Access - day 2 - Library Chatboxes in Electronic Reference
Library Chatboxes in Electronic Reference (link to podcast)
Link to slides (ppt)
Anne Christensen (Hamburg U)
Anne was sitting next to me at the conference, so I was very glad to see her do so well with this talk. The people in the Code4Lib chatroom all around us were starting to get a little on the silly side after several long days of work at this conference, and she seemed to completely win them over.
Anne presented Stella, a way to provide reference FAQ’s in a dynamic, interactive way. She pointed out first that several commercial enterprises use chatboxes - for example, Ikea’s Anna. This grant-supported project built and tested an interactive avatar type figure to provide help on the library’s homepage.
They named their chatbox “Stella” and worked very hard to give her a personality that would fit the culture of Hamburg, and the University in particular. Her personality is described as kind and discreet, with a sense of humor - or hanseatic. They were very firm that they did not want her to be a virtual version of any of the U Hamburg librarians. They wanted it to be obvious that Stella is a computer, not a person. She is supported by a knowledge base with 3,000 rules, and building the knowledge base was (of course) one of the most complicated parts of Stella’s creation.
The library’s goal was to provide a resource that would allow students to ask FAQ type questions in natural language, that would make the web-based resources provided by the library more visible, and that would be available to explain some of the complicated access-related processes related to electronic resources more clearly than a long list of text-based instructions. The finished chatbot can “follow” the user as they navigate the website, and she also refers questions to human librarians when her knowledge base doesn’t help her. By doing assessment and tweaking the knowlege base, they have been able to improve her performance from 30% “wrong” answers to 15% (and all of the interactions that conclude with her referring to her human colleagues get counted against her - so the 15% number is deceiving).
I was really interested in this talk because I thought some of OSU’s undergrads would really like an option like this. Anne said that one of the unanticipated outcomes they found was that there was a group of students who used Stella because she was a computer - the relative anonmyity of asking their question to the computer made them a lot more comfortable asking a question at all. That resonated with me; I think a lot of our students would have similar reactions. I also like the idea of a help devide on the homepage that can point students to resources they probably wouldn’t find on their own.