Chatbots are often spoken of in the context of artificial intelligence (AI). However, the field of AI covers a lot of ground. At best chatbots would be considered weak AI. They are in no way intended or expected to think, and certainly not to have self-awareness. They are simulators, capable of only appearing to understand questions and comments and giving appropriate responses. Even so, the simulations can be very convincing and useful. Besides providing information they are pre-programmed with, chatbots can extend their reach to search library systems and search engines such as WolframAlpha and Google.
Let us take a look at two library chatbots. Dewey (http://www.akronlibrary.org/dewey.html) is a chatbot at the Akron-Summit County Public Library in Akron, OH. Pixel (http://pixel.unl.edu/) is a chatbot at the University of Nebraska-Lincoln Libraries. As you experiment with asking Dewey and Pixel different questions, a few conclusions become quickly obvious. They are definitely not human and cannot carry on general conversation very well. They can, however, provide helpful responses to library-related, information seeking questions and provide these responses 24 x 7.
Many libraries already offer human-staffed chat, but the number of hours it is available is limited because the availability of human staff time is limited. Can a chatbot be used to fill in during the hours chat would normally have to be shut down? Or, if a library does not have a chat service, could a chatbot provide one? Chatbots are not perfect, and they cannot provide anywhere near the service and help a human library staff member can, but Emma and Pixel demonstrate they can provide a fair bit. I would argue that providing a service where none existed before, especially a service that helps even one more person than would have been helped otherwise, is worth it.
Chatbots are normally given a name and represented by an avatar. Having a well-conceived chatbot could transform a library, normally thought of as a place, into something more personal. I have observed a number of people who are regular users of both public and academic libraries that keep going back not to go to the library, but to seek help from a specific person they have developed a rapport with. They are not going to a place for help, but to someone. It is a bit of a stretch -- I admit -- as a chatbot can only take this idea so far, and different people will react differently to chatbots, but it does add some unique personality to the traditional library setting no matter what.
Chatbots are also fun. Anything that catches the attention and interest of an individual and engages them with the library is always a win for both.
The easiest way to begin experimenting with library chatbots is to download the free library chatbot infoTabby (http://code.google.com/p/aiml-en-us-ovrp-infotabby/) created by David Newyear of the Mentor Public Library in Mentor, OH, customize it to your library, and then host it for free on the chatbot hosting service Pandorabots (http://www.pandorabots.com/). Be aware that, especially with a potentially very popular chatbot, a point could be reached where it becomes necessary to move to one of the paid hosting plans available on Pandorabots, or undertake the more technically complex work of hosting it on your own as is done with Pixel at the University of Nebraska-Lincoln Libraries.
I believe chatbot technology has a lot to offer libraries, and I believe libraries can contribute a great deal to the future development and advancement of chatbots. In fact, I believe this to be true of all AI technologies in general. Is it time for libraries to begin moving in this direction?
David Ratledge is Associate Professor and Head of Systems at The University of Tennessee Libraries, Knoxville. firstname.lastname@example.org