1

IESD CASE STUDY:

CHILDREN ON THE AUTISM SPECTRUM SHOW IMPROVEMENT WITH ROBOTS4AUTISM IN SPARTANBURG, SOUTH CAROLINA

July 6, 2016

School Profile McCarthy Teszler School, Spartanburg, South Carolina Serving special needs children ages 3-20 Enrollment: approximately 240, but changes from week to week Draws students from 76 schools across 7 districts whose needs can’t be met in the home school

The Challenge Educators at McCarthy Teszler School in Spartanburg, South Carolina, have a lot of experience working with children on the autism spectrum. Drawing from 76 schools across 7 local districts, McCarthy Teszler provides services for special education students from preschool through early adult whose needs can’t be met effectively in their home schools. At any given time, this may include 90 or more students with diagnosed problems on the autism spectrum—the most challenging cases in the county Despite the best efforts of school staff, many of the most seriously challenged students weren’t experiencing success with existing approaches. So school leaders started looking for a new curriculum that would address the key areas of social skills, communication skills, and emotion regulation. Ideally, the approach would also incorporate support from technology but would not be strictly computer-based or tablet-based, since as one local specialist explained, “Left to their own devices, children with autism tend to interact with technology only and avoid human interaction.” The answer that McCarthy Teszler educators found was Robots4Autism, an autism intervention program developed by Robokind™ in collaboration with top autism experts.

The Solution Robots4Autism integrates a variety of evidence-based practices shown to improve important skills in students with autism, packaged together systematically and delivered using Milo—a highly expressive, advanced social robot designed specifically to teach critical skills to children with autism. The curriculum is delivered through verbal interactions and social narratives between the student and Milo, who is connected to a student iPad where the robot can display multiple choice options and show supporting text, images, and video modeling to enhance the lesson. Lessons can be repeated as many times as needed to help the student develop a specific skill.

IESD Case Study on Robots4Autism, 7/6/2016 2

An important reason why McCarthy Teszler chose Robots4Autism was the content focus. According to Elena Ghionis, lead autism specialist for Spartanburg County, Robots4Autism “was the only curriculum we found that combined instruction in all three key areas of need: social skills, communication skills, and emotion regulation.” Another key selling point was Milo itself. Ghionis explains that Milo “provides children with an almost human interaction experience”:

Milo is a bridge between a child interacting with technology and interacting with humans. Children who start out avoiding human interaction are attracted to Milo; they practice skills and appropriate behavior with him until they are ready to transition to engaging with the teacher and other children.

Implementation Autism specialists who were chosen to work with the program as facilitators participated in a live distance training session, in which they watched a demonstration of how to use Milo, then worked with their local Milo robot while the trainers observed via webcam and provided feedback. After the initial training, other online sessions were held every one to two months to answer questions and have facilitators share their experiences. 17 students in preK through grade 12 were chosen to interact with the program—children for whom other approaches had not been successful. All had large deficits in social skills, communication skills, and emotion regulation. Most were enrolled at McCarthy Teszler, although a few were from other local affiliated schools. Each week, the students would be pulled out for two to three 30-minute one-on-one sessions with an autism specialist who acted as the facilitator. This continued for three-fourths of the school year—about 27 weeks. Each session began with students playing a game involving Milo—Red Light, Green Light was a favorite. Students then completed between one and three modules chosen by the facilitator. During the sessions, students communicated with Milo using symbols on a student iPad, while the facilitator used a separate iPad to control various aspects of the lesson. For example, during lessons on communicating emotions, Milo would demonstrate facial expressions, the facilitator would assess the child’s facial behavior, and Milo would react appropriately. After the child had successfully demonstrated a skill, a short video would play showing a real-life situation for the child to interpret. In the final lesson for each emotion, Milo would ask the student to demonstrate his or her own facial expression into a camera on Milo’s chest, and the student performance was captured in the Robots4Autism data collection system.

