Design and Development of HRI-based Intervention for ASD Children using ADDIE Model and ABA: A Preliminary Study Sarah Afiqah Mohd Zabidia, Hazlina Md. Yusofb, Shahrul Naim Sidekc, Aimi Shazwani Ghazalid a International Islamic University Malaysia, Jalan Gombak, Kuala Lumpur, 0000-0002-9744-0429, Malaysia b International Islamic University Malaysia, Jalan Gombak, Kuala Lumpur, 0000-0002-6349-2028, Malaysia c International Islamic University Malaysia, Jalan Gombak, Kuala Lumpur, 0000-0002-3204-1347, Malaysia d International Islamic University Malaysia, Jalan Gombak, Kuala Lumpur, 0000-0001-5042-3536, Malaysia Abstract The growing interest in Human-Robot Interaction (HRI) based early intervention therapies for autism spectrum disorder (ASD) children brings about the prospect of adapting this platform as assistive tool during therapy. Though this field has garnered myriad of attention, more research is needed in developing systematic design for human-robot interaction using standardized process models. One of the most well-established systematic process models in developing instructions is the Analysis, Design, Develop, Implement and Evaluate (ADDIE) model. While the ADDIE model was used to design the interaction during the first phase, Applied Behavioral Analysis (ABA) technique, an evidence-based practice for ASD children were also integrated into our modified model. This preliminary study discusses three subjects: 1) guideline in adapting ADDIE model for HRI-based ASD intervention, 2) a preliminary modified ADDIE model for human-robot interaction with ASD children and 3) robot’s behavior integrated with ABA technique. Preliminary evaluations on the developed framework were done by a therapist with experience working with ASD children on the severe end of the spectrum. Keywords Autism Spectrum Disorder, ASD, Human-robot interaction, HRI, Applied Behavioral Analysis, ABA 1. Introduction Autism Spectrum Disorder (ASD) is a pervasive neurodevelopmental disorder that is characterized by difficulties with social and communicative skills, restricted interests, repetitive behavior, and sensory issues [1]. Current prevalence data estimated that 16.8 per 1,000 (one in 59) children aged eight years in 2016 has autism. This is approximately 2.5 times higher than the first ADDM Network ASD prevalence estimates of 6.7 (one in 150) from 2000 and 2004 [2]. As there are no known medical cure for ASD, main method for interventions are focused on behavioral modifications such as Applied Behavioral Analysis (ABA)[3], TEACCH Autism Program [4], LEAP [5] and Sensory Integration Therapy [6]. This has also caused for increasing popularity in utilizing technologies, especially robots as assistive tools for educators/therapies during therapy. Although there are various robotic platforms available, studies on ASD children in this domain usually focuses on humanoid robotic research [7] due to their ability to mimic, behave and interact like a human. Educators also assumed that ASD children will be able to generalize skills learned with robots due to their consistent predictability and human-like features [8]. However, as many studies are still focusing on robots’ technologies, more research is needed in systematic Human-Robot interaction (HRI) development using process models. This step is important in ensuring HRI-based interventions are accepted as an evidence-based practice (EBP) by the clinical community. ____________________________________ Proceedings of the Third Symposium on Psychology-Based Technologies (PSYCHOBIT2021), October 4–5, 2021, Naples, Italy EMAIL: sarahafiqah,zabidi@live.iium.edu.my (SAM Zabidi) ; myhazlina@iium.edu.my (HM Yusof); snaim@iium.edu.my (SN Sidek); aimighazali@iium.edu.my (AS Ghazali) ORCID: 0000-0002-9744-0429 (SAM Zabidi); 0000-0002-6349-2028 (HM Yusof); 0000-0002-3204-1347 (SN Sidek); 0000-0001-5042- 3536 (AS Ghazali) ©️ 2021 Copyright for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0). CEUR Workshop Proceedings (CEUR-WS.org) In instructional design community, the Analysis, Design, Develop, Implement and Evaluate (ADDIE) model is one of the most widely used systematic process model. There are five elements to ADDIE model: analysis, design, develop, implement, and evaluate. This method was chosen as the five elements are ongoing activities that continue throughout the process. The process can be referred to in Figure 1. Clear and effective intervention programs can be designed using this method. Although it is originally developed to be hierarchical, it can