Why Does This Robot Get ‘Sick’ on Purpose?

As a philosophy student, Dr.

Lola Cañamero was fascinated by how humans learn and understand things about the world.

When she studied artificial intelligence as part of her curriculum, she saw parallels between the two disciplines, and looked to computers and robots to tell that story.

Together with Dr.

Matthew Lewis, Dr.

Cañamero developed ROBIN, an autonomous robot “toddler” intended to help children with diabetes manage their condition.

She is currently at the University of Hertfordshire in the UK, where she leads the Embodied Emotion, Cognition and (Inter-)Action (EECAIA) Lab at the School of Engineering and Computer Science.

We spoke to her recently about the ROBIN project.

Dr.

Cañamero, what inspired you to test a robot 'toddler' that helps children with diabetes manage their condition?
[LC] ROBIN, which stands for ROBot INfant, was initially developed as part of an EU-funded international project called ALIZ-E to develop robot companions for children with chronic conditions such as diabetes, in order to support their understanding and management of their conditions. 

What does the robot do?
ROBIN is designed to behave like a toddler—a simple version of a toddler—walking around and doing things in a toddler’s playroom that contains different types of objects, such as toys, etc.

Children are invited to “look after ROBIN” in its playroom.

ROBIN has a simulated physiology that gives it internal needs of various types—physical, such as eating, drinking, and sleeping; curiosity-related such as exploring and playing; and social, such as seeing people and faces and interacting with them (like requesting hugs).

ROBIN has a number of behaviors that allow it to satisfy those needs, either on its own, or with the help of the children. 

How does the robot display the disease state of diabetes?
ROBIN has a simulated model of diabetes, which means that sometimes it displays symptoms of diabetes—such as being tired, thirsty, or dizzy—when its internal simulated glucose levels are too high, in which case children need to administer it “robot insulin” externally, using a “robot glucometer." If it's too low, children need to give ROBIN corrective foods.

Describe the gameplay aspect so children feel motivated to assist the robot.
Children have to try to identify the symptoms of diabetes when playing with ROBIN, and some of those symptoms are not always clear.

Like themselves, ROBIN might be sleepy because it is tired from walking around, or due to high or low glucose levels.

Therefore, they need to try to remember and decide what to do in such situations—whether they should measure the glucose of ROBIN and how to correct it, whether it would be “safe” to give ROBIN some food when it asks for it, etc. 

How are you measuring success?
We have not carried out long-term studies yet, so we cannot assess the potential efficacy of ROBIN as a long-term learning tool, but that is something that we would like to do in the future.

So far we have focused on exploring how to design ROBIN as a meaningful tool that children with diabetes can relate to and engage with, to support them not only in learning about their condition but also on the more affect-related aspects of diabetes management, such as building up confidence in their skills, and making their learning about diabetes less stressful.  

What made you decide on the NAO robot hardware platform from Aldebaran and Softbank Robotics?
NAO was chosen for various reasons, such as the fact that children seem to engage very naturally with it, as well the fact that it provides good functionalities for interaction with children. 

What’s the software platform that you use for this robot?
We wrote our own control software using mostly UrbiScript with some C++.

The software controller/decision-making architecture, as we call it, that drives the behavior of ROBIN builds on a “core” model that we have used in my group for different robots, research topics and applications related to decision making.

In this case, we adapted that model to the specific needs and behaviors that are appropriate for ROBIN as a playful and social toddler that also has (robot) diabetes.

ROBIN’s decision-making architecture is based on the notions of motivation and emotions in humans, and implements specific theories of how they interact and work in guiding behavior and making decisions about what to do, including how to interact socially. 

In your co-authored paper with Dr.

Lewis, Making New “New AI” Friends: Designing a Social Robot for Diabetic Children from an Embodied AI Perspective, you talk about the importance of making the ‘interaction unstructured and partly ambiguous and unpredictable’ - can you explain why?
We thought it would be important that the interaction was unstructured—as opposed to following a rigid script that the robot would be programmed to enforce—and partly ambiguous and unpredictable, as this would make this “play” experience feel closer to the complexity of real diabetes self-management.

