Abstract
Conversational virtual assistants have exploded in popularity in recent years. Developers have been working to create personality in virtual assistants, so that users treat them more than mere automation tools. Among other factors, motivation is considered playing an important role. However, previous researches focused mainly on extrinsic motivation, and ignored the fact that intrinsically motivated engagement cultivates better relationship of reliance. Intrinsic motivation is derived from inner urge to seek enjoyment in an action itself, while extrinsic motivation is oriented at consequences and rewards. In context of virtual assistant interaction, motivation is formed based on the user’s prior expectation: whether the assistant can cope with his/her intention. When the expectation is contradicted, the user is subjected with disruption and is bounced back to learning process about the assistant’s attributes. Related works in expectation effect found that the surprising amplitude of disruption is function of perception uncertainty and prediction error. The current research investigated effects of expectation towards the assistant’s capability and uncertainty during perception on intrinsic motivation. We conducted laboratory environment experiments using Amazon Echo smart speakers and extracted subjective assessment as well as used free choice paradigm as objective measurement. The result showed a significant trend that small uncertainty encouraged more intrinsically motivated interaction while large uncertainty motivated the participant to interact in order to resolve uncertainty. Neither uncertainty nor expectation showed main effect on subjective assessment on intrinsic motivation. Our findings suggest that trading consistency for performance is a risky strategy in designing virtual personality.
