RESNA Annual Conference - 2019

Comparison Between Two Different Prompting Conditions When Children Use A Lego Robot To Perform A Set Of Tasks

María F. Gómez-Medina, Javier L. Castellanos-Cruz, Adriana Ríos Rincón, Patrick M. Pilarski, and Kim Adams

University of Alberta, Canada


Play is a natural activity where children can develop social, cognitive, and linguistic skills [1]. When children play, they receive information via their senses to develop an awareness of relationships with people and objects in the environment [2]. Additionally, play promotes discovery, learning, creativity, adaptation, mastery and self-expression in children [3].

Some children with physical disabilities do not have the same opportunities to play as typically developing children due to their limitations, such as difficulties in reaching and grasping objects. These limitations can result in developmental delays across different areas, including sensory, cognitive, motor, communication, interaction, and social development [4]. During play activities, children with physical disabilities tend to be dependent on their playmates or parents, who oftentimes manipulate the toys [5].

Assistive robots, controlled by switches or other alternative access method, could be used by children to enable access to play.  It is important to consider the skills that children need in order to control the robots, since children should learn how to control the robot before they can use it to perform specific activities successfully [7, 8]. Forman [9] found that causality, spatial relations, binary logic relations, and the coordination of multiple variables are required cognitive skills that children need to successfully control a robot. These skills vary with children’s age, where older children demonstrate a better understanding of the skills than younger children [9]. Assistance such as prompting could be a way to help children at early ages to complete tasks that required the mentioned skills when using a robot.

When adults give assistance to children when they are playing, adults can enhance children´s capacity by performing different roles, such as directors, mediators, observers, or players [10, 5]. Prompting is one form of assistance and can be defined as actions given by adults to children to increase children´s engagement in a desired behavior [11]. Prompting from adults has been used to help children with disabilities when playing using a robotic system [12,7,13], to help children with disabilities when using a robotic arm to perform classroom activities [14], and to support people with intellectual disabilities to complete tasks [15]. However, oftentimes adults overprotect children with disabilities, limiting and inhibiting children’s play competences [2]. When adults overprotect children, they take away the opportunity for children to explore play situations [10].

Technology could be a way to give prompting to children with disabilities in a structured way, making a better match between the child’s abilities and the help the child needs. If the technology is providing the prompting, children could be more independent because they would not require support from adults to perform an activity when using a robot. Self-operated systems have been used to give prompts (e.g., “press the blue button” and “insert the coin”) to people with intellectual disabilities and have increased their performance on tasks and decreased their dependence on others [15]. This type of prompting assistance could be applied to robotic systems to give assistance to children with disabilities only when they need it.

The concept of a robotic system giving the prompts to children is aligned with the idea of automation in machines, in which machines can automatically execute tasks or functions for humans [16]. In this case automation in machines can have different continuous levels, where lower levels represent humans having most of the control over an action, and higher levels represent machines having more control over an action.

In this study, the robot acted according to different levels of automation. At the lowest level, the robot did not give prompting if children performed the correct action independently. At intermediate levels, the robot gave prompting when children made a mistake, for example, the robot said, “Do you think pressing a switch can help us move?”, giving children the opportunity to perform actions independently. At higher levels, the robot acted automatically and performed the action if children made the same mistake several times or if children were not doing anything.

The final goal of this study is to develop a robotic system that gives prompting to children to support them when they are using a Lego robot to perform a set of tasks. As a first step towards this goal, the objective of this study was to observe the performance of children when they obtained prompting from a robot. Thus, the performance of children when they used a Lego robot to perform a set of tasks was compared when they received no prompting from the robot and when they receive prompting from the robot (robotic prompting).



Seven children participated in this study, six typically developing children with ages ranged from three years and two months to five years and 1 month (50 ± 9.1 months) and a child with cerebral palsy. He was eleven years and four months old at the time of the study sessions. He was able to control two switches using his head and his left hand. The occupational therapist of the participant with disabilities was present during all the sessions to give him assistance when he was pressing the switches (e.g., stabilizing his hand/head). Ethics approvals were obtained from the Ethics Review Board of the University of Alberta in Edmonton, Canada, and from the Ethics Research Committee of the School of Medicine and Health Sciences at the Universidad del Rosario in Bogota, Colombia. Consent was obtained from the participants parents and verbal assent was obtained from the participant before each trial. To refer to participants in this paper a code including the age of the participant will be used like this: P1-5y, P2-5y, P3-3y, P4-4y, P5-3y, P6-5y for the typically developing children and P1D for the participants with disabilities.


A Lego Mindstorms (Lego Group, Billund, Denmark) robot assembled as a car-like vehicle with a basket on the back was used. The robot was controlled via Bluetooth by three switches (Ablenet, Roseville, California) connected to a Windows PC using a switch interface (Don Johnston, Illinois, USA). Additionally, ten wooden blocks were used so children could knock down piles of blocks during the tasks using the Lego robot.