The use of a motivationally and cognitively autonomous robot and an embodied AI approach is instrumental to this end, as these elements make each interaction unique, due to both the dynamics of the architecture in interaction with the physical and social environment (the robot never behaves in exactly the same way twice), and to the different ways in which each child treats the robot. 

Talk us through the trials to date.
As part of the EU-funded ALIZ-E project, we carried out pilot interactions with 17 diabetic children within our initial target age range of 7 to 12 at a hospital and a diabetes summer camp, both in Italy, to assess whether the elements of the interaction were appropriate, believable, and engaging with a variety of children in a real-world context.

These studies were not to assess the effectiveness of ROBIN as a tool to support self-efficacy in diabetes management.

Later on, we carried out also small pilot studies at a couple of hospitals in the UK, mostly to assess potential differences and changes we might need to make to the robot or the interaction scenario due to, for example, differences in the treatment of diabetes, and cultural differences.

What surprised you, if anything?
A number of things surprised us.

First, the enormous variety in the ways children interacted with ROBIN.

Another surprise was to see how willing children were to “play the game"—play with ROBIN and look after it, trying to make sure that ROBIN was “fine," despite the fact that they knew it was a machine.

We made this very clear to them, and explained how the sensors of the robot worked, how it was switched off and on.

We were also surprised by how easily many children related the game to their own experience of diabetes, and how it helped them reflect and in some cases talk about it.

I’ve written about several robots that are being used within pediatric scenarios, including RXRobots.

Would you say there’s something particularly suitable in using robots to help children?
Robots have a great potential in the education of children.

Nowadays, children are used to robots; in many cultures they grow up seeing them and playing with them.

This usually means that children are not afraid and interact with them quite naturally, whereas adults can be more wary.

Sometimes children see robots as a toy, sometimes as a playmate, sometimes as an interesting “gadget” that nurtures their natural “scientific” curiosity. 

Are you looking to take it into the commercial realm?
The next step would be to test it with a larger number of children and assess its effectiveness as an educational tool to support diabetic children in learning about their condition and how to manage it.

We would certainly love to see it used in real life if it proves to be useful. 

As a philosophy student, Dr.

Lola Cañamero was fascinated by how humans learn and understand things about the world.

When she studied artificial intelligence as part of her curriculum, she saw parallels between the two disciplines, and looked to computers and robots to tell that story.

Together with Dr.

Matthew Lewis, Dr.

Cañamero developed ROBIN, an autonomous robot “toddler” intended to help children with diabetes manage their condition.

She is currently at the University of Hertfordshire in the UK, where she leads the Embodied Emotion, Cognition and (Inter-)Action (EECAIA) Lab at the School of Engineering and Computer Science.

We spoke to her recently about the ROBIN project.

Dr.

Cañamero, what inspired you to test a robot 'toddler' that helps children with diabetes manage their condition?
[LC] ROBIN, which stands for ROBot INfant, was initially developed as part of an EU-funded international project called ALIZ-E to develop robot companions for children with chronic conditions such as diabetes, in order to support their understanding and management of their conditions. 

What does the robot do?
ROBIN is designed to behave like a toddler—a simple version of a toddler—walking around and doing things in a toddler’s playroom that contains different types of objects, such as toys, etc.

Children are invited to “look after ROBIN” in its playroom.

ROBIN has a simulated physiology that gives it internal needs of various types—physical, such as eating, drinking, and sleeping; curiosity-related such as exploring and playing; and social, such as seeing people and faces and interacting with them (like requesting hugs).

ROBIN has a number of behaviors that allow it to satisfy those needs, either on its own, or with the help of the children. 

How does the robot display the disease state of diabetes?
ROBIN has a simulated model of diabetes, which means that sometimes it displays symptoms of diabetes—such as being tired, thirsty, or dizzy—when its internal simulated glucose levels are too high, in which case children need to administer it “robot insulin” externally, using a “robot glucometer." If it's too low, children need to give ROBIN corrective foods.

Describe the gameplay aspect so children feel motivated to assist the robot.
Children have to try to identify the symptoms of diabetes when playing with ROBIN, and some of those symptoms are not always clear.