also be tailored to be a continuous ‘iterative’ approach. The first five phases are followed in order, then, once complete, the data obtained can be used as a guideline and researchers may restart from the analysis phase, improving the final product. Considering that HRI-based intervention requires continuous rigorous tests and evaluation, the systematic nature of ADDIE process model can be applied to the HRI-based intervention for future performance improvement. Analysis Evaluation Design Implementation Development Figure 1. The ADDIE model. While the ADDIE model was used during development and design of the research, Applied Behavioral Analysis (ABA) technique were used to program robot’s behavior. This is to ensure the robot is predictable and consistent, which is important in effective learning for ASD children. ABA is based on operant conditioning where consequences influence behavior. The first step in ABA clinical practice is to perform mental/developmental assessments for ASD children (ASDC) in determining the baseline and appropriate goals for each individual. The step is followed by the development of a clear intervention plan which includes clearly defined procedures for instruction, error correction, prompt levels, reinforcement, and performance data collection. Our project proposes on the development of HRI-based intervention by utilizing the systematic ADDIE model during research development and building framework for robot’s behavior using ABA’s core technique. 2. Human-Robot Interaction (HRI) HRI is an interdisciplinary study of interaction dynamics between humans and robots [9]. HRI has been recognized as a new probable approach in the research on autism research. It is the changing relationship between intelligent robots and humans, which are done through social interaction. HRI can be described both by the user’s behaviors and by the role of the robots during the therapy session. The goal of HRI for ASDC involves encouraging imitative behaviors, mediating turn-taking, extracting referencing and enhancing joint attention between ASDC and another human. 2.1. Robots for ASD children As mentioned in many studies, the use of HRI for ASDC is not to replace therapists, but the main aim is to successfully integrate robot into a normal therapy as a mediator between therapist and ASDC [10,11]. Based on previous literatures, various types of robot were used. The robots are generally used to improve foundational skills such as imitation (I), joint attention (JA), social (S) and other (O) skills. Other skills taught by the robot include reading, comprehension, and literacy skills. Example of robots used in the earlier studies, and the skills targeted can be referred to in Table 1. Table 1. List of current robots used and the skills targeted. Robot JA S I O Studies NAO / / / / [12–17] Pleo / [18] Probo / [11] QTRobot / / [19] Zeno / / / [20–22] RoboParrot / [23] TurtleBot / [24] Popchilla / [25] IROMEC / [26] 3. The ADDIE Model ADDIE model is the most common model used in instructional field design. This is due to it being generic enough to create any type of learning experience for many learners. Although this model is usually used by training instructors, we believe the model can be modified according to ASD children’s needs. The systematic nature of this model may help facilitate researchers in the HRI and ASD domain to move in the correct direction. The model comprises of five phases as shown in Figure 1. Each phase is critical as the researcher must make crucial decisions after thorough analyzation to deliver an effective training/intervention. Figure 2 defines each phase of the model. The entire intervention can be designed thoroughly by following the guideline. ADDIE model for HRI were referred from [27] and were further improvised to suit our research objectives. This phase involves: In this phase, the During development, the For implementation, the In the last phase, the specification of issues in researcher selects and designer develops and researcher conducts the HRI effectiveness of the HRI-based intervention for documents the objectives, completes the actual session to specified target interaction and the ASD (problem statement), evaluation tools (outcome creation of the HRI to meet (in this case: ASD children). achievement of goals are HRI goals and objectives are measures), HRI plan and the interaction objectives. A In this phase, training is evaluated. One unique specified, understanding of control modalities. This step by step procedure for provided to