliance on automated tools, and the erosion of critical thinking.
Izzie Clarke: 00:01
Hello and welcome. I’m science journalist Izzie Clarke and in this podcast presented in partnership with the International Science Council, with the support of the China Association for Science and Technology, we’ll be discussing the power of the digital aid and artificial intelligence known as AI, its importance to careers in science, as well as its potential threat to the scientific enterprise.
Today, I’m joined by Mercè Crosas, Director of Computational Social Science and Humanities at the Barcelona Supercomputing Center and President of the Committee on Data of the International Science Council, known as CODATA.
Mercè Crosas: 00:42
Hello.
Izzie Clarke: 00:43
And Mohammad Hosseini, Assistant Professor of Ethics at Northwestern University in Chicago, and member of the Global Young Academy.
Mohammad Hosseini: 00:51
Hi, how are you?
Izzie Clarke: 00:52
Very well, thank you. I think a question to both of you, to start things off, is why is now a critical moment to reflect on how digitalization and AI are shaping scientific careers?
Mohammad Hosseini: 01:05
I think we are seeing more and more data-driven decision-making by researchers, which sometimes also trickles down to national or local decision-making, which is good, but in terms of scientific careers, this means that we need to train researchers in new skills.
And this has always been the case. But because of the tipping point, things are moving so fast that we can hardly catch up. Machines are becoming so capable that they can displace or replace human workforce in science. We are now in a sort of critical moment to discuss digitalization and explore who benefits from these technologies, who may be left behind and how we can ensure transparency and equity in their use.
Izzie Clarke 01:54
Mercè, what are your thoughts?
Mercè Crosas: 01:56
One of the things first is that AI also has been used in science already for quite a long time, and the change has been happening progressively. It is true that now there is an exponential use of AI used as for methods in a lot of the scientific production.
So, from the exploration of the literature review to trying to figure out the research question, to data processing and data collection, and then the analysis itself, but also then the publication of the scientific results. I mean, I guess that tipping point that Mohammad was talking about, it has a much broader impact than ever before.
Izzie Clarke: 02:34
There’s a lot of things to consider here. You mentioned publication there and we will get onto that in a moment. But in terms of opportunities, what are the opportunities that you see emerging from this for early- and mid-career researchers and how that is changing that AI-driven scientific landscape?
Mohammad Hosseini: 02:55
I guess opportunities are mostly around making new discoveries and doing things that would be even a dream five years ago. Any area that could benefit from modelling, we are moving much faster now. This is an opportunity, especially for early- and mid-career researchers who may be more adept in using AI, but it comes with certain trade-offs. Finding opportunity in this new dynamic requires a new kind of curiosity that we are not trained in. But I think we should try to find tasks in research contexts that cannot be automated and try to excel in such tasks.
For example, my area of research, I’m an ethics researcher. Writing a well-argued paper is already automated. But mentoring, teaching an in-person class, which is also interactive and engaging, or conducting interviews to collect data and get new insights from people’s lived experiences — these are tasks that cannot be easily automated. And I think we need to find these group of tasks in our own research context and try to excel in that.
Izzie Clarke: 04:06
And Mercè?
Mercè Crosas: 04:07
I don’t see much the risk of scientists or early-career scientists, mid-career scientists, to be substituted. What I see is opportunities to new research questions that a lot of scientists from previous generations couldn’t even think of asking, right? So, no, it’s not so much just that, well, now we can apply these tools, but that we can think about some fields in a whole different way. In biomedicine, in climate change, in physics and biology for genetics, that can change with the use of AI and new types of data.
Izzie Clarke: 04:39
I think we are seeing that there are a lot of different ways that we can turn to AI and tackle different tasks, and we’ve talked about re-skilling. So, what do you think early- and mid-career researchers in the scientific fields need to be mindful of, and where can they get support?
Mercè Crosas: 04:57
It’s more important than ever to be very rigorous in science and to understand that, at the end, whether we use AI or we use other tools, science is what we do, and science is inference and science has to be public. The methods, the data and the way we do it has to be verified by others.
It means that, again, we don’t just use the AI tools to give us answers, but we need to become more specialists in how we validate those answers. And for that, we need to still be more prepared about the theory of the fields where that we do research and the rigorosity of the outputs.
Izzie Clarke: 05:33
Yeah, I mean, Mohammad, I’d love your thoughts on this as well because I know that this is something that you pay a lot of attention to.
Mohammad Hosseini: 05:38
Yeah, absolutely. And I also want to go back to what Mercè said here. Yes, it is important to think about theory, and at the same time, there’s a lot of people who now argue that because of this rise of data-driven science, we are seeing the end of social theory. Theory is not really as important because people can just collect data and do data mining to see what is relevant without even having had a hypothesis prior to their data collection.
