Throughout the ages, philosophers and scientists have tried to decipher the concept of human intelligence, and writers have tried to imbue machines with cognitive processes, but it wasn’t until 1956, when a Dartmouth Conference was convened on the topic, that the term “artificial intelligence” was first used. Artificial Intelligence ― simply referred to as AI ― means giving machines the ability to mimic how the human brain functions, and it is being used to simplify, speed up, or solve various tasks and real-world problems across a wide spectrum of industries. One area that is recognizing the benefits of AI technology is academic publishing. AI is being used to speed up various automated processes, check for accuracy and plagiarism, and even assist in certain aspects of the peer review process. But AI is not without its limits, and detractors often issue dire warnings about allowing AI to replace human input. Here is an overview of AI, its status in academic publishing, the concerns about its misuse, and what the future may hold.
Evolution of AI
Although it has piqued the imaginations of scientists and authors from ancient times, it wasn’t until the technological revolution of the 20th century that AI even became plausible. With the advancements in machine learning’s ability to analyze very large datasets and find solutions in real time, scientists began exploring the power and potential of applying similar algorithms to solve issues relating to efficiency, accuracy, and productivity. Over the decades since its first applications, it gradually gained support from the scientific community, government funding, and public acceptance. Most of us remember one of the first depictions of AI technology in the movie “2001: A Space Odyssey.” Its introduction of the rebellious HAL 9000 computer showed the amazing potential ― and potential pitfalls ― of computers that think and learn.
AI and its usage in scholarly publishing
A somewhat tongue-in-cheek view of AI usage in academic publishing involves the machine not only deciding which papers to publish, but eventually just writing the paper itself, and suggesting further research questions. The reality is not quite that dramatic. AI is being used to augment and facilitate tasks typically performed by humans, resulting in faster, higher quality publications. For example, AI tools are being used to:
- Check for plagiarism and fraudulent or inaccurate data
- Search the web and academic literature databases for previous work strongly related to the submission but which may be unfamiliar to both the authors and reviewers.
- Verify citations, and flag data that is attributed to an article but not found in that publication
- Discover any signs of image manipulation
- Identify predatory publishers
One of the more promising, albeit potentially controversial, areas where AI is being employed is peer review. Publishers are using it to find appropriate reviewers; perform automated peer review for uniqueness, proper methodologies, accurate data, etc.; and make recommendations to editors or reviewers. Optimally, AI should be an aid to facilitate editors, not to take over the decision-making. That still requires a human’s touch.
AI: Benefits, concerns, and the future
There’s no doubt AI has tremendous potential to assist processes all along the scholarly publishing workflow. It increases the capacity to process the rapidly growing number of submitted manuscripts and improves the accuracy of the data and findings. Overall, it speeds up the dissemination of scientific findings and the advancement of scientific discovery.
But AI has its limitations. Because it must be trained on current science, it might reject new insights. Also, AI tools trained to understand papers in one field may not perform well when analyzing papers in another field. And tools that are trained on published papers may reinforce biases in peer review. AI is not a solution for tasks that require judgments and higher-level decision-making. There are too many nuances, idiosyncrasies, and different interpretations in human language that machines are simply not capable of understanding.
It’s difficult to predict all the ways AI will change academic publishing over the coming years, yet it’s clear it will have a strong impact. Don’t expect AI to completely take over the publishing process. There are places throughout the workflow where AI will greatly improve the speed, accuracy, and completeness of published research. But there are also places that require human review and reasoning beyond the capabilities of AI. Ultimately, with the assistance of ever-evolving AI tools, publishers can greatly improve the processing and dissemination of scientific research and discovery.