the robot feature of this model is that Analysis Design Development Implementation Evaluation research gaps, robot’s process must be both implementation of the operator (should it not be it can be used iteratively. requirement, systematic and specific. interaction strategy is the same person For example, feedback from understanding of the gaps completed. This phase also developing the robot) on the evaluation phase can be between current condition includes the developmental the operation. used right back in the and desired outcomes. A Output: Module of work, debugging, testing, Analysis phase, which starts quality analysis can interaction, overall design review and revision of an entirely new iteration of contribute to identifying of robot’s behavior and strategies. Output: Conduct the HRI- the end product. learning goals and needs, outline of based intervention with objectives. intervention, outcome ASD children. measures. Output: A completed HRI Output: Detailed evaluation platform. on the efficacy of the Output: Clear views on the intervention, information research gaps, objectives, on aspects that may need desired outcomes. revision and can be used for to improvise future research. Figure 2. Details for each elements/phase in the ADDIE model. 3.1. Applied Behavioral Analysis (ABA) ABA is based on operant conditioning where consequences influence behavior. The first step in ABA clinical practice is to perform assessments in determining the appropriate goals for each individual using several tools such as the Verbal Behavior Milestones Assessment and Placement Program [28] and Vineland Adaptive Behavior Scale [29]. Assessment is then followed by the development of a clear intervention plan which includes clearly defined procedures for instruction, error correction, prompt levels, reinforcement, and performance data collection. 3.1.1. Teaching Technique (Discrete Trial Training) ABA approach uses discrete trial teaching (DTT) to teach skills. Discrete trials or the three- part teaching unit is a special behavioural sequence used to maximize learning. DTT is used to make teaching session clearer, let the child know when he/she is right or wrong, helps teacher maintain consistency and makes assessment of progress simpler [30]. Operant conditioning implies that a behaviour that results in something that is liked (reinforcement) will be repeated [31]. As reinforcer is something that is liked by ASDC, the reinforcers are a critical tool to be embedded in ABA program. The reinforcers are given as consequence whenever the child respond correctly. Figure 3 shows the sequence of DTT and its’ component. Discriminative Stimulus Response Consequence /Instruction (R) (𝑆 𝐶 ) (𝑆 𝐷 ) Figure 3. Components of DTT. 3.1.2. Prompting Techniques As shown in Figure 3, prompting is a means to induce an individual with added stimuli (prompts) to perform a desired behavior. Prompting is provided when an ordinary antecedent is ineffective and is extensively used in behavior shaping and skill acquisition. It provides learners with assistance to increase the probability that the desired behavior will occur. Successful performance of a desired behavior elicits positive reinforcement, therefore reinforcing learning. A prompt is used as a cue to support and encourage a desired behavior that otherwise does not occur. Discriminative Stimulus Response Consequence /Instruction (R) (𝑆𝐶 ) (𝑆 𝐷 ) Prompt (P) Figure 4. Prompting as part of DTT. 3.1.3. Application of ABA in HRI Existing literatures were reviewed to identify studies that adapted ABA technique into HRI applications in the past. There are a few techniques that have been frequently used by researchers in the past which are reinforcement (R), prompt levels (PL) and discrete-trial teaching (DTT) technique. Results may be referred to in Table 2. Table 2. ABA-based HRI studies. Studies Robot R PL DTT [32] NAO / [33, 34] NAO / / [7,33,35] NAO / / [36] PABI / [37] NAO / / [38–40] NAO / / [41, 42] NAO / / [43] KASPAR / / [44] KASPAR / / [45] ReRO / [46] NAO / / / All studies reviewed applied the reinforcement technique (reward). The reward can be in the form of social praise (“Good job!”, “Well done!”) or music played by the robot. Prompt levels are also one of the most applied ABA techniques. If the child answered wrongly, the robot will ask the child to try again and prompt will be provided. DTT is a structured teaching technique used to help ASD child learn effectively. However, there are more to DTT than following the special behavioral sequence. An effective DTT must follow the correct technique in terms of differential reinforcement, intonation, and timing. [47] focused more on how to use social stories and visual schedule instead of using those two techniques in a therapy. 