Results GARS-3 Data Of the 17 students using Robots4Autism, 8 showed improvements on both the 4-score index and the 6-score index on the GARS-3 (Gilliam Autism Rating Scale, 3rd ed.), comparing scores prior to use of the program with scores afterward. Another 4 students showed reductions specifically in the area of restrictive repetition. Lead autism specialist Elena Ghionis noted that this is a particular benefit of working with Milo:

IESD Case Study on Robots4Autism, 7/6/2016 3

Many children with autism tend to use repetitive behavior as a calming behavior. Milo teaches alternative, socially acceptable, age-appropriate, calming behavior—so children have less need for restrictive repetitive behavior.

Progress Toward IEP Goals Review of the students’ IEP data and information from teacher records by the school’s lead autism specialist showed that during the first nine week quarter before students started using Robots4Autism, all 17 students showed minimal progress toward their IEP goals. However, over the next three quarters after they started using Robots4Autism, students showed significant progress or mastery related to their social, communication, behavioral, and academic goals. General Impressions Autism specialists who worked with Robots4Autism as facilitators thought the program had helped many of their students to • Recognize and communicate about their emotions • Express and regulate their emotions • Apply calming down skills, thereby reducing behavioral issues • Maintain eye contact with other people in social situations • Engage in appropriate, two-sided conversation Specialists commented in particular on the value of Milo as a model for student interactions. For example, one teenage student working as a cashier in the teacher’s cafe was making change with her head down and no eye contact with customers when her teacher asked, “What would Milo say?” Immediately the student turned her head, made eye contact, and said “Here you go” as she was handing back change. Classroom teachers confirmed that students were enthusiastic about working with Milo, and that they were able to see the results of students’ work with Milo in their classrooms. The autism specialists reported that they now consider Robots4Autism an important therapy tool for selected students. Examples of Individual Progress Example #1: An elementary student on a half-day schedule due to severe behavioral issues was able to go to a full-day schedule after working with Robots4Autism. Tantruming and aggression decreased to no more than twice a day, and he is now able to self-regulate his emotions by removing himself from a whole-group setting to sit alone in a calming area. He has improved from being resistant to using the toilet to following a toileting schedule with minimal prompts and very few accidents. Where before he was able to follow along with a story that was being read to him, he can now read some books independently—a change the autism specialist attributes in part to interaction with Milo. Example #2: After working with Robots4Autism, another elementary student, whose only oral speech had involved repeating words and phrases without functional meaning (echolalia), was able to use verbal language for simple forms of communication. Communication using an assistive device also became

IESD Case Study on Robots4Autism, 7/6/2016 4

much more in-depth. The student’s behavioral issues (e.g., screaming, hitting, and kicking) were greatly reduced, and he was able to identify and communicate how he was feeling, where before he had made no attempts to play, interact, or communicate with others. Example #3: One elementary student who did not show a gain on the GARS-3 index nonetheless made an important transition while working with Robots4Autism. Prior to use of the program, he was able to use an iPad as an assistive communication device, but could not speak. After using the program, he was able to speak functionally. Example #4: A high school student who before working with Robots4Autism used to make multiple verbal outbursts in the classroom, did not readily make eye contact, and had problems interacting appropriately with others, has now reduced the number of verbal outbursts she makes, is doing a better job with eye contact, and engages more appropriately with both peers and adults. Where before she was unable to functionally communicate her emotions without shutting down, now she can identify and demonstrate multiple emotions, including happy, sad, angry, hurt, tired, and worried. Her interaction with others has improved to include socially appropriate greetings and reciprocal conversations.

Lessons Learned Learning how to implement Robots4Autism most effectively with students has been a learning process for the educators at the McCarthy Teszler School. In particular, they noted that Robokind was very open to making adjustments to the modules and program features on an ongoing basis in response to problems as they arose. Specific lessons mentioned by the McCarthy Teszler educators include the following: • In order to interact successfully with Milo, students need to be able not only to verbalize but also to

listen and talk/communicate with understanding. For some low-functioning students, the facilitator had to rephrase a question asked by Milo. In other cases, the facilitator had to switch to simpler modules when students couldn’t pick up on small details of the video interaction.

