New trends, challenges, and developments in robotics have led to the question whether a robot could become an effective teacher. In past studies, social robots have been used, for example, to foster language acquisition, and in support of science education, and technology education.

 

In this pedagogical process, the robots were used as a tool, a peer, a tutor or teacher. While actual robots can be employed in direct physical interactions with students, there is also a possibility of developing virtual agents with robotic appearance to be used in simulated environments.

 

For instance, Li et al. (2016) showed that it is possible to use an embodied pedagogical character with a robotic appearance as a virtual instructor. Li et al. (2016) suggest that the use of a virtual robot could have almost the same effect on students’ recall compared to a human teacher in instructional video content.

 

To test this claim, we examined whether a virtual robot with non-human traits could replace a human instructor in pre-recorded instructional video’s. The result of our investigation significantly contributes to research on pedagogical agents that can be used in the area of educational robotics.

 

Robot Instructors

 

One of the important goals in the field of social robotics is to design robots to be used as education companions and tutors.

 

The decision on whether a robot design is useful for educational purposes is highly dependent on its impact on learning outcomes. For example, Belpaeme et al. (2018) showed than the Nao (human appearance with arms and legs) and Keepon (yellow snowman appearance without arms and legs) robots had a medium-sized effect on cognitive learning gains in students. A virtual prototype of the Nao robot was used in an instructional video by Li et al. (2016), who found that a Nao as a virtual robot agent can have a positive effect on students’ knowledge recall compared to a non-virtual robot, and a similar effect to a non-virtual human teacher.

 

This study offers some important insights into the effect of robotic appearances as a virtual instructor and their wide possibilities in the pedagogical context. Robotic Appearance and the Uncanny Valley Effect Research in robotics on the effect of robot appearances has primarily been based on the “uncanny valley theory”. This theory explores the effect of human-like appearance on the affinity and/or familiarity of the users towards the character. An interesting aspect of the theory is that when the robot comes closer to having human traits, it may be perceived with an increasing revulsion. This effect is called the “uncanny valley” and it is found when the robot is close to but fails to attain a realistic appearance. While anthropomorphic features have been the subject of many classic studies in robotic appearance, recently, the focus has begun to shift toward animal appearances.

 

For example, Schneider, Wang and Yang (2007) suggested that a safe strategy preventing the uncanny valley effect is to design virtual animals with non-anthropomorphized appearance, in such a way that they could emote and communicate as a human. However, Schwind et al. (2018) note that, similarly to virtual/robotic humans, the level of realism of the virtual animal-like characters could generate negative reactions. 

 

Summary

 

To sum up, the literature shows that the theory of the uncanny valley can be applied to robots both with a human and non-human appearance, such as animals. However, currently very little is known on whether the familiarity’s perception described in uncanny valley theory can also affect a robot with animal appearance when it acts as a virtual instructor. In our experiment, we tested different appearances of a virtual animal instructor, compared to a human instructor, on students’ perception of the character and its possible effect on cognitive learning outcomes.