4. Preliminary output based on ADDIE model. 4.1. Analysis Phase In this phase, our focus lies on analyzing and identifying test subjects to determine interaction goals and learning context. More precisely, we assessed the developmental age of the children by using Vineland Adaptive Behavior Scale (VABS-3) to determine suitable module of interaction. Having a clear view on the baseline of each child is important as no ASDC is the same and their chronological age does not necessarily determine their developmental/mental age. Knowing the baseline can guide decision for the next phase and provide realistic approach for the intervention. Therefore, all available information on the ASD children from a center in Malaysia were identified. Three main issues that were analyzed in this phase include: a) Selection of center Due to the movement-controlled order (MCO) imposed by Malaysian government during the period of this study, a center for ASD children located near the university were chosen. This issue has limited the sample size of this study. b) Target age of ASD children In the center chosen, although all ASD children were diagnosed with ASD, no records on their exact developmental age were available. Therefore, help from the clinical team were obtained to administer VABS-3 to the children. While this center consists of child aged from 4-9 years old, only child with physical age 6-9 were selected as participants. Participants went through VABS-3 assessment and developmental age for all participants were obtained. c) Selection of robot Existing literatures were reviewed to obtain the most optimal robot suitable for ASD children. Studies in HRI for ASDC combines both robotic and clinical teams, therefore buying a commercial robot is the better choice to bridge this gap [18]. Besides, parents, clinicians and teachers can easily buy and customize the robot for certain needs due to the availability of the robot in the market. To summarise, commercial robots have more advantages in terms of cost, robustness, and their lower failure rate. During selection phase, recommendations from previous studies were adapted for inclusion criteria. Some commercial robots that have been shortlisted are Bioloid, DARWIN, NAO and QTRobot. However, for this study, Bioloid and Darwin were eliminated due to its small size (39.7cm and 45.5cm respectively). As our target audience are ASDC aged 6-9 years old, robot with a similar size with the child will be used. This is so that the child can see the robot at his or her eye-level. In addition, imitation tasks are intuitively easier for ASDC to follow compared to using robots that are extremely large or small in size [48]. The robot must also be easily integrated into the existing therapy framework and agenda (objectives, setting, timing, etc.). To conclude, the selection criteria for the work are as follows: (a) The robot must be easily programmable by therapists (has block-based coding); (b) Has developer mode for researchers to embed a more complex program using programming languages such as Python/C++/Java; (c) Able to display facial expressions, changes in emotional expression in the robot must be able to be subtly observed. Although both NAO and QTRobot has almost the same features, QTRobot was chosen because of its’ ability to subtly show emotional expression. The importance of this aspect is discussed in the next sub-section. 4.2. Design Phase In this phase, learning objectives were defined and modules of interactions were also constructed. Based on our preliminary VABS-3 assessment, most of the participants were mostly lacking in the social domain. Therefore, learning objectives and modules of interaction were designed to focus on learning emotions as it is an important aspect in building social reciprocity. Table 3. Modules of interaction developed based on findings from analysis phase. Modules Methods Objectives Introduction Robot slowly starts movement To measure robot’s according to child’s response and approachability and create a one-way communication using friendly environment for the voice recognition. child during interaction. Module 1 Robot will perform simple To assess child’s (Identify emotion) actions and child are expected to engagement with the robot and imitate. child’s ability to listen to robot’s instruction. Module 2 Robot asks the learner to (Express emotion) identify the picture that is the correct match of a sample picture. Goodbye Robot says goodbye and wave To communicate robot’s its arms. Thanks child and attempts limitation to the child. handshake. 