Conversational agents in the form of chatbots available in messaging platforms, or smartphone and home-based virtual assistants, are gaining increasing relevance in our communication environment. Based on natural language processing and generation techniques, they are built to automatically interact with users in several contexts. We present here a tool, the Conversational Agent Research Toolkit (CART), aimed at enabling researchers to create conversational agents for experimental studies using computational methods. CART is an alternative to Wizard of Oz studies, in which researchers usually resort to (human) research assistants to pose as automated conversational agents. This paper provides an overview of the tool, discusses the role of computational methods in enabling research with conversational agents, and provides a step-by- step tutorial of to design an experiment with a chatbot.
This work is licensed under a Creative Commons Attribution 4.0 International License.