User-Adaptive Personalized Chatbots for Conversational Information Seeking Tasks

A chatbot’s design affects how customers perceive its competence and usefulness, both significant predictors of technology acceptance. The preferred design principles vary with the customer’s emotional state and personality, making it unrealistic to come up with a static design that serves everyone. I propose a user-adaptive personalized chatbot framework where the most suitable conversational design cues are decided interactively. The framework is composed of multiple components, each one providing valuable information on how to modify the conversation style of the bot. Components include mechanisms such as a sentiment tracker for active feedback, and a retriever to check previous conversations for finding similar interactions. After receiving the context from the components, the chatbot actively adapts its conversation style for matching the customer’s emotional needs. With this framework, my goal is to create a truly personalized experience for customers, thus increasing the adoption of chatbots as customer support agents for information-seeking tasks.