The past 10-15 years has seen an increasing field of research in human-computer interaction (HCI) and more specifically, humanized agents known by many names, social model, pedagogical agent, character, and avatar, to name a few. A humanized agent whose role is to specifically facilitate learning, the pedagogical agent, has been of particular empirical interest. Their features, such as voice, facial expression, gesture, and affect have been investigated to determine which combination of which features in the computer-based environment are most facilitative to which kind of learners as well as which types of instructional strategies.
This study examined humanized agents as social models of behavior as well as their delivery of message content types on participants’ affective outcomes. In keeping with Bandura’s theory of social modeling (1986) and Reeves and Nass (1993a, 1993b, 1996) investigations into the Human-Computer Interaction (HCI), questions such as, “How do social models and the messages they deliver effectively influence participants’ attitudes toward ethnic groups historically associated with prejudice and discrimination?” were posed. In addition, this study sought to contribute to empirical research in the affective learning domain and spark new interest in what is traditionally perceived as difficult to define and measure.
In order to combine these variables, this research investigated the effects of ethnicity and message content on affective outcomes in a computer-based learning environment. Specifically, ethnicity was represented by computer-based models’ nativeness and non-nativeness (as it pertains to culturally-related content), and message content was represented by three levels: common membership, mutual differentiation content and value neutral content. A 2x3 factorial design was used to determine significant effects on attitudes toward persons of Arab descent and perceptions of computer-based models. Models, in the form of two dimensional animations, acted as mentors representing one of two ethnicities and engaged in dialogue, representing one of three message content levels. This research was conducted within Bandura’s (1977, 1986) social learning and Tajfel (1978) and Turner’s (1975, 1985) social identity theoretical frameworks.
Results indicated there were no overall significant main effects in attitudes toward Arabs or perception of models between students viewing native or non-native models and common membership or mutual differentiation message types. In addition, there were no overall significant main effects in attitudes toward Arabs or perceptions of models between students who viewed the treatment conditions and value neutral message control conditions.
Results did show significant differences in attitudes toward Arabs for two single variables (partially constructing the dependent variable, attitude) between students viewing the non-native model delivering the common membership message and students viewing the non-native model delivering the value-neutral message. Students viewing the non-native model delivering the common membership message reported more positive attitudes when hypothesizing themselves to be on a crowded bus with Arabs or when hearing that an Arab student received financial aid.
In addition, there were some significant differences for items relating to model perception. Students viewing one of the two treatment messages perceived the model to be more competent than students viewing the model delivering the value neutral content.
Lastly, there was a significant mixed effect between students viewing the non-native model delivering either the common membership message or the mutual differentiation message and the non-native model delivering the value neutral message content. Students viewing the non-native model delivering either the common membership message or the mutual differentiation message perceived the non-native model as keeping their attention significantly greater than students viewing the non-native model delivering the value neutral message. Anecdotal evidence indicated NSD interactions among the variables and supported the finding that students’s overall attitudes and perceptions were more positive after viewing the non-native model delivering the common membership message than participants who viewed other combinations of variables. Although more data are needed to clarify and strengthen the impact of these results, these findings have practical and theoretical implications when using social models and messages in informal computer-based learning environments for affective outcomes.