Academic Libraries Can Develop AI Chatbots for Virtual Reference Services with Minimal Technical Knowledge and Limited Resources
Bibliographic record
Abstract
A Review of: Rodriguez, S., & Mune, C. (2022). Uncoding library chatbots: Deploying a new virtual reference tool at the San Jose State University Library. Reference Services Review, 50(3), 392-405. https://doi.org/10.1108/RSR-05-2022-0020 Objective – To describe the development of an artificial intelligence (AI) chatbot to support virtual reference services at an academic library. Design – Case study. Setting – A public university library in the United States. Subjects – 1,682 chatbot-user interactions. Methods – A university librarian and two graduate student interns researched and developed an AI chatbot to meet virtual reference needs. Developed using chatbot development software, Dialogflow, the chatbot was populated with questions, keywords, and other training phrases entered during user inquiries, text-based responses to inquiries, and intents (i.e., programmed mappings between user inquiries and chatbot responses). The chatbot utilized natural language processing and AI training for basic circulation and reference questions, and included interactive elements and embeddable widgets supported by Kommunicate (i.e., a bot support platform for chat widgets). The chatbot was enabled after live reference hours were over. User interactions with the chatbot were collected across 18 months since its launch. The authors used analytics from Kommunicate and Dialogflow to examine user interactions. Main Results – User interactions increased gradually since the launch of the chatbot. The chatbot logged approximately 44 monthly interactions during the spring 2021 term, which increased to approximately 137 monthly interactions during the spring 2022 term. The authors identified the most common reasons for users to engage the chatbot, using the chatbot’s triggered intents from user inquiries. These reasons included information about hours for the library building and live reference services, finding library resources (e.g., peer-reviewed articles, books), getting help from a librarian, locating databases and research guides, information about borrowing library items (e.g., laptops, books), and reporting issues with library resources. Conclusion – Libraries can successfully develop and train AI chatbots with minimal technical expertise and resources. The authors offered user experience considerations from their experience with the project, including editing library FAQs to be concise and easy to understand, testing and ensuring chatbot text and elements are accessible, and continuous maintenance of chatbot content. Kommunicate, Dialogflow, Google Analytics, and Crazy Egg (i.e., a web usage analytics tool) could not provide more in-depth user data (e.g., user clicks, scroll maps, heat maps), with plans to further explore other usage analysis software to collect the data. The authors noted that only 10% of users engaged the chatbot beyond the initial welcome prompt, requiring more research and user testing on how to facilitate user engagement.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.012 | 0.034 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.010 | 0.009 |
| Science and technology studies | 0.002 | 0.001 |
| Scholarly communication | 0.005 | 0.009 |
| Open science | 0.002 | 0.004 |
| Research integrity | 0.002 | 0.001 |
| Insufficient payload (model declined to judge) | 0.050 | 0.035 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".