• For five of the lower-functioning students, the facilitator found it useful to preteach symbols used in Robots4Autism before having students use them with Milo.

• While some students loved Milo from the first, others needed more time. For example, one student was initially scared of Milo and wouldn’t go anywhere near the robot. Over time, the facilitator moved the child’s workstation progressively closer to Milo so he could see Milo interacting with another student. Then the facilitator had the student interact with Milo using the iPad from across the room. Now the student loves working with Milo.

The autism specialists and teachers at McCarthy Teszler School are looking forward to continuing their use of Robots4Autism for the benefit of their most needy students on the autism spectrum.

BackgroundIndividuals with Autism Spectrum Disorders (ASD) exhibit significant impairments in development in social relatedness, reciprocal social behavior, social communication, joint atten-tion, and language learning.

Children with ASD have a preference for objects over people (Lombroso et al., 2009) in addition to superior nonsocial skills constructing and analyzing systems (i.e., computers and ro-bots) (Baron-Cohen, 2005; Baron-Cohen, 2009).

Recent research suggests that despite individual differences in performance, children with ASD show more social engage-ment with a robot as compared to humans or other technologi-cal devices (Diehl et al., 2014; Bekele et al., 2013).

PurposeTo examine specific features of the robot-child interaction that may facilitate social engagement and understanding of social situations in children with ASD.

Specifically, to investigate differences in engagement related to the child’s level of functioning (i.e., minimally verbal vs. fluent) and therapy condition (i.e., student-led vs. robot-led) while playing a follow-the-leader type game.

HypothesisBoth minimally verbal and fluent children with ASD will be more engaged with both the therapist and the robot (i.e., Milo) in the student-led condition as compared to the robot-led condition.

MethodParticipants

• 9 children with ASD between 5-14 years •Minimally verbal • Fluent speech

• Inclusion criteria •Understand cause and effect • Understand how to use a tablet to communicate with Milo • Have picture symbol recognition • Answer yes/no questions

Procedure

• Randomized crossover design, with wash-out condition

• Two conditions were utilized: a robot-led condition in which Milo provided simple 1-step commands to the child, and a student-led condition in which the child provided simple 1-step commands to the robot via a tablet.

• The two conditions were separated by a 6-minute wash-out condition in which the child engaged in a low-key interaction with a therapist.

• Sessions were recorded for later analyses.

Analyses

• Engagement was coded based on continuous microanalytic coding with quarter-second precision. If an engagement state was less than a quarter second, it was not counted.

• Each session was coded for the following:

(1) Engagement with Milo

(2) Engagement with a research assistant

(3)Engagement elsewhere (disengaged)

• Due to the small sample size, Cohen’s d, a standardized measure of effect size that is not influenced by sample size like p –values, was used to analyze the data.

ENGAGEMENT WITH THERAPIST ENGAGEMENT WITH MILO

Using a Humanoid Robot as a Co-therapist with Children with ASDAllison Kroiss, B.S., Danielle Sonego, B.S., Pamela Rollins Ed.D, CCC-SLP

Callier Center

RESULTSQuantitative Analyses

• Fluent children had more engagement with the robot compared to minimally verbal children in both conditions.

• While minimally verbal children did not differ by condition, fluent children were moderately more engaged (d =.56) with Milo in the robot-led condition.

• Children rarely engaged with the therapist even though she tried to interact with them.

• When children were not engaged with Milo, they tended to be disengaged.

• Disengagement did not differ across condition for the minimally verbal children; however, the fluent children were more disengaged in the student-led condition (d =.783).

Qualitative Analyses

• Minimally verbal children took longer to warm up to Milo than did the fluent children.

• Once they warmed up, performance resembled the fluent children.

• Fluent children appeared to treat Milo as a friend.

SUMMARY & DISCUSSION Our research found that children with ASD are more engaged with Milo than with the therapist, especially when Milo was instructing them.

