Human and Humanoid Don't Match: Consistency Preference and Impact on Users' Trust

For many anthropomorphic representations on computer interfaces, there are choices of using mediacaptured human components and using computer-synthesized humanoid counterparts. This applies to pair the face and the voice of a talking-head character. It is not uncommon that designers mismatch the face and the voice of a talking head in the human vs. humanoid dimension. Based on the literatures on person perception and communication processing, consistency is proposed as a natural human preference which calls for matching the face and the voice. A 2x2x2x2 mixed-design (human face vs. humanoid face by human voice vs. humanoid voice by male vs. female users by attitudinal vs. cognitive task) experiment (N = 80) was conducted. The results supported the principle of consistency preference in terms of users’ trust and other attitudes. Attitudinal responses in accordance to consistency preference were stronger among female users than males. For the cognitive task, however, human voice was preferred over the synthetic voice regardless of the face.