Certified Professional Coach, Institute for Professional Excellence in Coaching
Interim Associate University Librarian for Health Sciences & Director, Health Sciences Library
UNC University Libraries, University of North Carolina at Chapel Hill
LinkedIn
“This is going to be interesting.”
Michelle Cawley, MLS, recently shared her thoughts on how Gen AI will alter librarianship and academia. Through varied career experience, the now Interim Associate University Librarian for Health Sciences & Director at UNC Libraries has an appreciation for new technology that she balances with the understanding that “sometimes, the best innovations don’t rely on technology at all”.
Michelle sees the potential for Gen AI to reduce manual workload for tasks such as data cleanup and improve accessibility. She shares some practical advice and insight for those looking to adopt and encourage the uptake of AI at their institution in our latest BMJ Insider interview.
Important Note: You can hear Michelle speak on this topic on March 5, 2024, at a complimentary MLA fireside chat: AI in Action: How Librarians are Implementing and Supporting AI in Their Institutions.
BMJ: What are the focus areas for your current role at UNC Libraries?
Michelle: I oversee the administration of the Health Sciences Library (HSL), which is the primary library for the University’s schools of dentistry, medicine, nursing, pharmacy, and public health. HSL also serves the UNC Medical Center.
I work closely with the University Libraries leadership team, contributing to library-wide strategic planning, program and policy development, change management, inclusive excellence, management of facilities and infrastructure, budget formulation, collections negotiations and analysis, fundraising and development, and allocation of resources in support of library mission and values.
I’m not interacting heavily with students or faculty in my current role, although I did as Head of Clinical, Academic, and Research Engagement, which is my permanent position.
BMJ: Are you involved with initiatives regarding data science?
Michelle: I’m closely involved with the University Libraries effort to develop and implement priorities around data science, develop resources and provide training to Libraries staff around Generative AI, and align library support of research data management with other campus units.
Day to day, I advance key partnerships by meeting with campus leadership and school administrators. I represent University Libraries in a variety of forums on campus, locally, and nationally, including UNC Research Deans, campus working groups to meet federal data management and sharing mandates, our regional consortium and association, and the Medical Libraries Association.
I’m also leading the Impact Measurement and Visualization team at HSL, which uses bibliographic data to visualize and explore publication output, collaboration patterns, and research focus areas to communicate and understand research impact or investigate a research domain.
BMJ: Your impressive librarian career began outside of academia. In which other industries have you worked?
Michelle: I started my professional career as a research assistant with ICF, a management consulting firm headquartered in Fairfax, VA. At that time, I was doing policy work largely in support of contracts with the U.S. EPA.
I moved to North Carolina to pursue a master’s degree in Ecology. After getting my degree, I spent about five years teaching high school science, including AP Environmental Science and Biology. I value the experience I gained as a teacher immensely and still love opportunities to develop interactive and engaging content for students.
I returned to management consulting (also at ICF) after leaving K-12 education. For this round, I worked as a librarian leading literature searches around human health risk assessment. I was also a project manager for large mission support contracts with the U.S. EPA and the National Institute for Environmental Health Sciences (NIEHS). As a project manager, I did a range of things, from managing our support for the risk assessment of Hexavalent Chromium, to developing risk assessment training materials, to creating an interactive tool for exposure assessors using data from the Exposure Factors Handbook.
BMJ: How do you think your previous roles have benefitted your understanding – and appreciation – of innovation, including AI?
Michelle: At ICF, there was a culture of innovation. We had full support from management to develop new ways of doing tasks more efficiently for our clients. There was an internal grant program to support innovation. I received a small grant through this program with another colleague, where we began to develop an AI tool to assist article screening efforts. I think innovation was facilitated at ICF by working closely with staff with complementary skill sets. I brought a unique perspective and set of skills and worked alongside colleagues with expertise in data science, toxicology, statistics, exposure science, visualization, and other fields.
I firmly believe we were able to do more together as a multi-disciplinary team approaching a problem than any of us could do alone. I started to understand how my colleagues could bring my ideas to life, particularly those colleagues with programming experience! We always sought to improve and develop better ways of doing things to save time and money for clients.
