In a unique partnership, staff from IT and reference departments at the University of Delaware piloted a low-cost AI-powered chatbot called UDStax to gauge campus interest and understand labor requirements. The article highlights the importance of involving staff in developing, refining, and reviewing training materials for chatbots. It discusses challenges faced during training, tool selection considerations, and the ongoing pilot phase to assess accuracy and user feedback.
The evolution of chat services in academic libraries is traced back to the mid-1990s, leading to the adoption of AI-driven chatbots due to changing user expectations. The team selected Chatbase for its user-friendly training methods using ChatGPT technology. Challenges such as fake links in responses were addressed through continuous training and documentation efforts.
The implementation process involved upskilling existing staff, creating base prompts, and utilizing various data sources for training materials. Weekly reviews of chatbot responses helped identify gaps and improve accuracy over time. The article emphasizes transparency, privacy, accountability to users, and ongoing assessment for future decision-making.
To Another Language
from source content
arxiv.org
Key Insights Distilled From
by Colleen Este... at arxiv.org 03-05-2024
https://arxiv.org/pdf/2403.01545.pdfDeeper Inquiries