Autism & Robots & Robbie the Robot

Posted by Erin, the RobotGrrl on Saturday, April 5th, 2008

Robbie the Robot, a human-machine conversation computer program I developed, has the capability aid autistic children. I’ve been thinking about this intensely for the past few weeks. Today (March 28), much to my excitement, my favourite podcast - Talking Robots - features guest Kerstin Dautenhahn!

This is what they say about her…

Kerstin Dautenhahn who is Research Professor in the School of Computer Science, Faculty of Engineering and Information Sciences at University of Hertfordshire in the UK, where she is a coordinator of the Adaptive Systems Research Group. Autism is a developmental disorder affecting around 91 people in every 10,000, mainly causing difficulties in social interactions, communication and imagination. Using therapy robots in the AuRoRA project, Dautenhahn has been pushing autistic children to learn essential social skills such as turn taking, joint attention and imitation. Armed with a lot of patience and zeal, her team has been adapting their robots and therapy sessions to each individual child, whether it’s about playing with wheeled robots or Dautenhahn’s toddler-sized Kaspar humanoid. She also presents her new project on Interactive RObotic social MEdiators as Companions (IROMEC) which looks at how autistic children can learn to cooperate and interact with each other through the introduction of a robotic mediator.

So, I listened to the podcast a few times and took some notes…

• A person with autism has several problems interacting with others
  • It’s a life-long development disorder

  • Problems social interaction and communication
  • Problems understanding facial expressions and body language
• It’s a spectrum dissorder
  • You can have withdrawn people to very able people (different sides of the spectrum)
• Children with autism enjoy interacting with robots as much as any other person. They’re realy fascinated by it. // This is a great fact, especially for the future development of Robbie the Robot
• Since they were interacting with an autonomous robot, it was completely open! - No concrete hypothesis.
  • Found they are very very very interested

  • Enjoy interacting with it
  • No fear interacting with the robot or touching
  • They were smiling, and very keen.
• Autistic children have problems with behaviours like joint attention and immitation. Designed trials, to test this. Particular aspects of behavioural aspects.
• What they did in their work with a humanoid that was able to move its arms and legs:
• Children could interact with robot by similarly moving arms and legs
• Child became more interested, immitated the robot and innitiated the interaction
• If the robot was unable to do that behaviour, child would try something different
• Imagine not being able to know turns for conversations… turn taking in special need schools is what they are teaching // A key target for Robbie the Robot!

  Robots are not a replacement for people. // AGREED!
  Robots simplify!

• Human-human interaction is very complex
  • Posture, body language, facial expression…

  • Tone, politeness, rules, queues
  • … we are to interpret them in a social way
  • For example: Do you know what time it is?
    Yes
    No
    Yes, it’s 4:30.
  • Social correctness is the last one
• You need to interpret what you hear and interpret their behaviour
  • Very difficult for a person with autism

  • Humanoids are not as complex
    • That way, you can use one to slowly teach the child

    • First you need to connect, and get the child interested
• Lots of results have to be published on how children can imitate or use joint attention
• More clinical study is needed
• Types of robots
  • Different types are suitable for different children
    • Low functional end - successful mobile robot engages attention. (Applied AI Systems).

    • Simple, robot moves around and has heat sensors and IR sensors
    • Chasing, tag, run around, etc. Very very very simple robot
    • More abilities, imitation … humanoid shape.
    • Designs and types of behaviours you want to explore.
    • Casper has facial expressions, it can turn its head, move arms, produce gestures, play peek-a-boo.
    • It really depends on what groups of children you are dealing with
    • Communication and interaction skills produce more challenges.
    • All in all, start with what you know about the child, and design objectives of trials for therapy for the children
    • Robots are individualized! This way, you can hope interaction capabilities develop
• Challenges from the Robotics point of view
  • Nature - important that robot is predictable

  • Scenerios are unconstrained
  • Sensor system has 100% reliability
  • More autonomy
• Summary of results
  • Children respond to a robot differently to a non-robotic toy

  • They are comfortable, they learn imitative behaviour, use robot to show joint attention
  • Emphasis on mediation aspect
  • Social mediator. 1 Robot, 2 children. Communication and interaction between the children
  • Collaboration is very difficult for a child with autism

  They use what they learn with the robot outside of the class.

• Although solid evidence is needed, there is big potential.

  Robot assisted therapy is a growing area! Commercialize these robots. Here comes Robbie the Robot!

  After breezing through all of my notes taken during this podcast, it becomes extremely obvious that Robbie can fit snuggly in a robot assisted therapy with an autistic child that is able to communicate, but sruggles slightly to keep a conversation going.

  Now, the big question is how am I going to make Robbie to the point where it is extremely intelligent and can assist autistic children?

• Use and exploit language models - be sure to include anything that can happen, as it might happen
• Work on continuous speech recognition - this way sentences can be heard
• Convert what is heard into a probable sentence - if what it said was wrong, the child might correct it, where then I can monitor and adjust it accordingly
• Add some algorithms to advance the conversation according to the number of times Robbie has conversed with that particular child, and for how long — this way, Robbie will be able to ‘coach’ them to have a normal conversation — Once they are able to do so, they won’t need Robbie anymore!
• Create a teddy bear, or bubble boy physical device to make it seem less frightening — however, that would have to be tested, as maybe they would be more frightening that just a laptop

  It definitely requires some thought, but I’m up for the challenge!
  If this post made no sense, I highly recommend listening to the podcast itself, which can be found here, and the mp3 here.

Posted in: News, Projects, Robot, Robot News.
Tags: autism, computer science, Kerstin Dautenhahn, Robbie, Robbie the Robot, robots

Leave a Reply