4.3. Development Phase 4.3.1. Robotic Platform (QTRobot) QTRobot is a 58cm tall humanoid robot that is. QTRobot Studio were initially used to program QTRobot’s behaviours. QTRobot Studio is a development environment provided by the robot manufacturer, LUX AI. The interface is mainly drag and drop, and allows the programmer to create a sequenced combination of predefined or custom behavior boxes to manipulate the QTRobot’s joints or attributes. QTrobot has an expressive social appearance and its screen allows the presentation of animated faces. It has 12 degrees of freedom to present upper-body gestures. Eight degrees of freedom are motor-controlled, two in each shoulder, one in each arm plus pitch and yaw movements of the head. The other four degrees of freedom, one in each wrist and one in each hand, are manually configured. QTrobot has a RealSense 3D camera mounted on its forehead and is provided with a microphone array. QTrobot is powered with an Intel NUC processor and Ubuntu 16.04 LTS, and provides a native ROS interface to program it in Python or C++ programming languages. QTrobot also provides a visual programming interface for IT non-experts, used in this study, to easily script custom applications and control the robot by an Android application from tablets and smart phones. In the present study, the robot’s interactions with the children were pre-scripted and controlled by the experimenter via a tablet. During the development phase, this robot will first be programmed using QTRobot Studio to test for the framework validity. Feedback on the framework validity will be done by a therapist. The equivalent module will then be converted to a Python/C++ equivalent program in ROS environment. 4.3.2. System Architecture The system architecture may be referred to in Figure 5. A researcher will operate the supervisory controller for procedure control, robot feedback decision and data acquisition. Figure 5. System Architecture. 4.3.3. Robot Teaching Structure Note that the robot follows the DTT sequence (Sd -> R -> Sc). Prompt level (PL) was programmed to help child feel successful. Reinforcement given in this context is social praise, gesture an expression. The previous three actions have categorized according to prompt levels as differential reinforcement Table 4. ASD child values consistency, the robot was made sure to give consistent feedback to the child based on their response. This aspect was something that the therapist considers important, as this is even difficult to maintain by experienced therapist. Consistency by human therapist may vary according to their level of stress, moods, or environment. Table 4. Differential reinforcement technique. Prompt Level Expression Gesture Tone/Praise No Prompt Clap Enthusiastic. (Exclamation Hands up mark used) First Prompt Clap Less enthusiastic response. (no exclamation mark) Second Prompt No gesture Praises are ended with ‘try’ at the end of sentence to signify this is not the expected behavior. Figure 6. Example of programming block for correct response. Figure 7. Example flowchart for imitation's interaction module. Figure 7 summarizes the flow of robot’s behavior. Sentence with exclamation mark at the end will be said by the robot with an enthusiastic tone. To summarize, the more PL the child need, the less enthusiastic the robot’s response will be. Being consistent in this aspect will help child learn targeted behaviors effectively. Even when the child does not give the correct answer, the robot will help ASD child feel successful by giving prompts. 4.4. Implementation Phase This phase involves the interaction between robot and ASDC. The experiment will be conducted at the child’s center to maintain familiarity and ensure ASDC is comfortable. Researcher will be hidden to minimize the number of unfamiliar people during HRI session. The flow of intervention, preparation and challenges that might emerge during intervention must also be considered. 