When children received robotic prompting, a technique called the Wizard of Oz was implemented to make it seem like the robot was giving the prompting (i.e., talking and giving the prompts) to the participants. Also, the switches to move the robot and the layout for the tasks were augmented so visual prompts could be given. In this case, each switch had three light-emitting diodes (LED) placed around it. These LEDs were controlled to turn on or off by an Arduino Microcontroller (Arduino, New York) in order to give the visual prompts. In addition, a mobile application was developed on a Samsung Tablet using Android studio to say the prompts and the instructions to the participants using voice output over a speaker.


A single subject design research design with a baseline and an intervention phase was conducted to observe the differences when participants received no prompting and when they received robotic prompting. In the baseline phase participants performed tasks using a Lego robot with no prompting. A total of 5 sessions were conducted in the baseline phase. In the intervention phase participants performed the same tasks and they received prompting from the researcher via the “robot” (robotic prompting) using the Wizard of Oz technique. In this case the researcher was the one controlling the prompts via the robot. A total of 3 sessions were conducted in the intervention phase. The sessions were videotaped for analysis.

Participants performed three tasks based on the protocol of Poletz et al. [8].  These tasks used the mentioned skills of causality, spatial relations, binary logic relations, and the coordination of multiple variables [9]. In the first task (task 1) participants were asked to knock down a pile of blocks by pressing and holding a switch to move the robot until it knocked the pile down. In the second task (task 2), participants were asked to build a tower of blocks using the robot. In this task, participants had to stop the robot (release the switch) by some blocks so the researcher could load the blocks on top of the robot. Then, they had to drive the robot until the end of the table so the researcher could build the tower. In the last task the researcher placed two piles of blocks at the left and right side of the robot and two more switches were added to the set up. In this task participants were asked to knock down one of the piles of blocks by first pressing the left or right switch, according to the pile of blocks they wanted to knock down, until the robot was facing the pile of blocks (task 3a) and then pressing the forward switch to knock down the pile (task 3b).

Data collection and analysis

The success rate of the participants in each of the tasks was collected from the videos of the session. It was calculated as the number of times the participants performed the task correctly over the number of attempts (i.e., the number of times the participant tried the task). Two raters were trained by the researcher to code the success rate of the participants during the tasks, and interrater reliability was 87,5% for one and 95.8% for the other.

This figure has 7 figures inside. Each figure has the results for the success rate in all the tasks of each participant during the baseline and the intervention phase. The results are presented in line charts and each task has a respective color to differentiate them from each other. The first graph shows that P3-3y obtained a success rate of 100% in task 1 and task 2 during the baseline.  The success rates for task 3a and 3b were about 65% and 20% respectively.  With the intervention the success rate of task 2, 3a and 3b decreased. The same happened with P5-3y, however his success rate decreased to 0% with the intervention phase for all the tasks.  P2_4y also obtained a success rate of 100% in task 1,2 and 3a during the baseline and about 70% for task 3b. However, with the intervention the success rate in all the tasks decreased to 0%. In the case of P4_4y, he reached a success rate of 100% in all the tasks during the baseline phase and these numbers remained the same during the intervention phase.  In the case of the five year old participants, both of them obtained a success rate of 100% in all the tasks during the intervention phase. P1_5y had a small decreased in task 2 and task 3b during the first session of the intervention phase but these numbers increased to 100% in the following sessions. For P6_5y results of success rate remained the same during the intervention. In the case of the child with disabilities, he had an improvement in the success rate for all the tasks during the baseline phase. This improvement continued during the intervention phase, observing an increase in the success rate for all the tasks.
Figure 1. Comparison of success rate at the baseline and intervention phases for all participants
To make the comparisons between the two phases, the success rate in each task for both phases was plotted and visually analyzed. The mean values of the success rate in each phase were calculated and compared. Also, statistical differences between the phases were determined using a two standard deviation band with respect to the baseline. In this case if, during the intervention, at least two consecutive data points were outside this band, it was considered that there was a significant difference from the baseline to the intervention [17].


In general, all the participants obtained a success rate of 100% in task 1 and 2 during the baseline, and most of the participants had difficulties with the performance of task 3a and 3b during the baseline. According to the two standard deviation method, P2-4y and P5-3y were the only participants who had a significant difference in success rate between the baseline and the intervention phase, in task 2 and task 3a, respectively. Despite the fact that the other participants did not have a significant difference between the phases, changes in the mean value of the success rate were observed when comparing the baseline and the intervention. Success rate was expected to increase or remain the same as the baseline during the intervention. However, the success rate decreased in all the tasks for P3-3y and P4-4y and in task 2 for P5-3y, even though they reached a success rate of 100% in the baseline for task 1 and task 2. Thus, the robot talking and giving the prompts to them seemed to have influenced their performance in the tasks.