And we know children who are more engaged learn better.

Our future research will focus on using Milo to facilitate understanding of social situations and generalization to humans.

REFERENCESBaron-Cohen, Simon. “Autism and the origins of social neuroscience.” In The Cognitive Neuroscience of Social Behaviour. New York: Psychology Press, 2005: 239-255.

Baron-Cohen, Simon. “Autism: The Empathizing–Systemizing (E-S) Theory. “Annals of the New York Academy of Sciences 1156, no. 1, 2009: 68-80.

Bekele, E. T., Lahiri, U., Swanson, A. R., Crittendon, J. A., Warren, Z. E., & Sarkar, N. (2013). A step towards developing adaptive robot-mediated intervention architecture (ARIA) for children with autism. Neural Systems and Rehabilitation Engineering, IEEE Transactions on, 21(2), 289-299. doi: 10.1109/TNSRE.2012.2230188.

Diehl, J. J., Crowell, C. R., Villano, M., Wier, K., Tang, K., & Riek, L. D. (2014). Clinical Applications of Robots in Autism Spectrum Disorder Diagnosis and Treatment Comprehensive Guide to Autism (pp. 411-422): Springer.

Lombroso, P. J., Ogren, M.P., Jones, W., Klin, A. (2009). “Heterogeneity and homogeneity across the autism spectrum: the role of development.” Journal of the American Academy of Child & Adolescent Psychiatry 48(5): 471-473.

Lord, C., Rutter, M., DiLavore, P.C., & Risi, S. (2002). Autism Diagnostic Observation Schedule (ADOS). Los Angeles, CA, Western Psychological Services.

MacWhinney, B. (2000). The CHILDES project: Tools for analyzing talk, vol 1: Transcription format and programs (3rd ed.). Mahwah, NJ, US: Lawrence Erlbaum Associates Publishers.

Child instructs Milo

Mean Percent Engaged

100.00

80.00

60.00

40.00

20.00

0.00Robot Therapist Disengaged

Minimally Verbal Fluent Speech

81.42

Milo instructs Child

PE

RC

EN

T 65.00

3.5415.04

3.00

32.01

PE

RC

EN

T

100.00

80.00

60.00

40.00

20.00

0.00Robot Therapist Disengaged

87.90

67.81

2.69 6.582.37

27.60

1

Using Humanoid Robots to Foster Social Skills in Children with ASD

Autism Spectrum Disorder (ASD) is a group of heterogeneous

neurodevelopmental disorders that severely compromise the development of social

relatedness, reciprocal social behavior, and communication. It is considered a spectrum

disorder due to the diversity of symptoms, skills, and disability. Nonetheless, persistent

deficits in social communication and interaction are the core features of ASD. An

estimated 1 in 68 American children (1 in 42 boys and 1 in 189 girls) are affected with

the disorder. The Centers for Disease Control and Prevention1 report that the prevalence

of the disorder has increased twentyfold in the past 40 years, with a 30% increase in the

past six years. While there is no cure for ASD, social skills intervention can improve the

social competencies of children and adolescents with ASD, thereby minimizing the core

features of the disorder.2

In response to the highly specialized needs of this population, researchers and

clinicians are grappling with innovative methods to teach social skills to children with

ASD. The past decade has seen a surge of technological advances and improved

affordability. More recently, several social robots have been developed, which has

created the possibility of robot-assisted therapy.3

Robots are a promising tool to facilitate social and communication skills in

children with ASD. They are simpler and more predictable compared to humans and

appear to be an intrinsic motivator.4,5 Many children with ASD show more social

engagement when interacting with a robot compared to when interacting with humans 6,7

or other devices.3 Further, children with ASD have been found to speak more to an adult

partner when the co-therapist was a robot compared to another human or a tablet.4,5

2

To our knowledge, the RoboKind Milo R25 robot is the only robot with a fully

integrated social skills curriculum (i.e., Robots4Autism). Robots4Autism’s innovative

robot-generated instruction teaches social and emotional understanding by integrating

several different evidenced based practices. Rollins & McFarlin8 found that school aged

children with ASD have a high level of engagement (i.e., they look at and/or gesture

toward, initiate comments, make contingent responses, or engage in an activity) with

Milo when he delivers lessons.