As a K-12 teacher, I had the freedom to be highly innovative in my approach to teaching the curriculum. I loved that. I loved that I could learn something new at a professional development training and bring it to the classroom almost immediately. I don’t think innovation has to be technology-based. One of the most innovative things I’ve done in my career involved zero technology or even equipment, only imagination. So, instead of teaching by telling students through a slide deck or even something hands-on – I developed lessons where it was fully in the imagination of students. For example, we did a lesson in my department on the parts of the microscope without a microscope. I would wheel out an empty cart and tell students I was going to teach them the parts of the microscope by holding up an imaginary microscope. You can imagine that got their attention as I was absolutely entertaining them! I’d go over each part, giving it a noise and a gesture. It is deceptively simple but incredibly powerful – I know because, in our training as teachers, we learned the geography of Spain on an invisible map from an instructor who only spoke in Italian (which none or few of us could understand). I was blown away and went home that night (the last day of the school year) and wrote a lesson for the fall. Full credit to Louis Mangione who is a brilliant master teacher and taught the professional development I’m referencing back in the mid-2000s.
BMJ: Although AI has recently accelerated in popularity, you note in a chapter of the Handbook of Research on Academic Libraries as Partners in Data Science Ecosystems, that using machine learning (ML) to assist in article selection has been well documented for 20+ years. Why do you think research libraries have been slow to adopt ML for this purpose?
Michelle: I think it is partly due to the Diffusion of Innovation Theory. The theory was developed by E.M. Rogers in 1962 and provides a framework to understand how a new idea or technology (like AI) is adopted over time and begins to gain momentum and diffuse through a population. The curve is bell-shaped, with innovators being the first to try something new, followed by early adopters. Over time, the idea, innovation, etc., spread throughout the population with “laggards” at the tail end. Laggards are individuals who tend to be conservative and more skeptical and are the final group to adopt something new.
You can imagine how this is relevant to the spread of AI. I’m not a scholar in this area, so I can’t tell you where we are on the curve with my use of AI, but what I am observing is that with the advent of Gen-AI diffusion, the acceptance of this older application of AI seems to be accelerating.
I also think the application of AI I refer to in that chapter would diffuse more quickly with access to free or low-cost resources to apply the technology. Lastly, training for health sciences librarians that allows them to explain and apply the approaches could help to advance acceptance among research teams.
BMJ: What are the current barriers to AI application among research teams?
Michelle: In my opinion, the main barriers are access to a tool or resource that incorporates AI-assistive article screening and training among health sciences librarians on how it works so they can consult with research teams. I say health sciences librarians because the application is most logical for systematic reviews and other comprehensive literature searches. That said, other applications could be in any domain that requires a comprehensive literature search for a research question. I just think health sciences is a natural place to find early adopters.
BMJ: Do you have any recommendations for colleagues interested in increasing the uptake of AI at their institution?
Michelle: I recommend starting small with a pilot project to demonstrate efficacy. If you can partner with a student or other staff with programming experience, a lot can be done quickly. An easy entry point might be to start experimenting with clustering algorithms (i.e., unsupervised machine learning). My sense is this is fairly easy programming in Python or R. For the application of ML that I’ve been using, I tried to write the book chapter referenced above as somewhat of a “how-to” geared specifically for librarians.
I saw a suggestion in one of the groups I’m a part of to pair librarians with ML or AI experience with those less experienced. I don’t imagine it would be 1:1, but rather small groups because AI literacy among librarians likely isn’t widespread enough yet. I would really like to see something like this come to fruition. It’s how I trained the librarians at UNC – by creating a mentoring network. I would recommend pushing their professional associations to offer such training and mentoring.
For Gen-AI, I’m interested in developing and piloting Gen-AI use cases for our work and also for how we do our work. Luckily, there’s interest all across my campus, so this doesn’t feel like a hard sell.
BMJ: What are your go-to resources to stay on top of this quickly changing technology?
