AI: Transforming Reference
The volume of reference questions has been declining over the years in many libraries, and the experiences of the University of Nebraska Libraries (UNL) are no different. So UNL has investigated and developed a new model of reference. Deann Allison and Lorna Dawes from the University of Nebraska are using an innovative “chatbot” called Pixel to handle reference inquiries.
Chatbots are software applications designed to emulate conversations with human beings. They are frequently text-based but can include sound and visual effects. The software includes a database of answers. There are a variety of ways to create chatbots, including commercial sites, open source projects, pandorabots, or the open-source program, Program O. Here are examples of some chatbots:
Why use artificial intelligence (AI)?
Sources for the database
Pixel was built on Program O with adaptions for the local environment. Here are some usage statistics.
37% of Pixel users are returning visitors so it is attracting a following.
Pixel responses are built on AMIL (AI Markup Language), which looks somewhat like XML. There are 3 ways to ask a question. It is difficult to match on exactly what people ask, so wildcards are liberally used (for example, “Recycling”, “* Recycling”, “* Recycling *”). Scripts can be written to accommodate complex answers. The program flow is books, articles, libguides, and finally the librarian. The librarian is last because people don’t want to talk to librarians. Subject categories were derived from LC classification areas. Chat logs are excellent sources for categories. Pixel also has an internal log of questions asked.
Pixel has a personality. It is usually polite and is patient. Some people use it to practice English, and others think a person is within the system, so they argue with it. It is capable of general conversation but tries to bring the chatter back to library subjects if possible. It can also handle difficult patrons without getting emotional.
Some reference librarians were concerned that Pixel would replace them. But Pixel is a help to them–it is just a tool to search a complex website. Students are taught to be wise selectors of resources; sometimes they use Pixel. It is good for answering FAQs and simple questions. It also can provide simple database searching instruction.
Pixel follows guidelines for reference interviews:
- It is approachable and friendly and encourages the user to feel comfortable. It is in a prominent position on the website to encourage use. It knows how to respond to difficult users.
- It shows interest and demonstrates a commitment to providing effective information assistance by answering in a timely manner and maintaining word contact. It always takes the approach of directing the user to library resources.
- It listens effectively and asks questions as the user states the information needed. Most of Pixel’s questions are open-ended or closed clarifying questions to understand the need while keeping the user at ease.
- It conducts searches and gives the user a printout, so it explains the steps and results to the user.
- It follows up by asking the user is satisfied and the search is complete. If the user is not satisfied, the user is directed to a librarian.
Pixel is open to anyone on the Internet. Try it out!