This Soap-Dispenser Robot Wants to Help Your Kids Learn—But Only If It Feels Like It

Table of Links

• Abstract and I. Introduction
• II. The Casa Paradigm and Gamification
• III. Land of Hands: The Game
• IV. Methods
• V. User Study
• VI. Results
• VII. Discussion

  • VIII. Conclusion

  • IX. Acknowledgments and References

IV. METHODS

A. Study Design

To explore the influence of a pro-social robot’s behavior on learning and interactions between the robot and the child in a collaborative gameplay setting, we split our study into two conditions namely - “With-Nudges” and “Without-Nudges”.

In the “With-Nudges” scenario, the children were provided two kinds of nudges — visual nudges and verbal nudges. Visual nudges were provided in the game in the form of a “Ask Robot for Help” button that glowed and shook at regular intervals. For the verbal nudge, HakshE acted as a pro-social robot exhibiting traits such as proactively offering help. HaKsh-E verbally asked the child “Do you need any help?”, when he/she got stuck even if the child did not explicitly ask for the robot’s help. HakshE also provided constant positive feedback to the child when he/she completed a task.

In the “Without-Nudges” scenario, HakshE did not display pro-social traits. Unless children specifically called out to HakshE for help, the robot did not reach out to them. The levels still contained the “Ask Robot for Help” button, but it did not glow or shake and was activated only when a child clicked on it.

B. Participants

A total of 32 children (15 girls and 17 boys) between the ages of 6 to 10 years (M=8.22 years, SD=1.07) were recruited for the study. All the children are students in schools in two different regions of Kerala, India, namely Puthiyakavu and Kochi (Kochi=20 children, Puthiyakavu=12 children). The participants were chosen through convenience sampling to ensure that every child had access to a working laptop and a stable internet connection. As our study consisted of a two-way split, to maintain equality in the distribution of the 32 children, we recruited 16 children for the “WithNudges” condition and 16 children for the “Without-Nudges” condition. The study was undertaken only after the verbal and written consent of the participants’ parents, guardians, and the participants themselves was taken. We ensured that the confidentiality and anonymity of the data collected from the children were maintained throughout the study

C. Materials

1) HakshE - The Robot: We custom-designed HakshE through a co-design study we conducted with children [2]. HakshE’s shell resembles the shape of a soap dispenser to meet its intended requirement of promoting good hand hygiene practises. HakshE’s face is an LCD screen that displays a wide range of emotions through its animated facial features. We designed the robot to appear very minimalistic to avoid any unmet expectations that children might have of social robots (refer to Figure 4). The robot is capable of handling verbal interactions with children on the topic of hand hygiene and essential day-to-day conversations.

2) The Wizard of Oz Platform: A common criticism against Wizard of Oz [26] studies in HRI is that such studies can potentially become a Human-Human Interaction scenario [27], [28]. This can be tackled by ensuring that the Wizard’s interactions are bound to the robot’s intended interaction capabilities. Moreover, the interactions of the Wizard should be streamlined to minimize variability across interactions and to reduce the cognitive load for the wizard [27]. With these two design requirements, we developed a custom Wizard of Oz dashboard for Haksh-E as shown in Figure 7.

We developed the dashboard client to work in any modern web browser, making it platform agnostic. We used d3.js to generate an interactive dialog tree with collapsible branches to make it easier for the wizard to keep track of the dialogues. The client communicates with the server via web-socket protocol providing a full-duplex communication channel with minimum latency. The web-socket server, which runs onboard the robot, connects with the robot’s API to play the required animation on the LCD screen.

To avoid inflicting biases on children’s behavior after the study, we revealed to them that HakshE was teleoperated after the study was conducted. Prior research also shows that in child-robot interaction studies, children’s interaction with social robots did not change in any manner when they were informed about the teleoperation post the study [29], [30].

D. Ethics Statement

The study was approved by the Clinical Trials Registry (ICMR-NIMS) of India (CTRI/2022/01/039654) and the Institutional Ethics Committee of Amrita Vishwa Vidyapeetham, India (IEC-AIMS-2021-AMC-306). The study was conducted in accordance with the Declaration of Helsinki (DOH).