There are several reasons why children with ASD may be engaged with Milo.

First, Milo has a simple design with exaggerated facial features – these are the attributes

of robots that many children with ASD seem to prefer. 9 Second, there is consistent

evidence that children with ASD have impaired and/or slower processing of auditory

stimuli, particularly speech.10 Milo’s synthesized speech is 20% slower than human’s

speech, which may provide children with ASD additional time to process the auditory

speech stream. Finally, during lessons, a 2.4-inch LCD touch screen on Milo’s chest

displays, picture symbols of core vocabulary that coincide with Milo’s speech (see Figure

1). This simultaneous pairing of picture symbols with Milo’s speech enhances

comprehension of the material. 11,12

------------------------insert Figure 1 about here---------------------------------------------

While interacting with a robot may have its advantages, the ultimate goal is to

generalize social understanding to interactions with humans. Vernon, Koegel,

Dauterman, & Stolen 13 suggest that the combination of motivational and social

intervention components can create meaningful changes in social function. Further, the

3

reciprocal and interactive nature of social robots may serve as a social mediator and

bridge to interactions with humans.

This idea was supported in a recent single subject design study.14 Specifically,

the author found a four-year-old boy with ASD was able to generalize information from

the Robots4Autism greeting lessons to adults in his preschool class. In a single subject

experimental design, each child serves as their own control by receiving both no

treatment (baseline) and treatment conditions. This research design is particularly

appropriate for defining evidence-based educational practices.15 The participant, C, was

enrolled in a University-based preschool program where graduate students in

Communication Disorders help maintain a one-to-one clinician-to-child ratio. Greeting

was omitted from the classroom goals and the graduate clinicians and parents were

blinded from the objectives of the study. The Robots4Autism lesson targeted the verbal

and nonverbal components of saying “Hi.” Specific target behaviors were (1) look at the

other person, (2) smile, and (3) say “Hi.”

During the study, the child was brought to a quiet room twice a week for 15-20

minutes. During baseline (weeks 1-4), the child worked with Milo on identifying

emotions (i.e., Happy, Sad, Angry). During the intervention (weeks 5-10) the child

worked with Milo on the first greeting lesson (i.e., look, smile and say “hi”). Results of

intervention were examined in terms of how C generalized the Robot delivered lessons

into the classroom situation. Therefore, generalization data C’s ability to look, smile and

say “hi” was collected each morning when he entered his classroom. C was given the

opportunity to respond to four different graduate clinicians’ initiation of “hello”. Figure

2 displays C’s use of the verbal and nonverbal components of greeting. The percent of

4

opportunities C looked at a clinician, smiled and said “hi” (y-axis) was graphed by the

week number (x-axis) across baseline (blue line) and two generalization phases (red line

and green line respectively). In phase one, C received no additional support when

responding to a graduate clinician’s greeting. In phase two, if the C did not respond by

looking, smiling, and saying “hi” the graduate clinician silently held out picture symbols

used in the Robots4Autism greeting lesson (Figure 1) in order to create a bridge from the

robot-led curriculum to the classroom. Effectiveness was measured by percentage of

non-overlapping data points (PND). The intervention with robot was minimally effective

for generalizing to the classroom without added support (PND=.66) and highly effective

for generalizing to the classroom when visual supports were added (PND=100). These

results are striking when considering C was only receiving a half hour a week of

intervention on greeting starting at week 5. In addition, C’s father reported that C was

saying “hi” to more people at home.

While robots like Milo are not intended to replace human-led intervention,

existing research shows that humanoid robots are an intriguing tool to help children with

ASD. Organizations have made strides to make this intervention more accessible

because it has high potential to improve social and communication skills. Programs like

Robots4Autism are designed to provide individuals with skills that can be translated into

their daily lives — a priceless gift for students and their families.

