Content area
Full Text
Introduction
In the last few years, especially during the COVID-19 outbreak in 2019 and resulting increase in virtual interaction across sectors, a growing adoption of conversational agents that utilize artificial intelligence, machine learning and natural language processing, largely known as “chatbots,” has emerged. We are currently in what has been described as a “chatbot tsunami” (Grudin and Jacques, 2019). Adoption of chatbots has been accelerated by tools that promote interoperability and integration of these into all communication platforms, from Facebook Messenger to your organization's website.
The growing interest and adoption of chatbots and other artificial intelligence and machine learning tools can also be seen in libraries (Bilal and Chu, 2021). Through chatbots, libraries may increase engagement with patrons, provide enhanced information services and better understand patron information-seeking behaviors (McNeal and Newyear, 2013; Vincze, 2017; Shi et al., 2021). In addition to the stated benefits, San Jose State University's (SJSU) Dr. Martin Luther King, Jr. Library desired a solution to increase interaction with students seeking help during overnight and weekend hours when no reference librarians or circulation staff were available.
While there are many full-service chatbot options available for purchase or subscription, these services can run well into the tens of thousands a year depending on the number of users and transactions. These costs may not be sustainable at public or non-profit organizations such as libraries; lack of means and resources are often cited for the slow adoption of AI and chatbots in libraries (Bilal and Chu, 2021; Ehrenpreis and DeLooper, 2022). Additionally, many libraries are discouraged by the perceived technical requirements of chatbot implementation and upkeep.
To investigate the possibility of building a chatbot with minimal cost and coding, SJSU King Library sought two virtual interns from SJSU's iSchool to perform in-depth research into current chatbot practices, explore readily available chatbot tools, and attempt a chatbot pilot. Dialogflow was selected for the pilot chatbot based on a variety of factors, including cost, ease of use and ability to train the chatbot (Rodriguez and Mune, 2021). This article builds on the authors' previous work by providing in-depth instructions into creating chatbot content, training chatbot responses and embedding chatbots using SpringShare chat widgets. It also presents the first year of user–chatbot interactions, offering insight into possible...