14th Scandinavian Conference on Information Systems

Abstract

Chatbots are becoming increasingly popular. However, little is known about the way chatbots should be designed. Whether the users should be informed or not beforehand that they are chatting with a chatbot is an open question. Similarly, questions related to the level of ‘humanistic’ tonality in interactions with chatbots are unanswered. In this paper, we present a controlled experiment in which 40 individuals participated. Their user experience was compared depending on whether they knew that they were chatting with chatbots before or afterwards. Two different versions of chatbots were tested (one with mechanical tonality and one with humanistic tonality). Our findings illustrate that: i) it is vital that the users enter the conversation knowing that they are chatting with a chatbot; ii) tonality matters, the way chatbots are designed is pivotal for the user experience, the ‘human-like’ and friendly chatbot was preferred over the mechanical, task-oriented chatbot.

Download

Share

COinS