In the first part of this article, we explained our research project whereby a RoBoHoN was placed at a kindergarten and children's interaction with it observed during free activity time (Study 1). In this second part, we will introduce another study where we asked some questions to children, who communicated with RoBoHoN, to find out how they recognized the robot (Study 2). Based on our research findings, we will discuss the characteristics of children's literacy and cognition towards such a new AI agent within the context of ECEC.

Study 2: Young Children's Interaction and Cognition Towards RoBoHoN in an Experimental Setting

Purpose and Method

The purpose of Study 2 is to set an experimental setting where young children can interact with RoBoHoN and find out how they recognize the robot.

This study targeted 28 children aged between five and six years who attend kindergarten. The experiment was conducted with three to four children each time for about ten minutes in a separate room within the kindergarten. Then, we analyzed the results of 22 children who clearly responded to the robot.

In the experiment, the participating children were first provided with a list of questions (e.g., "What is your favorite XX?"). Next, each child asked a question to RoBoHoN and received its response. Then, we asked these children to explain to RoBoHoN about their activities and play on that day. The RoBoHoN was programmed to nod in response to the children's words, aiming to create pseudo interactions between the children and RoBoHoN.

After the experiment, each child was asked whether they thought they could conduct kindergarten activities (play) with RoBoHoN.

Results and Discussion

At the first stage of the experiment, we observed many children's polite (self-conscious) reactions, in particular, girls' comments saying, "How cute!" when they interacted with RoBoHoN (Case 1). These reactions indicate that the children somehow recognized RoBoHoN as their junior (for example, like a baby or a pet) while having a sense of awe towards the robot. At the second stage, children explained to RoBoHoN about their activities and play on that day. Again, we confirmed that children were delighted with the robot's response to their stories. Thus, they seemed to recognize that RoBoHoN could communicate with them.

However, it should be noted that only a few children could describe their activity with a proper sentence such as "I did XXX play today." Instead, most children explained to the robot by just saying a word or phrase such as "XXX play!" This may be one of the characteristics of communications with RoBoHoN, different from those with their peers. In addition, this behavior probably reflects typical difficulty children have in describing their experiences.

Furthermore, a lot of children were observed making experimental attempts to elicit a response from RoBoHoN (Case 2). Gradually, they carefully chose their words according to the robot's response. In other words, the children noticed that they were interacting with something having distinctive characteristics from themselves. Because they recognized this difference, they tried to adjust their method of communicating with the robot. Such behavior may be part of their media literacy development. It should be noted, however, that most children were satisfied with the robot's reaction to their experimental attempts (in the form of questions and answers) and did not wish to continue or deepen the conversation.

Next, children were asked whether they thought they could play with RoBoHoN. All participating children answered "Yes." Then, they were asked what kind of play they could do with RoBoHoN. Their answers were analyzed according to the following two categories: one was the answers reflecting the RoBoHoN's functions, and the other was the answers just picking up their favorite activities without reflecting RoBoHoN's functions. The former category includes answers such as "I want RoBoHoN to give us directions when we go on a school excursion," "I want to dance with RoBoHoN," and "I want to sing a song with RoBoHoN." These answers indicate that children want to play with RoBoHoN utilizing its functions. The latter category includes answers such as "play tag" and "draw pictures." Meanwhile, the response rate of each category is as follows: the answers of three children were classified in the former category (14%), the answers of 14 children in the latter category (64%), and the answers of five children that do not belong to either category (23%). The results revealed that the majority of children's answers were not related to the RoBoHoN's functions.

Based on the above results, it was confirmed that young children considered RoBoHoN as something close to human beings that could play with them, realizing its functions such as body movements and speech. Nevertheless, when children were asked what they would do with RoBoHoN, they seemed to have difficulty in visualizing whether it could actually take part in such play.

Conclusion

The Outcome of Two Studies: Children's Reactions through Interaction with the RoBoHoN

The results of both studies discussed in this article indicate that young children were strongly motivated to interact with RoBoHoN, a humanoid AI robot. It is confirmed that children gradually sought to communicate with RoBoHoN by adjusting their way of interacting with the robot based on its responses. We also observed animistic behavior (see below) in children when they were interacting with the robot. This indicates that children recognize the robot as something different from human beings but which has the ability to think and communicate.

Child animism is typically observed in the developmental process of young children, which is children's tendency to perceive things as being alive by applying the animate characteristics of human beings. Therefore, for young children, human-like interactive media such as humanoid robots (e.g., the RoBoHoN used in our studies) are more accessible to be perceived as human beings than other types of interactive media. Furthermore, a humanoid robot has an autonomous movement function in response to children's actions, which can capture young children's attention and curiosity more effectively than other types of AI media.

Considering these characteristics, a humanoid robot is more suitable for young children to become familiarized with interactive AI media.

Based on the above results, it can be said that young children recognize RoBoHoN as something having distinctive characteristics from themselves. Therefore, they seek to communicate with it by adjusting their way of interacting with the robot based on its responses. Such behavior may be considered an experience fostering their media literacy development (cognitive skills). In addition, the children's interaction with the AI robot resembles those with human beings more than other types of AI media. Thus, such interaction may affect the development of their non-cognitive skills as well. It should be noted, however, that children's experience with the AI robot cannot be a substitute for communication with human beings. They will acquire different skills through interaction with the AI robot from those obtained from human communications.

Future Outlook

There have been a substantial number of basic experimental studies on AI robots focusing on the development of children's cognitive skills. Nevertheless, much fewer applied studies were performed from the standpoint of educational practices in the preschool and school settings. Therefore, we believe that it is essential to accumulate practical research data obtained in the settings of ECEC, like our studies described in this article. Nowadays, AI-related technologies are rapidly advancing and used in numerous ways in our daily lives. Therefore, we can think of various possibilities of such technologies being used in ECEC. For example, by extending and refining the studies discussed in this article, existing research findings can be re-examined in the context of ECEC. Furthermore, it may be possible to clarify the impact of AI robots on children's living and learning environments as well as the possibility of its educational effects through the growth (evolution) of both AI robots and human beings. This can be conducted based on progressive research studies on AI technologies, such as deep learning/autonomic interaction programming and cognitive developmental robotics (e.g., a study to clarify the effects of cognitive developmental robotics on human's cognitive skills through the development of robots; Asada, 2009).

In addition, AI technologies can be utilized to alleviate the work burden of kindergarten teachers, in addition to their use as children's educational materials, thus improving the quality of ECEC. The Japanese government is currently promoting the incorporation of new IT technologies in our society under the theme of "Society 5.0" (Cabinet Office, 2016) to drive revolutionary changes in our living and working environments. Under such circumstances, we believe that we are reaching a tipping point where we should examine the proactive use of AI/ICT devices in various educational activities in our daily lives, such as preschool's educational practices, management and administrative tasks, based on the findings of past studies.

References

• Asada, M. (2009). Approach to Understand and Design Cognitive Developmental Robotics that Enhances Human's Body, Brain, and Mind. Japanese Psychological Review, 52, 5-19.
• Cabinet Office, Government of Japan. (2016). 5th Science and Technology Basic Plan.
  https://www8.cao.go.jp/cstp/kihonkeikaku/index5.html (Accessed September 5, 2020; in Japanese)

Supplementary Note:
The research study discussed in this article is part of the research project (Basic Research B-18H01064; representative: Hiroshi Hotta), which received the Grants-in-Aid for Scientific Research from the Japan Society for the Promotion of Science.