And I think that’s a remarkable development that requires a lot of careful consideration and attention. I think one of the challenges I also want to highlight is the fact that we have access to different resources, depending on location. We also have disparities in terms of what institutions provide. I have the privilege to be based in an affluent private university in the US that offers free access to various AI models, but this is not the case for millions of other researchers.
And this disparity puts many other people in a disadvantaged position. Many universities don’t even have a general policy for the use of AI models. If I was in such university, I would really try to speak with the university administration or library to ask them to provide guidance and training.
Mercè Crosas: 06:54
To follow up on the danger of becoming too data-driven. I don’t accept that that’s the way that we need to go, right? The results is the intersection between the theoretical model and this data-driven approach. But in terms of using generative AI or new types of AI tools, I think that Europe has pretty different approach than other places.
And there is now undergoing the development of a new strategy of AI in science and science for AI. We need to be careful about what kind of AI tools we use, whether they have clear definition of what data has been used, whether they are open source, whether they focus on trustworthy AI, and I think that’s very important.
Izzie Clarke: 07:36
I wanted to pick on something there as well. We talk about how we are using AI in work and publishing, as well. So Mohammad, what are the things that you think early- and mid-career researchers should be mindful of when it comes to publishing and the use of AI?
Mohammad Hosseini: 07:54
Yeah, I think one of the things that we should be really mindful of is what is the task that we are offloading to AI? What is the task we are asking AI to do? When this AI boom began, AI was mostly being used at the end of your research process, like at the point of copy-editing and improving readability and so on.
But now we are offloading these important tasks to AI, and next time when you want to think about your next research question, instead of thinking deeper about the textbooks you read or the new articles you read, you’re like, ah, let me ask what AI has to say about it. It becomes very addictive, and I would encourage researchers to be aware of the tasks that they are delegating and ask themselves, is it worth it?
My suggestion is don’t just publish something for the sake of publishing something unless you have something really important to say. Think about who are you citing. If you’re using AI to find literature, make sure that you read the content that you are citing, because many times these citations are irrelevant.
Izzie Clarke: 09:03
And I think that’s a good point. Yes, there are ways that we can use AI that might be helpful in some points, but keep some of those skills active and to make sure that you are doing due diligence in other ways, as well.
And I think that probably brings us onto a discussion on credibility. So, within your field and to the wider public, what does it take to maintain credibility in this digital age? Mercè?
Mercè Crosas: 09:30
Well, I think it’s very easy. I mean, you had credibility when you can communicate it, when you fully understand it and what you’re working on and it’s not been generated by something else that you don’t understand. Going back to the values of science and open science, that it is as transparent as possible, that anybody else can verify what you’ve done from how you have applied the AI model, the method, the data that you’ve used, the workflows, fair principles for findable, accessible, inter[operable], reusable data. But also software so that what you’re using is shareable, is findable by others and can be verified.
Izzie Clarke: 10:06
But there are lots of exciting ways that this can be a tool for transforming science and digitalization, as well. So, Mercè, how do you see the role of science communication growing as technology grows, as well?
Mercè Crosas: 10:20
Well, so, science communication, we still need to do a lot of work on that for society. And there are already expectations that are possibilities or opportunities for AI to play a role also in helping summarizing a lot of the science output and make it more accessible to broader audience. So, I think that can be interesting.
Izzie Clarke: 10:41
And finally, what gives you both hope for the future of science in this digital world? Mohammad?
Mohammad Hosseini: 10:47
I think what gives me hope is a new generation of researchers who speak up. We are observing a new generation who dares to say what it thinks and is willing to pay a price for it. I’m in the US and I see all kinds of big companies and how they can influence the research landscape and universities and all of that. So, it’s very important for me to see that.
Izzie Clarke: 11:12
And Mercè?
Mercè Crosas: 11:15
So, I think that we have more tools to understand how we work, how we collaborate, what new questions we can ask in science. And I think that gives hope for better science if we don’t lose what science is and we don’t lose these values of open science, but also taking advantage of this new type of AI methods.
Izzie Clarke: 11:34
Thank you both so much for joining me.
If you’re an early- or mid-career researcher and you want to be part of the conversation on the future of AI, join the International Science Council Forum for emerging scientists.
Visit: council.science/forum to find out more.
I’m Izzie Clarke, and next time we’ll be discussing how early- and mid-career researchers can help protect our ocean and the power of a transdisciplinary approach to do so. Until then.
Read more about Science without borders: International collaboration for the next generation of researchers 
Read more about Rethinking growth and resilience in the research ecosystem 
Read more about Transdisciplinary science: Early- and mid-careers shaping the future of science 