Skip to content

[ Publishing Workflows ] February 25, 2023

Power, potential, and questions: Artificial intelligence in scholarly publishing

There aren’t concrete answers to the question: how will machine learning impact scholarly publishing? Instead, that one question leads to a thousand more, and those questions lead to even more (so many questions that perhaps the best way to categorize and synthesize the sheer volume of queries is with an AI algorithm).

Chatting with ChatGPT

ChatGPT, and other AI language models, synthesize information to provide comprehensive answers. We thought we’d join the ranks of articles asking ChatGPT about its perspectives on the future of AI in scholarly publishing.

As we’ve seen elsewhere, ChatGPT often begins responses to ethical or opinion questions with phrases like: “As an AI language model, I do not have personal beliefs or values.” It takes this further when asked about what it thinks about AI authorship: 

"Overall, while AI authorship has the potential to be a useful tool in certain contexts, it is important to carefully consider the ethical implications and limitations of this technology before embracing it as a solution for content creation. Additionally, it is important to recognize the value and importance of human creativity and nuance in writing, and to not rely too heavily on AI-generated content as a substitute for human writing."

Despite this call for nuance, trade publishers are seeing a sharp rise in AI-generated pitches, and scholarly publishers are seeing some of the same

We’re seeing a massive influx in experimentation with AI to test its limits, its possibilities, and more, from plagiarism detection to semantic tools that summarize research articles in a sentence. When asked how machine learning might impact scholarly publishing, ChatGPT responded: 


  • Automated peer review

  • Improved manuscript editing

  • Content recommendation

  • Text mining

  • Plagiarism detection

Going into further detail, ChatGPT stated that some of these use cases are already in place. Indeed, these impacts may not seem all that revolutionary on their own, because, as ChatGPT itself shared, “it’s important to recognize the value and importance of human creativity and nuance.”

 

A nuanced review

Many of the discussion on AI today either worries about its potential or celebrates its possibilities. The debate is polarizing: many believe AI will improve every aspect of our lives, while others expect it to destroy and supplant our species. 

Let’s take a nuanced approach, putting on our futurist hats and playing out those arguments from a dystopian and utopian lens. Will we find one scenario more likely, or is our future somewhere in the unknown middle, between the extremes of machine-learning discourse?

 

A dystopian future ruled by machines

As the typical argument goes, machines will steal our jobs and take over our world. What does that look like for scholarly publishing? 

Maybe scientists submit datasets to AI, which then synthesizes, evaluates and draws conclusions from that data. Think about the impact on misinformation and research integrity if AI publishing workflows could eliminate human biases (a promise which we know does not always come to pass). Would AI be able to cut through human bureaucracy and apply research findings to problems such as climate change? Would it only solve problems that also impact the survival of AI, as a dystopia implies?

 

A utopian future of harmony and coexistence

A utopian ideal of a post-AI world could either mean a future in which AI is in charge and solves all our problems, freeing up humanity to lives of leisure and whatever fulfills us individually, or it could mean a future where AI is tightly controlled by humans and only performs set tasks as we desire. 

Focusing on the second scenario, it is all about cooperation. Humans provide the creativity, insight, and inspiration that sparks scientific breakthroughs and innovative ways of publishing those ideas, and AI helps validate, mine massive datasets to help inform future research, and so much more. Both AI and scientist work toward the same goal: sharing high quality information with the world.

 

Somewhere in the middle

In reality, an AI-controlled publishing workflow would likely perpetuate some of the same biases we see in current publishing workflows if given free rein. 

The path forward will likely be full of experimentation, trials, and errors. We’ll develop new regulations as an industry, and we’ll embed new tools that help us make science publishing faster and better quality.

 

Conclusion

Perhaps most importantly, we think the path forward with AI will require an unprecedented level of collaboration, debate, and creativity on the part of the stakeholders in the scholarly publishing industry. 

We will build powerful partnerships to transform our publishing workflows, whether that means embedding AI to streamline the process or regulating the use of AI-authorship. It's easy to be worried about the uncertainty of the future, but with an interoperable set of tools, and a highly engaged community all working to improve our publishing process, there’s no limit to our potential.

Microscope cutting of a cell with messaging about the future of publishing infrastructure