Finding improvement areas in human-chatbot conversation experience by isolating limitations to learning and context: a quantitative and qualitative study

Abstract

Chatbots are becoming increasingly common recently, and everyone is in a hurry to get their own. Organizations use them to alleviate workload and improve services. At the same time, individuals and content creators who are making a living out of the Web 2.0 and social media phenomena can utilize them for audience engagement and click-through rates. Tech-giants such as Facebook, Google, Microsoft, IBM all have their version of AI-based virtual assistants and chatbot frameworks. With the current popularity and vast scope, chatbots deserve an in-depth study of their application, implementation, and interactions to clear a path for their normalization, so they do not end up like Sony’s Betamax, the apparent superior technology that lost to the VHS videotape format, i.e., the rule-based chatbots winning over superior AI technology. In the following study, we try to sum up the different factors in building, selecting, and training a chatbot, and how users perceive them.