The perception of artificial intelligence as "human" by computer users

Abstract

This paper deals with the topic of 'humanness' in intelligent agents. Chatbot agents (e.g. Eliza, Encarta) had been criticized on their ability to communicate in human like conversation. In this study, a CTT approach was used for analyzing the human and non-human parts of Eliza's conversation. The result showed that Eliza could act like a human as if it could greet, maintain a theme, apply damage control, react appropriately to cue, offer a cue, use appropriate language style and have a personality. It was non human insofar as it used formal or unusual treatment of language, failed to respond to a specific question, failed to respond to a general question or implicit cue, evidenced time delays and phrases delivered at inappropriate times. © Springer-Verlag Berlin Heidelberg 2007.