In November 2022, when ChatGPT burst onto the scene, the academic world experienced "the fastest moral panic in human history." Within just seven days of its launch, we went from ChatGPT's debut to declarations that "the college essay is dead." Faculty members found themselves caught between apocalyptic fears and eager adaptation, while institutions scrambled to develop policies in response to this disruptive technology.
Now, over two years into the generative AI revolution, a clearer picture is emerging of how these tools can be integrated into academic research—and librarians are proving to be vital leaders in this transformation.
Despite widespread fears that students would immediately embrace AI tools to circumvent traditional learning, research from Vanderbilt University reveals a more nuanced reality. When surveying science and engineering students, Joshua Borycz, MSIS, PhD, Librarian for STEM Research, and Alex Carroll, MSLS, AHIP, Associate Director of Science & Engineering Library, discovered that prior to receiving instruction, only 11.8% of students reported using AI tools "often" or "very often" for research-related assignments.
Why this hesitation? Many students expressed uncertainty about the ethical boundaries of using these tools, unsure whether their use constituted plagiarism or academic dishonesty. Without clear guidance from their institutions, students often avoided potentially helpful technologies out of caution rather than embracing them recklessly.
In response to this uncertainty, Vanderbilt librarians developed an elegant framework that acknowledges both the potential and limitations of AI tools in academic research. The 5 Is Framework highlights five critical limitations of large language models:
What makes this framework particularly valuable is that it isn't prohibitive. Rather than telling students "never use AI," it acknowledges that students "will, can, and should use AI tools as part of their research workflows," while encouraging selective rather than exclusive usage.
One compelling metaphor used by the Vanderbilt librarians compares traditional information retrieval to a chamber orchestra—highly structured, consistent, and producing the same results each time when given the same inputs. AI tools, in contrast, function more like a jazz band—improvisational, creative, and never playing the same solo twice even when given the same starting point.
This distinction helps students understand when each approach is appropriate. Boolean searches in library databases provide precision and consistency, while AI tools offer creative connections and natural language processing advantages. The key is knowing which "musical ensemble" suits your current research needs.
After implementing this framework in curriculum-integrated instruction sessions across multiple science and engineering courses at Vanderbilt, the results were striking. Following the sessions:
Most importantly, this learning was achieved with minimal time investment—typically just 15 minutes of a standard information literacy session. The framework integrated seamlessly into existing instruction rather than requiring separate workshops.
If you're looking to integrate AI tools into your research workflow and teaching, consider these approaches from the Vanderbilt model:
For Your Own Research:
For Teaching Students:
This case highlights an important evolution in the role of academic librarians. Rather than being displaced by AI, librarians are becoming essential guides to its thoughtful application. As Carroll notes, "For us, it's been a really, really great way to reinvigorate our information literacy practice."
Traditional information literacy instruction focusing on search strategies now has a compelling and contemporary context. Students who might have been uninterested in learning about engineering handbooks or database search techniques are now actively engaged in critical discussions about information evaluation.
For institutions just beginning their AI integration journey, the Vanderbilt librarians offer three key pieces of advice:
As specialized AI tools for academic purposes continue to develop, we're likely to see more robust retrieval augmented generation (RAG) models that can access closed-access literature and provide more reliable citations. Borycz has observed this evolution firsthand, noting that "Scite is the best tool for research that I've seen so far." Unlike general-purpose AI platforms, specialized academic tools like Scite offer superior source attribution through direct access to full-text research articles and publisher-indexed citation statements—addressing many of the "Incomplete" and "Incoherent" limitations outlined in the 5 Is framework."
However, even as these tools improve, the core need for critical thinking and information evaluation skills will remain essential.
The Vanderbilt case demonstrates that even a brief educational intervention can significantly shape students' understanding of AI's capabilities and limitations. By embracing a balanced approach that acknowledges both benefits and drawbacks, academic libraries can lead their institutions toward a future where AI enhances rather than undermines scholarly work.