Science communication does more than explain science — it creates connections between people. Where does AI fit into that relationship?
Read about current trends in science communication and science communication-related activities in the SciComm Close-Up, a monthly deep dive into science communication from the University of Maryland School of Graduate Studies’ Science Communication (SciComm) certificate program. The SciComm Close-Up is part of the program’s monthly newsletter "SciComm Spotlight." Sign up for monthly updates on SciComm.
In the past three to four years, I’ve been on the job market twice and participated in countless job interviews. One of the interview questions I started dreading, but was bound to get, was some variation of “tell us how you use AI to accelerate your work in science communication.”
I don’t doubt I was passed over by some potential employers because of my response to this question. I would try not to audibly sigh and roll my eyes, and to answer honestly. I would explain that, despite being tech-savvy and forward-thinking, and of course using AI to the extent that we ALL do now in the course of using everyday web tools and searches, I’m not a proponent of heavy AI use in my science communication.
One interviewer asked me what my favorite AI tool was in my work, and I was honestly stumped. I was like, well, I just canceled my subscription to ChatGPT because, well … that doesn’t need an explanation. And I avoid most AI features when I can. I don’t know what you want me to say … Google AI summaries?! lol.
Social Media and Science Communication
It’s a strange place to be, because I’ve always considered myself an early adopter when it comes to tech in science communication. I started my first science blog (actually, it was a newsletter for a hot minute first) in 2011. Shortly thereafter, I started using Twitter (forever Twitter in my mind, not X) and became a huge proponent of scientists using social media to share their science, results in progress, the scientific process, and their lives as scientists. I was adamant that all scientists could benefit from engaging on social media.
That is, until we were several years into the growth of the social media landscape, and researchers and practitioners alike realized that online social networks were no panacea for science communication. Social networks can isolate people in informational “bubbles,” making it harder than we first assumed to reach truly new audiences with science. Social media also tends to suppress minority voices, as do many communication mediums, and comes with mental health and social isolation pitfalls. The impacts of social networks really depend on how you use them, and they can be both positive and negative.
The same is true for AI. But perhaps I’ve been more aware (and outspoken) about AI's downsides because I did experience that shift in perspective as I adopted social media for SciComm early in my career. Of course I’m still a heavy social media user, but I’m much more aware that its impacts are not all positive (the rapid spread of misinformation being just one example). And that there are plenty of reasons to opt out of using social media, or limit its use.
Impact of AI on Fundamental SciComm Practices
But why exactly am I so cautious about leveraging AI for science communication in particular? Classic arguments for caution revolve around inaccuracies, hallucinations, and potential misinformation. Understandably so. But let’s say those concerns can ultimately be overcome, and AI content is human-checked for accuracy. Then it’s fine, right?
Well, I’m increasingly AI-hesitant for many other reasons, even if AI-generated or AI-assisted content is accurate and useful.
For starters, I’m a communications skill junkie. I love hands-on work: writing, photography and videography, illustration, fine-tuning videos and graphics. It keeps my brain engaged and challenged, and I’m always eager to gain a new creative skill. I’ve learned enough about myself in 42 years to know my passion is on-the-ground creative work. I don’t crave moving up the chain to primarily manage others rather than having a hand in the creative work myself (ideally, I’d do plenty of creative work while collaborating with other creatives and mentoring others).
Growing evidence suggests that offloading cognitive or creative tasks to AI reduces learning and new skill development, at least for those offloaded. I am looking to build, grow, and expand my hands-on and creative SciComm skills. It makes no sense to me to offload and let them languish. But I also ENJOY the creative work, and even the tedious edits, especially when I’m working on things I really enjoy and find fascinating, like my science podcast. I feel like the more I fine-tune, tweak, and fiddle with the content to get it just right before it goes out, the more of “me” is in the content and the prouder I am of the result.
Ethical Concerns around AI
Another big reason for caution involves the ethical quandaries associated with generative AI, how it is trained, and how it affects the livelihood of current creatives. It’s difficult to articulate how much I appreciate and love science art, and the artists who take science as their inspiration. I absolutely believe art, and particularly the process of creating art, can be transformative. I think every scientist should, at some point, collaborate with an artist and perhaps also create their own art related to their science. Looking at scientific questions and findings with an artistic lens, I think, can deepen one’s understanding of the science and what it might mean for other humans. I think doing art and/or collaborating with artists can literally expand one’s mind and reveal new perspectives on scientific findings.