Michelle: I go to webinars and other trainings offered whenever possible. I subscribe to Ben’s Bites, which provides daily curated news and products related to AI. I follow updates from the Association of Research Libraries (ARL) and Coalition for Networked Information (CNI), among others. But one of the best resources is my colleagues and sharing information through a Gen-AI Teams group we’ve set up, including an “early adopters” channel to help sift through the firehose of information.
BMJ: How might the introduction of Gen AI further alter librarianship and academia?
Michelle: This is going to be interesting for sure! Domain-specific chatbots will likely be adopted and will hopefully advance quickly. They could become a viable (and perhaps better) option for online reference or other questions compared to live staffing by librarians.
We’re also interested in how Gen-AI could be used to improve accessibility. For example, University Libraries are involved in a project to demonstrate the efficacy of describing images using Gen-AI. The descriptions created are far more detailed and robust than the short, simple version provided by accessibility software and have the potential to save significant time for staff who develop metadata. In this example, the images are more accessible in multiple ways, including directly for people with limited or low vision, but also more findable through searching metadata.
I’m also interested in how data clean-up or data wrangling could be more efficient using Gen-AI. For example, the team at my library who do bibliometrics and research impact spend most of their time on data cleaning tasks and relatively little on data analysis and visualization. In other words, the 80-20 rule of data science applies here and in other data services offered by academic libraries in that 80% of the time is spent on getting the data ready for analysis and 20% on analysis and other tasks. We are exploring whether Gen-AI can reduce the effort and time for data cleaning tasks.
For my application of AI, it will take some time until we see Gen-AI able to perform at scale for a reasonable cost to assist with literature screening tasks. With the technology I currently use, we are already able to achieve a very high recall of relevant articles (95% or higher). Still, we have room to improve in terms of better precision, so lowering the number of articles that must be screened to reach 95% or higher recall.
BMJ: What sparked your interest in a library science career?
Michelle: I originally got into librarianship to work as a K-12 librarian, but then I returned to ICF, which pulled me more in the direction of academic librarianship. When I got the opportunity to explore AI and ML for bibliographic data, my career took off. I had so many ideas that I was able to explore and being able to do something new and different was thrilling.
BMJ: How has the COVID-19 pandemic affected the UNC University Libraries?
Michelle: Probably in many of the same ways it has affected other academic libraries. I’m proud of how my organization prioritized health and well-being for staff and that staff were able to effectively work from home. Most librarians continue to have a hybrid work arrangement, which has been deeply appreciated among HSL staff. I think we were also better prepared in some ways as we had been using Zoom and Teams effectively before 2020, so the transition there was easy.
We also began offering our library workshops via Zoom in early 2020 and found a better turnout than in person alone. Today, we only offer via Zoom, although we do teach in person for curriculum-integrated instruction.
BMJ: What is the most rewarding aspect of your career?
Michelle: Bringing ML and other innovations to research teams and training staff at HSL to apply these approaches has been a highlight for me.
BMJ: What is the most challenging aspect of your career?
Michelle: It is sometimes challenging for me to accept the slow pace of change. As mentioned above, being a K-12 teacher was thrilling because I could have an idea and enact it almost immediately. In general, that’s something I love about teaching. Trying something new to reach and engage learners will always be a passion for me. I have many ideas, not all of which are good, but I like to try new things. In an academic setting, I am learning to be more patient and persistent in pushing for changes. It’s exciting when something like Gen-AI comes on the scene and everyone pays attention to the potential.
What do you do for fun?
I love spending time with my daughters and enjoy creative pursuits such as writing, drawing, and improv. I’ve been doing improv for nearly five years, and I’m part of a team that performs regularly in Durham, NC. I am also a longtime runner and regularly run trails near my house. I’ve finished two marathons (NYC and Richmond) and the Blue Ridge Relay, and I’ll be doing the Ville to Ville relay this April, which goes from Asheville, NC, to Greenville, SC!
Interviewed by Lauren Jones, Head of Marketing, BMJ Americas