Like themselves, ROBIN might be sleepy because it is tired from walking around, or due to high or low glucose levels.

Therefore, they need to try to remember and decide what to do in such situations—whether they should measure the glucose of ROBIN and how to correct it, whether it would be “safe” to give ROBIN some food when it asks for it, etc. 

How are you measuring success?
We have not carried out long-term studies yet, so we cannot assess the potential efficacy of ROBIN as a long-term learning tool, but that is something that we would like to do in the future.

So far we have focused on exploring how to design ROBIN as a meaningful tool that children with diabetes can relate to and engage with, to support them not only in learning about their condition but also on the more affect-related aspects of diabetes management, such as building up confidence in their skills, and making their learning about diabetes less stressful.  

What made you decide on the NAO robot hardware platform from Aldebaran and Softbank Robotics?
NAO was chosen for various reasons, such as the fact that children seem to engage very naturally with it, as well the fact that it provides good functionalities for interaction with children. 

What’s the software platform that you use for this robot?
We wrote our own control software using mostly UrbiScript with some C++.

The software controller/decision-making architecture, as we call it, that drives the behavior of ROBIN builds on a “core” model that we have used in my group for different robots, research topics and applications related to decision making.

In this case, we adapted that model to the specific needs and behaviors that are appropriate for ROBIN as a playful and social toddler that also has (robot) diabetes.

ROBIN’s decision-making architecture is based on the notions of motivation and emotions in humans, and implements specific theories of how they interact and work in guiding behavior and making decisions about what to do, including how to interact socially. 

In your co-authored paper with Dr.

Lewis, Making New “New AI” Friends: Designing a Social Robot for Diabetic Children from an Embodied AI Perspective, you talk about the importance of making the ‘interaction unstructured and partly ambiguous and unpredictable’ - can you explain why?
We thought it would be important that the interaction was unstructured—as opposed to following a rigid script that the robot would be programmed to enforce—and partly ambiguous and unpredictable, as this would make this “play” experience feel closer to the complexity of real diabetes self-management.

The use of a motivationally and cognitively autonomous robot and an embodied AI approach is instrumental to this end, as these elements make each interaction unique, due to both the dynamics of the architecture in interaction with the physical and social environment (the robot never behaves in exactly the same way twice), and to the different ways in which each child treats the robot. 

Talk us through the trials to date.
As part of the EU-funded ALIZ-E project, we carried out pilot interactions with 17 diabetic children within our initial target age range of 7 to 12 at a hospital and a diabetes summer camp, both in Italy, to assess whether the elements of the interaction were appropriate, believable, and engaging with a variety of children in a real-world context.

These studies were not to assess the effectiveness of ROBIN as a tool to support self-efficacy in diabetes management.

Later on, we carried out also small pilot studies at a couple of hospitals in the UK, mostly to assess potential differences and changes we might need to make to the robot or the interaction scenario due to, for example, differences in the treatment of diabetes, and cultural differences.

What surprised you, if anything?
A number of things surprised us.

First, the enormous variety in the ways children interacted with ROBIN.

Another surprise was to see how willing children were to “play the game"—play with ROBIN and look after it, trying to make sure that ROBIN was “fine," despite the fact that they knew it was a machine.

We made this very clear to them, and explained how the sensors of the robot worked, how it was switched off and on.

We were also surprised by how easily many children related the game to their own experience of diabetes, and how it helped them reflect and in some cases talk about it.

I’ve written about several robots that are being used within pediatric scenarios, including RXRobots.

Would you say there’s something particularly suitable in using robots to help children?
Robots have a great potential in the education of children.

Nowadays, children are used to robots; in many cultures they grow up seeing them and playing with them.

This usually means that children are not afraid and interact with them quite naturally, whereas adults can be more wary.

Sometimes children see robots as a toy, sometimes as a playmate, sometimes as an interesting “gadget” that nurtures their natural “scientific” curiosity. 

Are you looking to take it into the commercial realm?
The next step would be to test it with a larger number of children and assess its effectiveness as an educational tool to support diabetic children in learning about their condition and how to manage it.

We would certainly love to see it used in real life if it proves to be useful.