4.5. Evaluation Phase The interactions will be recorded via Intel Realsense D435 embedded with the robot. The child’s behavior will be analyzed via post session video analysis. In measuring the behavioral engagement, proxemics imaging method is used to measure the distance between the subject and the robot to identify their interest with the robot and levels of engagement throughout intervention. The same method was previously applied by [49] and [50] who also measured the engagement between children with autism and a humanoid robot using distance. Proxemics behavior includes interpersonal distancing, eye gazing, and body gesture that can express the behavior of the person towards the engagement. 5. Conclusion The present work introduces the design of a HRI platform aimed for ASD children by using ADDIE model and ABA. Our purpose is to support children with ASD and the educators by introducing robots as assistive tool. We would also like to propose for future researchers in this discipline to follow a more standardized model to increase replicability and acceptability by the clinical community with our design. 6. Acknowledgements The authors gratefully acknowledge the Ministry of Education Malaysia (MOE) for funding the research project through the Transdisciplinary Research Grant Scheme (TRGS) [Ref. No TRGS/1/2019/UIAM/02/4/3]. 7. References [1] M. DeFilippis and K. D. Wagner, “Treatment of Autism Spectrum Disorder in Children and Adolescents.,” Psychopharmacol. Bull., vol. 46, no. 2, pp. 18–41, Aug. 2016. [2] M. J. Maenner et al., “Prevalence of Autism Spectrum Disorder Among Children Aged 8 Years- Autism and Developmental Disabilities Monitoring Network, 11 Sites, United States, 2016 Centers for Disease Control and Prevention MMWR Editorial and Production Staff (Serials) MMWR Editor,” Morb. Mortal. Wkly. Rep. Surveill. Summ., vol. 69, no. S-4, pp. 1–12, 2020, doi: http://dx.doi.org/10.15585/mmwr.ss6904a1. [3] O. I. Lovaas, “Behavioral treatment and normal educational and intellectual functioning in young autistic children.,” Journal of Consulting and Clinical Psychology, vol. 55, no. 1. American Psychological Association, US, pp. 3–9, 1987, doi: 10.1037/0022-006X.55.1.3. [4] S. Ozonoff and K. Cathcart, “Effectiveness of a home program intervention for young children with autism.,” J. Autism Dev. Disord., vol. 28, no. 1, pp. 25–32, Feb. 1998, doi: 10.1023/a:1026006818310. [5] P. S. Strain and M. Hoyson, “The Need for Longitudinal, Intensive Social Skill Intervention: LEAP Follow-Up Outcomes for Children with Autism,” Topics Early Child. Spec. Educ., vol. 20, no. 2, pp. 116–122, Apr. 2000, doi: 10.1177/027112140002000207. [6] N. Ghotbi, N. N. Ansari, S. Naghdi, S. Hasson, B. Jamshidpour, and S. Amiri, “Inter-rater reliability of the Modified Modified Ashworth Scale in assessing lower limb muscle spasticity,” Brain Inj., vol. 23, no. 10, pp. 815–819, Jan. 2009, doi: 10.1080/02699050903200548. [7] Z. Zheng, H. Zhao, A. R. Swanson, A. S. Weitlauf, Z. E. Warren, and N. Sarkar, “Design, Development, and Evaluation of a Noninvasive Autonomous Robot-Mediated Joint Attention Intervention System for Young Children with ASD,” IEEE Trans. Human-Machine Syst., vol. 48, no. 2, pp. 125–135, 2018, doi: 10.1109/THMS.2017.2776865. [8] A. M. Alcorn et al., “Educators’ Views on Using Humanoid Robots With Autistic Learners in Special Education Settings in England.,” Front. Robot. AI, vol. 6, p. 107, 2019, doi: 10.3389/frobt.2019.00107. [9] A. Zaraki, L. Wood, B. Robins, and K. Dautenhahn, “Development of a Semi-Autonomous Robotic System to Assist Children with Autism in Developing Visual Perspective Taking Skills,” in RO-MAN 2018 - 27th IEEE International Symposium on Robot and Human Interactive Communication, 2018, pp. 969–976, doi: 10.1109/ROMAN.2018.8525681. [10] S. Costa, “Robots as Tools to Help Children with ASD to Identify Emotions,” Autism. Open. Access, vol. 04, no. 02, pp. 8–9, 2014, doi: 10.4172/2165-7890.1000e120. [11] R. E. Simut, J. Vanderfaeillie, A. Peca, G. Van de Perre, and B. Vanderborght, “Children with Autism Spectrum Disorders Make a Fruit Salad with Probo, the Social Robot: An Interaction Study,” J. Autism Dev. Disord., vol. 46, no. 1, pp. 113–126, 2016, doi: 10.1007/s10803-015- 2556-9. [12] Z. Zheng, G. Nie, A. Swanson, A. Weitlauf, Z. Warren, and N. Sarkar, “A Randomized Controlled Trial of an Intelligent Robotic Response to Joint Attention Intervention System,” J. Autism Dev. Disord., 2020, doi: 10.1007/s10803-020-04388-5. [13] D. O. David, C. A. Costescu, S. Matu, A. Szentagotai, and A. Dobrean, “Effects of a Robot- Enhanced Intervention for Children With ASD on Teaching Turn-Taking Skills,” J. Educ. Comput. Res., vol. 58, no. 1, pp. 29–62, 2020, doi: 10.1177/0735633119830344. [14] U. Qidwai, S. B. A. Kashem, and O. Conor, “Humanoid Robot as a Teacher’s Assistant: Helping Children with Autism to Learn Social and Academic Skills,” J. Intell. Robot. Syst. Theory Appl., 2019, doi: 10.1007/s10846-019-01075-1. [15] E. Y. han Chung, “Robotic Intervention Program for Enhancement of Social Engagement among Children with Autism Spectrum Disorder,” J. Dev. Phys. Disabil., vol. 31, no. 4, pp. 419–434, 2019, doi: 10.1007/s10882-018-9651-8. [16] E. Alarfaj, H. Alabdullatif, H. Alabdullatif, G. Albakri, and N. S. A. Karim, Analysis of the use of a nao robot to improve social skills in children with ASD in Saudi Arabia, vol. 868. 