Written By Morgan Rafferty ‘28
Edited By Allison Shea ‘28
Artificial intelligence (AI) is currently the most contested new technology in the 21st century. Its use has some praising the boost in productivity, while others argue that it only makes individuals lazy. This same discussion has reached the field of research, which has scientists questioning if it’s safe to use. While many argue in favor of its application, critics are always ready with a long list of cons to warn against it.
One of the more obvious applications of AI in the research field is during the writing process. According to a recent survey of academic researchers in various career stages, 57% of respondents used AI for writing assistance, 47% used it to remedy errors in writing, and 35% used it to check for possible instances of plagiarism (2). While these numbers may seem surprising to some, individuals in the field see this as no surprise. A follow-up survey asked about the performance of AI in these instances, and 62% of researchers responded that they believe AI can perform writing-related tasks better than humans (2). The high capacity for academic writing among chatbots is promising, as a reduction in the amount of time spent writing allows for more hands-on research to be done. In this way, AI allows researchers to focus more time on wet lab processes.
Many are excited to see that AI can help speed up the procedural process, too. Thirty percent of researchers claimed that they use artificial intelligence to help review a larger range of scientific literature, affording them greater exposure to pre-existing findings and techniques (2). Quantum physicist Mario Krenn at the Max Planck Institute for the Science of Light is a prime example of this. When faced with a procedural issue no one on his team could figure out, he created his own AI algorithm to help him find the solution. After a few hours, the software came up with a solution by suggesting an older, lesser-known technique and applying it to the new situation (3). The possibilities are not just limited to methodology development either. 35% of researchers use AI for data processing and number crunching, saving them hours of painful calculations (2). Researchers have also begun to use AI to analyze qualitative data, as it can sense differences in cell scans that the human eye cannot. This is currently being applied to the study of neurodegenerative diseases, as the process can identify cells that are destined to die (3). Thus, by integrating AI into all facets of the research process, new discoveries can be made at an increased rate.
While these benefits undoubtedly seem promising, others are not so convinced. For one, there seem to be many tasks that humans are better suited for. Respondents said that they believed humans were better at creating new methods and forming collaborations 58% of the time and conducting peer reviews 59% of the time (2). It's also important to note that AI can currently only recycle old ideas, and not generate new ones. A simple solution is to not use AI for these parts of the research process, but the problem lies in the fact that a significant percentage of respondents claim that they already use AI for these same reasons. The dips in quality in these areas thus seem inevitable, but new research shares that the harms could be more than just quality-based.
For one, AI can inadvertently shape our body of understanding. Much like standard search engines, AI’s search responses will focus on more popular works and methodologies, recommending those to researchers first. Research will then be made based mostly on those works, and not studies favoring different outcomes or niche methods. As a result, Messeri and Crockett claim that “scientists incorrectly believe that AI tools advance scientific understanding, failing to appreciate that these tools instead narrow the scope of scientific knowledge production” (4). By creating research based on limited knowledge, the possibility for discovery only shrinks, forcing researchers to choose between fast results or important ones.
Researchers produce work under the guise of the “publish or perish” framework, but is it possible that in their haste they have only managed to both publish and perish? The uses of AI undoubtedly seem promising, but it’s important to recognize that with this gain comes the potential for narrow minded and potentially even false research. In its current state, artificial intelligence is able to produce fast research, but only time will tell if it can produce quality research.
References
Sakana AI creates an “AI Scientist” to automate scientific research and discovery - SiliconANGLE [Internet]. [cited 2025 Feb 22]. Available from: https://siliconangle.com/2024/08/13/sakana-ai-creates-ai-scientist-automate-scientific-research-discovery/
Naddaf M. How are researchers using AI? Survey reveals pros and cons for science. Nature [Internet]. 2025 Feb 4 [cited 2025 Feb 22]; Available from: https://www.nature.com/articles/d41586-025-00343-5
Frueh S. How AI Is Shaping Scientific Discovery. National Academies: Science Engineering Medicine [Internet]. 2023 Nov 6; Available from: https://www.nationalacademies.org/news/2023/11/how-ai-is-shaping-scientific-discovery
Messeri L, Crockett MJ. Artificial intelligence and illusions of understanding in scientific research. Nature. 2024 Mar;627(8002):49–58.