Goren Gordons’ research focus is on the intersection of artificial intelligence, developmental psychology, education and engineering. He studies curiosity using a formal mathematical framework which attempts to explain curiosity-driven behavior in children. The same framework is: (i) implemented in social curious robots that learn about themselves and people around them; (ii) used to assess children’s curiosity, and (iii) used to construct improved curriculum. The social curious robots are then introduced as social companion learners for children, in the attempt to promote learning and curiosity during long-term interaction and play.
What have I achieved during my fellowship?
During my Jacobs Foundation Research Fellowship, I have worked on developing curiosity assessment tools. I have started with developing the mathematical framework for model-based curiosity assessment, based on my curiosity-loop concept. The first studies were aimed at adults, to better validate the framework using self-reporting questionnaires, that are more difficult with children. I have shown that the framework indeed works, with two studies regarding physical and social curiosity. Moreover, I have developed a tablet app that assesses other aspects of curiosity and have conducted a large-scale study with university applicants. I have shown, for example, that priming for curiosity results in more efficient learning and that enabling curious exploration results in more learning.
With these frameworks I have started to build a toolbox for curiosity assessment for children using age-appropriate tablet games. We have conducted a study in the Youth University to validate the games, using the students’ teachers as external validation. One such game, called Questions Worlds, assesses question-asking behavior of children. Our research question was whether curious children ask different questions than less curious children. We have indeed found that questions can be parameterized by unique network-theory measures and that curious children ask questions which have high specificity, namely, that their answers cannot be reached by other questions.
We have designed a large-scale study in elementary schools using four such tablet games. The goal is to have children in elementary schools be assessed by the games and have their teachers and parents fill validated curiosity questionnaires on themselves and their children. This way we aim to create a curiosity map of schools.
However, the study was about the begin in the school when the COVID19 pandemic started. All interaction with schools has then ceased. We have pivoted to make all the games and questionnaires online and have worked on this since then.
Concurrently, we have been working on curious social robots for curious children. We have developed an interaction with children based on tangram games in which the child and robot take turns in solving these spatial puzzles. We have designed an expressive curiosity cognitive architecture for the robot which not only learns to solve puzzles with the child, but also expresses its internal curiosity-based parameters as verbal and non-verbal behaviors of enthusiasm of learning. We have also used the same algorithm to assess children’s curiosity and have found that children were significantly affected by the curiosity of the robot.
We have also developed other types of interaction between social robots and children, not focused on curiosity. The rise of social robots as a promising medium for teaching children has enabled us to explore new interaction. We have conducted a study on teaching the Hebrew language to children ages 6-8 using a social robot. We have shown that while younger children start off with less morphological awareness, after a few short interactions with our social robots, they have learned as much as their older peers and achieved amazing post-test results.
Furthermore, we have developed a unique triadic interaction between robot-toddler-caregiver for English as a second language tutoring. We have compared the interaction with the social robot to that of a tablet and have shown that interaction with a social robot promotes significantly more social interaction, as measured by joint gaze and facial expressions, compared to the tablet. This social interaction is of paramount importance for the scaffolding of learning at this early age.
Finally, we have started developing a novel architecture wherein social robots are facilitating small group activities in schools. We have worked in the last one year and a half with the Ministry of Education on this system. We have first conducted a study with the teachers on the system, in which the teachers insert content of group activities, and we supply the pedagogy of the group facilitation. Again, we were ready to begin pilots in schools when the Pandemic started. So, we have also pivoted this project from social physical robots to avatars in the online domain.
My plans for the future
I plan to finalize the studies that have been delayed by the Pandemic. First, I will finalize the study of the curiosity map in schools using online format of the games and the questionnaires. Hopefully, this study will give a definitive answer to the question “Does school kill curiosity?”
Next, I will continue with the flagship project with the Ministry of Education to introduce social robots, as well as virtual avatars, as small group facilitators. These interactions are of paramount importance, as small group activities have been shown to be an extremely effective mode of learning. Through this interaction, important 21st century skills can be introduced, such as collaboration, communication, curiosity, growth mindset and others.
One of my main goals is to take these platforms “out of the lab” and implement them in low SES schools. I believe that these technological advancements, which I open-source and thus make affordable, can drastically reduce the gap that has been ever increasing since the Pandemic has arrived. Enabling state-of-the-art pedagogy with facilitation of socio-emotional emphasis can truly make a difference.