All that is to say, I think we too often undervalue the work of artists and creatives (and miss out on what they can bring to science). That’s already a problem. But using generative AI to circumvent a human creative professional in creating visual science content just puts this problem on steroids. Not to mention the fact that today’s visual GenAI tools have been trained on existing artwork and creative content without permission. Unless it’s a professional creative responsibly using AI tools to expedite their art-making workflow, I honestly don’t want to see it.
SciComm Storytelling and Trust
So yes, inaccuracy, cognitive offloading, and undervaluing the role of creative professionals in science communication are all very relevant concerns when it comes to AI for SciComm. But perhaps most fundamental to my AI-for-SciComm hesitancy is my concern over how it impacts human relationships, mutual understanding, and trust.
There’s something so inherent about storytelling that humans have been doing it since we first developed complex language. From ancient oral storytelling traditions to prehistoric visuals on cave walls to modern media, storytelling is so core to being human that it’s hard to imagine one without the other. So core that we are teaching our machines to tell stories, and that becomes a signal to us that our machines have become intelligent. But what does a story mean, and what is its purpose, without a human storyteller? Without the person whose perspective, life experiences, internal struggles, and insights we are (even subconsciously) craving when we get pulled in by a good story?
“Humans tell stories to predict surroundings, understand agents and events, and learn patterns of situations and personalities for cooperation.” – The Ancient Art of Storytelling, University of California Television
Science has long been plagued with a “trust” problem (or at least a “warmth” problem). Scholars often argue over whether this is a glass-half-full or glass-half-empty situation. Scientists do generally enjoy decent levels of public trust — the problem is that many people just don’t really know if scientists share their values, perspectives, and concerns. They aren’t sure whether most scientists are particularly altruistic, friendly, and driven by public good. Frankly, most people likely just don’t know a scientist at all (or at least don’t know that they know a scientist). How can you trust someone you don’t really know, understand, and empathize with? The answer is, you can’t. That is why relationship-building matters if we want society to implement science-led solutions in everyday life. And good relationships, well, they form between people willing to wear their hearts on their sleeves, so to speak. Machines? Yeah, they don’t count here.
Maybe you see where I’m going. Trust is critical for effective science communication. Trust requires relationships. Relationships require humanity, a willingness to open up and be vulnerable. To me, the biggest question is this: Why would we give over such a critical function as trust and relationship building to AI? And yet this is precisely the function that the best science communication serves.
Final Thoughts
I suppose one might ask, OK, how human is human? Yes, science communication is about storytelling and relationship building, but exactly how human do you have to be to make this happen? At what percentage of AI involvement in a creative task does the process and outcome become less than fully “human”? At what point does an AI-assisted communication product fail to really hear the other person and communicate your true and honest knowledge, insights, perspectives, findings, struggles, and joys? It’s hard to say.
We don’t see art created with handheld art utensils, from paintbrushes to Apple Pencils, as anything less than human art. We don’t see a blog post typed on a machine, as opposed to being handwritten, as anything less than fully human-created. (Although I suppose at some point in history, maybe people did). Most of us don’t take issue with spelling auto-correct. Surely there are ways to use AI in science communication that don’t take anything away from human self-expression and the connection between the communicator and audience. That don’t erase the influence of human touch and emotion in the final communication product or artwork. But the line is increasingly fuzzy and contentious.
Let’s just say, I’d rather be generously on the better side of that line. What about you?
At some point, the tools (aka AI) become so heavy-handed that everything starts looking the same … and I think we can all agree that we are worse off when that happens.
Paige Brown Jarreau, PhD, is a science communication researcher and practitioner with a background in biomedical engineering and mass communication. Her work focuses on the intersection of science communication and new media, and she has held communication roles across academia, health, and research organizations. Now a consultant and freelance science communicator, she creates science stories across formats ranging from articles and newsletters to infographics, videos, and comics.