2018. [17] D. Conti, A. Di Nuovo, G. Trubia, S. Buono, and S. Di Nuovo, “Adapting Robot-Assisted Therapy of Children with Autism and Different Levels of Intellectual Disability: A Preliminary Study,” ACM/IEEE Int. Conf. Human-Robot Interact., pp. 91–92, 2018, doi: 10.1145/3173386.3176962. [18] E. Kim, R. Paul, F. Shic, and B. Scassellati, “Bridging the Research Gap: Making HRI Useful to Individuals with Autism,” J. Human-Robot Interact., vol. 1, no. 1, pp. 26–54, 2012, doi: 10.5898/jhri.1.1.kim. [19] A. P. Costa et al., “More Attention and Less Repetitive and Stereotyped Behaviors using a Robot with Children with Autism,” RO-MAN 2018 - 27th IEEE Int. Symp. Robot Hum. Interact. Commun., no. September, pp. 534–539, 2018, doi: 10.1109/ROMAN.2018.8525747. [20] G. Palestra, G. Varni, M. Chetouani, and F. Esposito, “A multimodal and multilevel system for robotics treatment of autism in children,” 2016, doi: 10.1145/3005338.3005341. [21] V. Banthia et al., “Development of a graphical user interface for a socially interactive robot: A case study evaluation,” 2016, doi: 10.1109/IEMCON.2016.7746294. [22] K. Jordan, M. King, S. Hellersteth, A. Wirén, and H. Mulligan, “Feasibility of using a humanoid robot for enhancing attention and social skills in adolescents with autism spectrum disorder,” Int. J. Rehabil. Res., vol. 36, no. 3, pp. 221–227, 2013, doi: 10.1097/MRR.0b013e32835d0b43. [23] P. Soleiman, S. Salehi, M. Mahmoudi, M. Ghavami, H. Moradi, and H. Pouretemad, “RoboParrot: A robotic platform for human robot interaction, case of autistic children,” in 2014 2nd RSI/ISM International Conference on Robotics and Mechatronics, ICRoM 2014, 2014, pp. 711–716, doi: 10.1109/ICRoM.2014.6990987. [24] J. Lee, H. Aoki, D. Stefanov, T. Yamamoto, and G. Obinata, “A study on the relationship between robotic movement with animacy and visual attention of young children,” 25th IEEE Int. Symp. Robot Hum. Interact. Commun. RO-MAN 2016, pp. 682–687, 2016, doi: 10.1109/ROMAN.2016.7745192. [25] O. Hawks, C. J. Dunst, D. W. Hamby, C. M. Trivette, J. Prior, and G. Derryberry, “Effects of a Socially Interactive Robot on the Conversational Turns Between Parents and Their Young Children with Autism,” no. 6, pp. 1–8, 2013. [26] I. Iacono, H. Lehmannt, P. Marti, B. Robinst, and K. Dautenhahnt, “Robots as Social Mediators for Children with Autism - A Preliminary Analysis Comparing Two Different Robotic Platforms,” 2011 IEEE Int. Conf. Dev. Learn. ICDL 2011, 2011. [27] R. E. Mohan, C. A. A. Calderon, C. Zhou, and T. Yang, “Adapting ADDIE Model for Human Robot Interaction in Soccer Robotics Domain,” pp. 166–176. [28] M. L. Sundberg, VB-MAPP Verbal Behavior Milestones Assessment and Placement Program : a language and social skills assessment program for children with autism or other developmental disabilities : guide LK - https://Iiumx6.on.worldcat.org/oclc/298338098. Concord, CA SE - vii, 239 pages : color illustrations, color forms ; 28 cm + Protocol (72 pages : color illustrations, color forms ; 28 cm): AVB Press, 2008. [29] K. L. Salekin, T. M. S. Neal, and K. A. Hedge, “Validity, interrater reliability, and measures of adaptive behavior: Concerns regarding the probative versus prejudicial value.,” Psychol. Public Policy, Law, vol. 24, no. 1, pp. 24–35, 2018, doi: 10.1037/law0000150. [30] K. Lear, Help us learn : a self-paced training program for ABA LK - https://Iiumx6.on.worldcat.org/oclc/49629919. Toronto, Ont. SE - 2 volumes (loose-leaf, in binders) : illustrations ; 30 cm + 1 computer disc: [publisher not identified], 2000. [31] M. Wolf, T. Risley, M. Johnston, F. Harris, and E. Allen, “Application of operant conditioning procedures to the behavior problems of an autistic child: A follow-up and extension,” Behav. Res. Ther., vol. 5, no. 2, pp. 103–111, 1967, doi: 10.1016/0005-7967(67)90004-6. [32] M. Dimitrova, N. Vegt, and E. Barakova, “Designing a system of interactive robots for training collaborative skills to autistic children,” 2012 15th Int. Conf. Interact. Collab. Learn. ICL 2012, no. July 2014, 2012, doi: 10.1109/ICL.2012.6402179. [33] Z. Zheng et al., “Impact of robot-mediated interaction system on joint attention skills for children with autism,” IEEE Int. Conf. Rehabil. Robot., 2013, doi: 10.1109/ICORR.2013.6650408. [34] E. Bekele et al., Multimodal interfaces and sensory fusion in VR for social interactions, vol. 8525 LNCS, no. PART 1. 