The Efficacy of Robot-enhanced Therapy for Children With Autism Spectrum Disorders (DREAM-RCT)

DREAM Project Trial: Testing the Efficacy of Robot-enhanced Therapy for Children With Autism Spectrum Disorders in a Randomised Clinical Trial

University of Skövde, Vrije Universiteit Brussel, University of Plymouth, University of Portsmouth, De Montfort University, SoftBank Robotics

This trial will test the efficacy of robot-enhanced interventions for developing social skills in children with Autism Spectrum Disorders (ASD). The study will compare, in an equivalence design, an innovative intervention delivered by a semi-autonomous robotic agent with standard behavioral intervention. The target group is children between 3 and 6 years old which will be randomly allocated to one of the two treatments. Each treatment will be delivered over 8 bi-weekly sessions of 45 minutes each.

Autism Spectrum Disorders (ASD) is characterized by a pattern of persistent deficits in communication and abnormal behaviors (i.e., repetitive and restrictive). Previous research has shown promising results for the use of social robots to develop social skills in ASD children. However, most of the research is based on case studies and single case experiments. Rigorous clinical data comparing robot--enhanced therapy with standard therapy is needed to establish the efficacy of this innovative intervention. The present study is an equivalence randomised clinical trial which plans to include children with a diagnosis of ASD between 3 to 6 years old. Eligible children are randomly allocated in one of two groups: (1) standard therapy, derived from applied behavior analysis (ABA) and (2) a robot--enhanced intervention in which the same treatment is delivered by a semi--autonomous robotic mediator agent (the robot acts as a mediator while the therapist is supervising the session). Both interventions are delivered over 8 bi-weekly sessions of about 45 minutes each. The present study will provide some of the first rigorous data regarding the efficacy of robot--enhanced therapy delivered by an intelligent robotic agent which require less direct input from an operator to deliver the intervention, as compared to previous research. If the results demonstrate the equivalence of the two interventions, this could open the door for the dissemination of evidence--based interventions for ASD children through the use of robotic agents.

Random allocation to one of two intervention conditions: (1) standard therapy, derived from applied behavior analysis (ABA) and (2) a robot-enhanced intervention in which the same intervention is delivered by a semi-autonomous robotic mediator agent (the robot acts as a mediator while the therapist is supervising the session).

Principal investigator, the research team, and data analyst are all blind to the treatment conditions.

Standard therapy based on Applied Behavioral Analysis (ABA) and Cognitive-Behavior-Therapy (CBT). It implies structured activities in which the therapist reinforces the adaptive behavior of the child.

A treatment developed on the same principles as standard therapy, based on Applied Behavioral Analysis (ABA) and Cognitive-Behavior-Therapy (CBT). It implies structured activities in which the robotic agent reinforces the adaptive behaviors of the child, under the supervision of the therapist.

The tasks included in the treatment follow a structured behavioral approach, called the discrete trial training (DTT). The learning process is guided by the robotic agent (Softbank Robotics Nao robot). All activities follow a gameplay scenario and they take place at a table, which in the case of joint attention and turn taking tasks incorporates a large touchscreen. Each behavior presented to the child is preceded by instructions and followed by a contingent reinforcement. The contingent reinforcement and prompting are used to train each targeted behavior/action. The behaviors are presented over multiple and successive trials and explicit prompting is giving when the child doesn't succeed to accomplish the targeted behavior. Each action is repeated three times. For every child, the intervention for each ability starts from the level determined in the first two sessions, but as the child's performance improves, the task moves to a more complex level.

The tasks included in the treatment follow a structured behavioral approach, called the discrete trial training (DTT). The learning process is guided by an interaction partner (human), under the supervision of the therapist. All activities follow a gameplay scenario and they take place at a table, which in the case of joint attention and turn taking tasks incorporates a large touchscreen. Each behavior presented to the child is preceded by instructions and followed by a contingent reinforcement. The contingent reinforcement and prompting are used to train each targeted behavior/action. The behaviors are presented over multiple and successive trials and explicit prompting is giving when the child doesn't succeed to accomplish the targeted behavior. Each action is repeated three times. For every child, the intervention for each ability starts from the level determined in the first two sessions, but as the child's performance improves, the task moves to a more complex level.

Imitation skills will be assessed trough a clinician's rating of the child performance in imitating the behavior of the play partner (robotic or human agent), based on an observation grid.

The measurement of Imitation skills will be conducted before the intervention and after the intervention (after 8 sessions / after 4 weeks). The outcome will be represented by change scores on this measure.

This measure is a clinician rating of the child performance in demonstrating and initiating joint attention with the play partner (robotic or human agent), based on an observation grid.

The measurement of Joint-attention skills will be conducted before the intervention and after the intervention (after 8 sessions / after 4 weeks). The outcome will be represented by change scores on this measure.

This measure is a clinician rating of the child performance in demonstrating turn taking skills in the gameplay with the partner (robotic or human agent), based on an observation grid.

The measurement of Turn-taking skills will be conducted before the intervention and after the intervention (after 8 sessions / after 4 weeks). The outcome will be represented by change scores on this measure.

This measure is a clinician rating of the child engagement in the game play with the partner (robotic or human agent), based on an observation grid.

This measure will be assessed 10 times: once, one week before the intervention; 8 times in 4 weeks (twice every week) during the intervention; once, one week, after the intervention.

This measure is a clinician rating of the child child expression of verbal utterances in relation to the gameplay with the partner (robotic or human agent). The frequency of verbal utterances is recorded for this outcome.

This measure will be assessed 10 times: once, one week before the intervention; 8 times in 4 weeks (twice every week) during the intervention; once, one week, after the intervention.

Performance in a task where the child is asked to share some personal information regarding own preferences by choosing a picture on the touchscreen.

This measure will be assessed 10 times: once, one week before the intervention; 8 times in 4 weeks (twice every week) during the intervention; once, one week, after the intervention.

Performance in a task where the child is asked to categorize some objects by moving a picture on the touchscreen over another picture depicting and object that belongs to the same category.

This measure will be assessed 10 times: once, one week before the intervention; 8 times in 4 weeks (twice every week) during the intervention; once, one week, after the intervention.

Performance in a task where the child is asked to chose an object from several others that match the pattern of a pre-existing series of objects (e.g., based on geometrical shape or color).

This measure will be assessed 10 times: once, one week before the intervention; 8 times in 4 weeks (twice every week) during the intervention; once, one week, after the intervention.

ADOS (Lord, Rutter, DiLavore, & Risi, 2008) is a clinician rated semi-structured set of activities designed for observing behaviors that are relevant to judge presence of the diagnostic criteria for autism spectrum disorder.

The test will be applied twice: one week before the intervention and one week after the intervention.

SCQ (Rutter, Bailey & Lord) is a other-report questionnaire used for autism spectrum disorders screening and assessment. The questionnaire is filled by the parents or another tutor in relation to the social skills and social functioning of the child.

The test will be applied twice: one week before the intervention and one week after the intervention.

Inclusion Criteria: to be diagnosed using DSM-V criteria by a psychiatrist; the diagnostic is confirmed by the Autism Diagnostic Observation Scale (ADOS; scores are in the clinical range).

Lord, C, Rutter, M, DiLavore, PC, Risi, Sl. Autism diagnostic observation schedule: ADOS manual. Los Angeles: Western Psychological Services, 2008.

Rutter, R, Bailey, A, Lord, C. Social Communication Questionnaire. Los Angeles: Western Psychological Services, 2003.