GenAI@BCLibraries: Pedagogy & Information Literacy
Approaching the question of AI in relation to teaching information literacy to students raises a few questions. Four of these are detailed below, but weaving through them all is the question of when AI can help build robust research skills rather than detract or distract from it. In a class recently, after a discussion of complex theory and how it could benefit research, a student asked, but isn’t this actually just asking good questions?
How can AI help us to ask good questions rather than distract us from learning how to ask those questions? How do teachers of information literacy ensure that we continue to teach the complexity of the research process and provide skills–including AI skills–to navigate that complexity?
Information Privilege
Free GenAI tools are widely available, but they generally have use restrictions (e.g. a monthly limit), limited features, or are trained on older or more limited data. It’s important to maintain a level playing field for students, and use tools that don’t disadvantage students who can’t afford to pay premiums for better access. Here at BC, that could mean using the tools supported by ITS & provided to students for free, which include Copilot, Gemini, and NotebookLM. There are also tools that BC doesn’t support, but which are free and robust. For instance, Semantic Scholar, focused on sciences, is one such tool.
Searching for Sources
Though some tools that draw specifically on academic articles as data sources are improving on the “hallucination” problems of ChatGPT, there are still issues. The beta Research Assistant in ProQuest databases regularly suggests similar articles that bear little relation to an article record, and provides reasonably correct but nuance-free summaries of articles. JSTOR’s interactive research tool does considerably better at finding similar articles. However, similarity algorithms have existed for years in many databases, often resulting in closer similarity than the secret sauce of GenAI, without the environmental computing costs. Databases like Web of Science and Scopus are standouts because they draw on the depth and breadth of citation data: shared citations are a bedrock indicator of similarity.
Information Literacy
Librarians help students build their understanding of key features & processes of information that are not intuitively obvious. The Association of College & Research Libraries (ACRL) developed a Framework of concepts that seasoned researchers understand intuitively through long experience but are often outside the grasp of novices:
- Authority Is Constructed and Contextual
- Information Creation as a Process
- Information Has Value
- Research as Inquiry
- Scholarship as Conversation
- Searching as Strategic Exploration
Using AI to find, summarize, and synthesize articles could circumvent some learning processes necessary for changes in perspective that facilitate understanding of these core research concepts.
Wrestling with Ideas: Reflection and Threshold Concepts
The ACRL Framework was adapted from Myer and Land’s (2003) “threshold concepts,” an approach to learning in higher education that emphasizes those “aha!” moments when students who have wrestled with a troublesome disciplinary concept suddenly cross a threshold to understanding.
Wrestling with ideas is an important part of the process of learning, and fits well into the Jesuit approach to pedagogy, which values reflection: making time for students to metabolize information long enough to bring changes in perspective that lead to understanding. This transition requires space to wrestle with ideas, to recognize emotional responses to them, and to synthesize across disciplines or concepts.
How does this all connect back to AI? Reflection and threshold concepts, as well as the Framework above, all require time, focused attention, and a commitment to accepting the discomfort of struggling with troublesome concepts.
A recent library article about demonstrating GenAI chat’s capacities and limitations for generating search keywords illustrates how groups of students brainstorming keywords outperformed GenAI chat in one key dimension: culture. Students were familiar with the cultural landscape of sample search topics, and because of that, suggested culturally-specific search terms that GenAI didn’t. The author’s point was to emphasize how empowering it was for students to see their competitive advantage, but the lesson also illustrated the obverse: how turning to GenAI in place of effortful thought could both nudge students to undervalue their own knowledge and stifle more creative thinking.
Learners grapple with ideas and internalize them so that they can relate them to other ideas, measure new ideas against them, and begin to form new ways of understanding the world around them. GenAI unfortunately can create a mirage of an easy exit from the more thorny, troublesome parts of learning by providing quick, seemingly authoritative answers. But we all know that in formative education, the answers themselves aren’t the goal; the goal is to learn processes of finding and evaluating answers. Librarians, therefore, have a role in teaching how AI can be an assistive tool for some aspects of research, but also in reminding students of the importance of working through and reflecting on troublesome concepts in a process of exploration, discovery, and creation, rather than leaning into AI’s escape hatch.
Further Reading
- Meyer & Land original (2003): original article on threshold concepts
- Townsend, Brunetti, and Hofer (2011): article that led to the ACRL Framework