2014. [35] Z. Zheng, S. Das, E. M. Young, A. Swanson, Z. Warren, and N. Sarkar, “Autonomous robot- mediated imitation learning for children with autism,” in Proceedings - IEEE International Conference on Robotics and Automation, 2014, pp. 2707–2712, doi: 10.1109/ICRA.2014.6907247. [36] L. Dickstein-Fischer and G. S. Fischer, “Combining psychological and engineering approaches to utilizing social robots with children with Autism,” in 2014 36th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBC 2014, 2014, pp. 792–795, doi: 10.1109/EMBC.2014.6943710. [37] S. M. Anzalone et al., “How children with autism spectrum disorder behave and explore the 4- dimensional (spatial 3D + time) environment during a joint attention induction task with a robot,” Res. Autism Spectr. Disord., vol. 8, no. 7, pp. 814–826, 2014, doi: 10.1016/j.rasd.2014.03.002. [38] M. A. Miskam, S. Shamsuddin, M. R. A. Samat, H. Yussof, H. A. Ainudin, and A. R. Omar, “Humanoid robot NAO as a teaching tool of emotion recognition for children with autism using the Android app,” 2014 Int. Symp. Micro-NanoMechatronics Hum. Sci. MHS 2014, pp. 2–6, 2015, doi: 10.1109/MHS.2014.7006084. [39] S. Shamsuddin, N. A. Malik, H. Yussof, S. Mohamed, F. A. Hanapiah, and F. W. Yunus, “Telerehabilitation in robotic assistive therapy for children with developmental disabilities,” in IEEE TENSYMP 2014 - 2014 IEEE Region 10 Symposium, 2014, pp. 370–375. [40] S. Shamsuddin, H. Yussof, L. I. Ismail, S. Mohamed, F. A. Hanapiah, and N. I. Zahari, “Humanoid robot NAO interacting with autistic children of moderately impaired intelligence to augment communication skills,” in Procedia Engineering, 2012, vol. 41, pp. 1533–1538, doi: 10.1016/j.proeng.2012.07.346. [41] M. Begum, R. W. Serna, and H. A. Yanco, “Are Robots Ready to Deliver Autism Interventions? A Comprehensive Review,” Int. J. Soc. Robot., vol. 8, no. 2, pp. 157–181, 2016, doi: 10.1007/s12369-016-0346-y. [42] M. Begum et al., “Measuring the Efficacy of Robots in Autism Therapy: How Informative are Standard HRI Metrics’,” in ACM/IEEE International Conference on Human-Robot Interaction, 2015, vol. 2015-March, pp. 335–342, doi: 10.1145/2696454.2696480. [43] G. Palestra et al., Social robots in postural education: A new approach to address body consciousness in ASD children, vol. 8755. 2014. [44] M. J. Salvador, S. Silver, and M. H. Mahoor, “An emotion recognition comparative study of autistic and typically-developing children using the zeno robot,” in Proceedings - IEEE International Conference on Robotics and Automation, 2015, vol. 2015-June, no. June, pp. 6128–6133, doi: 10.1109/ICRA.2015.7140059. [45] N. I. Ishak, H. M. Yusof, S. N. Sidek, and Z. Jaalan, “Interactive robotic platform for education and language skill rehabilitation,” 2017 IEEE Int. Conf. Smart Instrumentation, Meas. Appl. ICSIMA 2017, vol. 2017-Novem, no. November, pp. 1–5, 2018, doi: 10.1109/ICSIMA.2017.8312031. [46] W. Y. Geoffrey Louie, J. Korneder, I. Abbas, and C. Pawluk, “A study on an applied behavior analysis-based robot-mediated listening comprehension intervention for ASD,” Paladyn, vol. 12, no. 1, pp. 31–46, 2020, doi: 10.1515/pjbr-2021-0005. [47] P. Chevalier, J.-C. Martin, B. Isableu, and A. Tapus, “Impact of personality on the recognition of emotion expressed via human, virtual, and robotic embodiments,” in Proceedings - IEEE International Workshop on Robot and Human Interactive Communication, 2015, vol. 2015- Novem, pp. 229–234, doi: 10.1109/ROMAN.2015.7333686. [48] N. Giullian, D. Ricks, A. Atherton, M. Colton, M. Goodrich, and B. Brinton, “Detailed requirements for robots in autism therapy,” Conf. Proc. - IEEE Int. Conf. Syst. Man Cybern., pp. 2595–2602, 2010, doi: 10.1109/ICSMC.2010.5641908. [49] E. Short, D. Feil-Seifer, and M. Matarić, “A comparison of machine learning techniques for modeling human-robot interaction with children with autism,” HRI 2011 - Proc. 6th ACM/IEEE Int. Conf. Human-Robot Interact., pp. 251–252, 2011, doi: 10.1145/1957656.1957756. [50] M. D. Manner, M. Gini, and J. Elison, “Graphically representing child-robot interaction proxemics,” Work. Thirty-Second AAAI Conf. Artif. Intell. Graph., pp. 470–478, 2018.