2.2.4. Measurements

First, we checked whether the participants were adequately stimulated from the following two points of view. Namely, we checked whether the contents of the conversation presented in the experiment were too difficult for the participants to understand and whether the participants in the semi-passive condition were able to engage in the conversation. Regarding the difficulty of the conversation, it was checked whether the contents of conversation made sense (D-1) to the participants. Meanwhile, it was supposed to not be too difficult for participants (D-2) to avoid a floor effect in evaluating how much objective information was conveyed to participants.


Regarding the degree of engagement to the conversation, it is important for the system to correctly handle the replies from the participants so as to prevent them from feeling ignored to provide the experience of semi-passive conversation. If the system failed to do that, it was considered that the androids' attitude toward engagement in the conversation with the subjects, i.e., the feeling of being given attentive (E-1) and interested (E-4), as well as the participants' attitude to engage in the conversation with the androids, i.e., the motivation to speak (E-3) and to join the conversation (E-2), would not be recognized. Note that only the participants in the semi-passive condition were evaluated for this as a comparison to the passive condition in this regard was considered to be unfair because the androids never paid their attention to the participants in the passive one. The degree of engagement with the conversation was surveyed using the following questions:


To verify the first hypothesis regarding objective information, we conducted a recall test to confirm how much the participants memorized the content of the conversation. We created ten questions to ask about the contents that the androids conveyed in the conversation. For example, the participants were asked: "How many androids existed in this laboratory?", "To which country did the female android want to go?", and so on. We counted how many questions the participants could correctly answer. To easily judge whether the answers were correct, questions were chosen so that the correct answers could be described as a single word.

The second and third hypotheses are related to the effective conveyance of the subjective information. The second one concerns whether participants regarded messages from the androids more strongly in the semi-passive condition than in the passive one. To verify this, the strength of messages from the androids about the subject's experience in the conversation was evaluated by using the below questionnaire:


Furthermore, as predicted in hypothesis (iii), the participants were expected to be moved to follow the subjective messages in the semi-passive social conversation than in the passive social conversation. To verify this, we measured how much empathic concern the androids elicited from the subjects by using a questionnaire. In this experiment, we focused on empathic concern as a typical example of emotional response because the androids talked about their story to effect empathic concern in the conversation. Specifically, the following two questions were used:


A seven-point Likert scale was adopted for questions about the strength of the message, empathic concern, the difficulty of contents, and engagement, which ranged from 1: Strongly disagree, 2: Disagree, 3: More or less disagree, 4: Undecided, 5: More or less agree, 6: Agree, to 7: Strongly agree.