This decrease could be because it was the first time that participants had the experience with the robot talking to them. In the case of P4-4y, he had a marked decreased in all the tasks with the intervention phase and it was because he did not do anything when the robot asked him to perform a task. It seems like he needed approval from the researcher to start pressing the switches because he waited for her to tell him that it was ok to press the switches.

When comparing the baseline and the intervention phase results of P1D it was observed that the intervention did not influence the success rate. In this case, the mean value of the success rate increased with the first intervention for all the tasks, Nevertheless, these results could be due to a learning effect since his improvement was observed in his sixth session using the robot.


The present study aimed to observe if there was a difference in the performance of participants when doing a set of tasks when they received no prompting and when they received prompting from the researcher via the robot

using the Wizard of Oz technique. Results showed that the prompting via the robot had a negative effect on the performance of young typically developing children (3 and 4 year old participants) during the first sessions. A possible reason is that participants were not used to following instructions from the robot. Thus, a familiarization session in which the participants get used to the robot talking and giving instructions should be performed where participants can interact with the robot in less structured activities.


  1. Besio, S., Dini, S., Ferrari, E., & Robins, B. (2007). Critical Factors Involved in Using Interactive Robots for Play Activities of Children with Disabilities. Assistive Technologies Research Series: Challenges for Assistive Technologies, 505–509. Retrieved from
  2. Missiuna, C., & Pollock, N. (1991). Play Deprivation in Children with Physical Disabilities: The role of the Occupational Therapist in Preventing Secondary Disability. The American Journal of Occupational Therapy, 45(C), 882–888.
  3. Ferland, F. (2003). Le Modèle Ludique (Third). Montreal, Canada: Les Presses de l’Université de Montréal.
  4. Robins, B., Dautenhahn, K., Ferrari, E., Kronreif, G., Prazak-Aram, B., Marti, P., … Laudanna, E. (2012). Scenarios of Robot Assisted Play for Children with Cognitive and Physical Disabilities. Interaction Studies, 13(2). Retrieved from
  5. Musselwhite, C. (1986). Adaptive Play for Special Needs Children. College Hill Press.
  6. Cook, A., & Polgar, J. (2015). Principles of Assistive Technology: Introducing the Human Activity Assistive Technology Model. In Assistive Technologies (Fourth, pp. 1–15). Missouri: ELSEVIER, Mosby.
  7. Cook, A., Adams, K., Volden, J., Harbottle, N., & Harbottle, C. (2011). Using Lego robots to estimate cognitive ability in children who have severe physical disabilities. Disability and Rehabilitation: Assistive Technology, 6(4), 338–346.
  8. Poletz, L., Encarnação, P., Adams, K., & Cook, A. (2010). Robot skills and cognitive performance of preschool children. Technology and Disability, 22(3), 117–126.
  9. Forman, G. (1986). Observations of Young Children Solving Problems with Computers and Robots. Journal of Research in Childhood Education, 1(2), 60–74.
  10. Blanche, E. (2008). Play in children with Cerebral Palsy: Doing With-Not Doing To. In D. Parham & L. Fazio (Eds.), Play in Occupational Therapy for Children (2nd ed., pp. 375–393). USA: Mosby Elsevier.
  11. Lang, R., Reilly, M. O., Rispoli, M., Shogren, K., Machalicek, W., Education, S., … Regester, A. (2016). Review of Interventions to Increase Functional and symbolic Play in Children with Autism. Education and Training in Developmental Disabilities, 44(4).
  12. Besio, S., Carnesecchi, M., & Converti, R. (2013). Prompt-fading strategies in robot mediated play sessions. Assistive Technology Research Series, 33, 143–148.
  13. Encarnação, P., Alvarez, L., Rios, A., Maya, C., Adams, K., & Cook, A. (2014). Using virtual robot-mediated play activities to assess cognitive skills. Disability and Rehabilitation. Assistive Technology, 9(3), 231–41.
  14. Cook, A., Bentz, B., Harbottle, N., Lynch, C., & Miller, B. (2005). School-based use of a robotic arm system by children with disabilities. IEEE Transactions on Neural Systems and Rehabilitation Engineering, 13(4), 452–460.
  15. Savage, M. N., & Taber-Doughty, T. (2017). Self-operated auditory prompting systems for individuals with intellectual disability: A meta-analysis of single-subject research. Journal of Intellectual & Developmental Disability, 42(3), 249–258.
  16. Sheridan, T. (1992). Automation and Human Supervisory Control. The MIT Press.
  17. Portney, L., & Watkins, M. (2015). Foundations of Clinical Research: Applications to practice (3rd ed.). New Jersey: Prentice-Hall.