Resolving the Paradoxical Effect of Human-Like Typing Errors by Conversational Agents

Conversational agents (CA) interact with users via natural language in a turn-taking communication. They have been shown to trigger anthropomorphism in users, leading to the perception of humanness. However, this effect is influenced by various aspects, such as the CA having a name and its choice of words. In this context, the expected effect of typing errors is paradoxical. On the one hand, making typing errors is human-like and should, therefore, increase the perception of humanness. On the other hand, users understand that CAs are not human, and typing errors can be perceived as unexpected. To investigate the effect of typing errors on perceived humanness, we employ the dual processing theory of human cognition and propose a NeuroIS experiment to test our hypothesis. In this research in progress paper, we present the experimental design, expected outcome, and reflect on possible implications for CA research.