Quality Assessment of Conversational Agents

In this poster I will present my work on quality assessment of chatbots (conversational agents, or CA). I have worked in 3 directions: (i) natural language understanding where I test the quality of chatbot models by introducing artificial faults (e.g., typos, accents) to identify limits in the quality of these software systems; (ii) user experience design and evaluation where I test the impact of chatbot personalities; and (iii) conversational AI ethics where I question the impact of chatbots on users, in particular marginalised groups. I will present in this poster my publications in the domain as well as my work with Microsoft and NRC (an NGO) on chatbots design in particular for young refugees – as well as my many EPE activities.