The role of Animated Pedagogical Agents (APAs) depends on an understanding of the persona effect as a mechanism for increasing student engagement and motivation. We argue that historical figure applications of APAs may be helpful to identify the parameters that give rise to a persona effect. Given the importance of visual information, an experimental approach was used to examine how different image conditions would affect perception of a historical figure APA interaction. Eighty-eight participants were randomly assigned to one of three conditions; no image, static image, or animated image. Contrary to expectations, the no image condition was associated with significantly higher ratings for 6 of the 12 measures, including 3 measures of social presence. These findings stand in contrast to previous research and suggest that historical figure applications may be unique in their evocation of a persona effect and valuable for understanding the nature of the persona effect.
TopThe Persona Effect
Although there have been numerous investigations of the persona effect in APAs (e.g. André, Rist, & Müller, 1999; Atkinson, 2002; Dirkin, Mishra, & Altermatt, 2005; Heller, Procter, Mah, Jewell, & Cheung, 2005; Johnson et al., 2000; Koda & Maes, 1996; Mayer, Sobko, & Mautone, 2003; Sproull, Subramani, Keisler, Walker, & Waters, 1996), a close examination of the literature concludes that the evidence for a persona effect is mixed (Clark & Choi, 2005; Dehn & van Mulken, 2000). Dehn & van Mulken (2000) identify 3 possible effects on the user as a result of an APA interaction. Evidence for a persona effect could be found in the user’s (1) subjective experience and perceptions of the system (2) behavioural record produced during the interaction, and (3) learning outcome as revealed by performance measures taken subsequent to the interaction.
In terms of subjective experience, as assessed through self report, there are numerous reports that interactions with animated agents are perceived by users to be engaging and entertaining (Takeuchi & Naito, 1995; Koda & Maes, 1996; Van Mulken, André, & Müller, 1998), believable (Lester et al., 1997), comfortable (Koda & Maes, 1996), and useable (Cassell & Thórisson, 1999; Lester et al., 1997; Takeuchi & Naito, 1995). Although the use of self-report data is important, the construct validity of many of the self-report measures has not been established (Dehn & van Mulken, 2000). Moreover, many of the studies lack suitable control conditions to establish the degree of attitudinal preference (Clark & Choi, 2005). Importantly, Gulz and Haake (2006) have pointed out that issues of visual rendering have been virtually ignored despite the close relation between visual information and attitude formation. Finally, an evaluation of the agent will be closely related to the overall performance of the system with respect to task goals and objectives (Dehn & van Mulken, 2000).