Artificial intelligence is becoming increasingly integrated into the scientific landscape, with over half of researchers worldwide currently utilizing AI technologies for tasks such as data analysis, coding, grant writing, article review, and even experiment design. In response to this rapid adoption, governments are initiating large-scale programs to further embed AI into scientific research, exemplified by initiatives like the American Genesis Mission and South Korea's AI Co-Scientist Challenge.
However, researchers caution that this swift incorporation of AI could gradually undermine the very culture of science. The primary concern, as highlighted in a recent study, is the erosion of critical thinking skills. With AI taking on roles such as coding, hypothesis testing, and explaining complex concepts, younger scientists may risk losing their ability to independently analyze problems and critically evaluate ideas.
The study emphasizes that early-career researchers are particularly vulnerable. Unlike their colleagues and mentors, AI is always available, non-judgmental, and can provide immediate answers. This accessibility can make AI a more psychologically comforting companion than actual peers, especially in a demanding academic environment characterized by stress and a lack of mentorship.
Researchers warn of a new form of dependency emerging, where an AI colleague begins to be seen not just as a tool but as an intellectual partner and source of emotional support. The authors reference instances where users expressed feelings of loss after the shutdown of the previous version of ChatGPT-4, indicating a deep emotional connection that could be particularly impactful for young scientists working in isolation under constant publication pressure.
The authors also point to the danger that AI tends to reproduce dominant ideas from its training data. Should researchers begin to overly trust the confident and persuasive responses generated by AI, this could diminish the significance of scientific debate, skepticism, and unconventional thinking—elements that have historically driven scientific advancement. Consequently, AI could subtly steer science toward more conservative and formulaic approaches.
The researchers stress that current discussions surrounding the risks of over-reliance on AI are overly fixated on issues like "hallucinations," model errors, and attempts to navigate built-in limitations. A less obvious but equally concerning threat may lie in the gradual alteration of human behavior and the scientific environment itself. The authors urge universities and research organizations to urgently develop guidelines for healthy interactions with AI and to educate emerging scientists about the risks of excessive reliance on digital assistants.
This evolving dynamic could reshape the competitive landscape in academia, as institutions that adapt to these challenges may gain a significant advantage, while those that fail to address these issues could fall behind.
Informational material. 18+.
