How social robots can make children more curious

Goren Gordon with robots George and Howie
Goren Gordon with robots George and Howie

In the Information Age, when every answer is just a click away, curiosity – the intrinsic drive to learn and explore – is becoming ever more important. But what is curiosity? What happens to children’s curiosity as they grow older and enter the formal educational system? Are there ways to encourage curiosity?

These are some of the questions that we study in the Curiosity Lab at Tel Aviv University.

Most previous research on curiosity in children has been qualitative and subjective, and has focused on observing children’s behavior and coding for such expressions of curiosity as the number of questions a child asks. The new field of artificial curiosity, that is the curiosity of artificial agents such as computers and robots, is inspired by new insights in neuroscience. It proposes a mathematical model of curiosity based on the idea that learning is an intrinsic reward that changes exploration behavior. This new model makes it possible to take a more quantitative and digital approach to assessing and promoting curiosity in children.

In my lab, we are currently developing tablet games based on this new mathematical understanding of curiosity. These games assess various expressions of curiosity – from free to structured exploration and from diverse to specific curiosity, and taking into account physical, cognitive, and social curiosity.

Each of the games involves learning something new ­– such as how an imaginary planet is organized or what a monster likes to eat. All of them include several possible actions, each supplying a different type of data, so that players can choose the kind of information they want to obtain. Since some of the actions in these games reveal more information than others, there is an optimal sequence of actions that will generate the most information and allow the player to learn the most.

We use computer algorithms to determine which sequence of actions yields the most information. Comparing a child’s solution in these learning games with the algorithmic solutions gives us a quantitative measure of a child’s behavior by showing how close the child comes to the optimal solution. Repeating the game several times also allows us to measure improvements in the child’s exploration strategy – in other words, the extent to which the child has become better able to learn. This unique game design gives us a better understanding of children’s expressions of curiosity.

“Repeating the game several times allows us to measure improvements in the child’s exploration strategy – in other words, the extent to which the child has become better able to learn.”

We plan to use these novel assessment tools with children between five and twelve years of age. One of the main goals of the study is to gain a more profound understanding of the dynamics of curiosity as children grow older. To that end, we plan to assess all the children in several elementary schools in a large municipality in Israel, from first through sixth grade.

This will allow us to analyze, in a novel way, how various types of curiosity change as a function of grade and age. In other words, we will be able to determine whether a child’s question-asking behavior is changing, and how cognitive curiosity – how the child analyzes new learning tasks and selects which actions to perform in order to obtain the most information – changes over the course of elementary school.

Previous studies have shown that curiosity wanes as children grow older or, perhaps more precisely, as they spend more time in the formal educational system. This new study is designed to be much broader in scope, and will assess a wide variety of dimensions of curiosity on a large sample of children spanning a wider age range.

Curiosity is contagious

Assessment, however, is just the first step. The project’s second goal is to encourage children who show a lack of curiosity to be more curious. In previous studies, conducted in collaboration with Susan Engel of Williams College while I was working as a postdoc in the Media Lab at the Massachusetts Institute of Technology (MIT), I found that curiosity is contagious; in other words, children can “catch” curiosity from their peers. To study this effect, I developed a unique experimental design that uses a social robot as a peer companion and then evaluates the effect of the robot’s behavior on children’s curiosity.

Some robots were programmed to behave like a curious child – expressing a love of learning, demonstrating resilience in response to mistakes and showing enthusiasm for novel things. The children who played with the curious robot were found to exhibit greater curiosity after the interaction, relative to children who played with a non-curious robot.

“The children who played with the curious robot were found to exhibit greater curiosity after the interaction, relative to children who played with a non-curious robot.”

In the years to come, I intend to continue this work on a larger scale, and will also include older children. Using the comprehensive assessment toolbox, I plan to conduct a thorough pre- and post-intervention study, assessing children’s curiosity before and after long-term interactions with curious social robot peers.

The primary goal of my research is to maintain and increase children’s curiosity throughout their elementary school years in the formal educational system. To that end, I plan to develop unique and comprehensive assessment tools for use with all children in several elementary schools in a large municipality in Israel.

The games are being developed and the assessment study will commence in the following months. It is my hope to show that social robots, employed as peer companions, will help children retain their curiosity as they grow older, so that they will remain eager to explore and learn about this amazing world.