5

Figure 1: Visual support using picture symbols from the Robots4Autism curriculum.

Figure 2: Percent C generalized the robot delivered lesson of (look, smile and say “hi”)

when greeting across baseline and two generalization phases

References:

1. Centers for Disease Control and Prevention. (2014). Prevalence of Autism

Spectrum Disorder among children aged 8 years—Autism and Developmental

Disabilities Monitoring Network, 11 Sites, United States, 2010. Morbidity and

Mortality Weekly Report (MMWR), 63(2), 1–21.

2. Reichow, B., Steiner, A. M., & Volkmar, F. (2013). Cochrane review: Social

skills groups for people aged 6 to 21 with autism spectrum disorders (ASD).

Evidence‐Based Child Health: A Cochrane Review Journal, 8(2), 266–315.

3. Diehl, J. J., Crowell, C. R., Villano, M., Wier, K., Tang, K., & Riek, L. D. (2014).

Clinical applications of robots in autism spectrum disorder diagnosis and

treatment. In V. B. Patel, V. R. Preedy, & C. R. Martin (Eds.), Comprehensive

guide to autism (pp. 411–422). New York, NY: Springer.

6

4. Kim, E., Berkovits, L., Bernier, E., Leyzberg, D., Shic, F., Paul, R., & Scassellati,

B. (2012). Social robots as embedded reinforcers of social behavior in children

with autism. Journal of Autism and Developmental Disorders, 43(5), 1038–1049.

5. Srinivasan, S., & Bhat, A. (2013). The effect of robot-child interactions on social

attention and verbalization patterns of typically developing children and children

with autism between 4 and 8 years. Autism-Open Access. Paper 18. Retrieved

from http://digitalcommons.uconn.edu/libr_oa/18

6. Bekele, E., Crittendon, J. A., Swanson, A., Sarkar, N., & Warren, Z. E. (2014).

Pilot clinical application of an adaptive robotic system for young children with

autism. Autism, 18(5), 598–608.

7. Pop, C., Pintea, S., Vanderborght, B., & David, D. (2014). Enhancing play skills,

engagement and social skills in a play task in ASD children by using robot-based

interventions: A pilot study. Interaction Studies, 15(2), 292–320.

8. Rollins, P. R., McFarlin, M. N,.(2015). Innovations Using Robotics Technology

to Improve Social Skills in Children with ASD. Texas Speech and Hearing

Association. San Antonio, TX (April 2015).

9. Peca, A., Simut, R., Pintea, S., Costescu, C., & Vanderborght, B. (2014). How do

typically developing children and children with autism perceive different social

robots? Computers in Human Behavior, 41, 268–277

10. O’Connor, K. (2012). Auditory processing in autism spectrum disorder: a review.

Neuroscience & Biobehavioral Reviews, 36(2), 836-854.

11. Rollins, P. R. (2014). Facilitating early social communication skills: From theory

to practice. Shawnee Mission, KS: AAPC.

12. Rollins, P. R., & McFarlin, M. N. (2015). AAC Strategies to improve social skills

in children with Autism. Paper presented at ATIA 2015: Orlando, FL, January

2015

13. Vernon, T. W., Koegel, R. L., Dauterman, H., & Stolen, K. (2012). An early

social engagement intervention for young children with autism and their parents.

Journal of Autism and Developmental Disorders, 42(12), 2702–2717.

14. Rollins, P.R., & Blohm, C., (2016). Securing National Visibility for Autism

Research Results. Paper presented at ATIA 2015. Orlando Fla (January 2016).

15. Kratochwill, T. R., Hitchcock, J., Horner, R. H., Levin, J. R., Odom, S. L.,

Rindskopf, D. M., & Shadish, W. R. (2010). Single-case designs technical

documentation. What Works Clearinghouse. Retrieved from

http://ies.ed.gov/ncee/wwc/pdf/reference_resources/wwc_scd.pdf.