Sinem Aslan, Nese Alyuz, and Sangita Sharma are research scientists in the Intelligent Systems Research Lab, led by Lama Nachman at Intel Labs.
Highlights
- The final Kid Space research study shows the value of the immersive learning environment in supporting collaborative problem-solving (CPS) processes in young learners.
- Scientists involved in the multi-year research project completed several prototype studies, demonstrating the potential of the program to facilitate student engagement in various learning environments, including classrooms, after-school programs, and home-based learning.
- The combination of physical interaction, real-time adaptation, and collaborative learning positions Kid Space as a distinct contribution to the future of educational technology.
Researchers from Intel Labs and Purdue University recently collaborated on the final Kid Space study using the school prototype, demonstrating how responsible artificial intelligence (AI) has enormous potential in helping teachers facilitate student engagement and nurture collaborative problem-solving (CPS). Scientists and educators know that behavioral, emotional, and social engagement all play a vital role in learning, especially for young children. Understanding learner engagement also serves as an important metric, helping teachers deliver more personalized learning experiences. By combining these metrics in a collaborative learning environment facilitated by Oscar, Kid Space’s AI-based learning companion, students can augment both academic and social skills with real-time, personalized feedback.
Over the course of several years, scientists have completed a total of four studies, including an ethnographic study with teachers and three experimental studies with students and parents conducted in elementary schools and home-based learning environments. While the immersive Kid Space research started as a school prototype, with the onset of COVID-19, the needs of educational stakeholders changed towards online learning. In response, the researchers created a version for home-based learning.
Kid Space is designed to engage children both physically and socially with play-based learning. Oscar, an animated teddy bear, is a multi-modal conversational AI agent who interacts with students as a peer learner. Using context-aware multimodal AI technologies, Oscar serves as an encouraging companion in increasing student engagement in math learning. Kid Space features a multi-modal dialogue system facilitating Oscar’s intelligence (see Figure 1).
Figure 1. Kid Space multimodal dialogue system.
Using off-the-shelf components that can be set up in any learning environment, the interactive Kid Space environment understands the physical space and interactions by utilizing several audio and video hardware devices, including multiple cameras for pose, gesture, and object tracking. RPLIDAR mounted on the wall provides wall touch detection, while overhead microphones detect learners’ voices and a ceiling projector displays a 3D scene on the wall. By capturing the children's verbal and non-verbal interactions during the math game sessions, Kid Space can generate Oscar's responses in real-time. Using multiple algorithms that work together to compute vast amounts of contextual data, the system can understand and interpret multiple modality inputs from the environment, including language, visual, and acoustic cues.
Kid Space is part of Intel Labs' research in Human-AI Collaboration.
Final Study: Students Solving Problems Together
The final Kid Space study explores the system’s potential to support collaborative problem-solving behaviors, joint engagement, and pedagogical interventions. By engaging in collaborative problem solving from an early age, children learn to communicate effectively, listen to others’ ideas, and work together toward a common goal.
To analyze joint engagement and how well the students worked together during the session, the system tracked the students’ joint attention on visual cues like the projected wall, their shared interest levels when collaborating on tasks, and their mutual emotional state during activities.
In addition, an instructional assistant made pedagogical interventions to tailor teaching methods, activities, and materials to suit the unique developmental stages, interests, and abilities of each child. These different types of interventions included asking the students questions, scaffolding techniques such as repeating a question or providing hints, providing positive reinforcement, and promoting social collaboration.
Figure 3. Sample views from the Kid Space learning sessions: (a) flower color selection, (b) flower planting game, and (c) flower watering game.
Conducted with 14 seven-year-old first-grade students, eight girls and six boys with bilingual backgrounds were divided into pairs based on math mastery levels. Based on teacher insights, the participating students had strong social skills, high attention span, a positive attitude toward math with good math skills, and access to diverse technological devices at home for both educational and entertainment purposes. The sessions focused on math lessons and collaborative problem-solving through interactive games such as flower planting and watering.
The results of the study showed significant correlations between collaborative problem-solving behaviors and joint engagement of students working together, and pedagogical interventions provided by the instructional assistant facilitating the learning experience. Joint engagement metrics can serve as indirect measures of CPS, providing valuable insights for educators. Scaffolding and other pedagogical interventions provided by the instructional assistant were positively associated with CPS skills such as establishing shared understanding and representing information.
The results imply that, with necessary design augmentations, Kid Space could potentially play a valuable role in supporting CPS processes in young learners. Oscar could complement human facilitators by monitoring group dynamics and offering real-time interventions to support CPS as well as generating insights to empower educators to incorporate advanced pedagogical interventions.
Past Kid Space Studies Show High Physical and Social Engagement, Low Screen Time
Utilizing both elementary schools and home-based learning environment prototypes, the past user studies conducted by the researchers resulted in several key learnings. Below is a summary of these findings.
Kid Space at home pilots improved learner engagement metrics in the following ways:
- The parents’ concerns about screen time (p<0.05), lack of physical activity (p<0.1), and lack of social interactions (p<0.05) about digital learning environments were all significantly lowered after experiencing Kid Space.
- The children dominantly demonstrated states of on-task (100% of the time) and satisfied (nearly 94% of the time). The children also demonstrated excitement (approximately 5% of the time), a rare emotion to observe in learning context.
- Despite the high levels of engagement, the children’s screen time was as low as 50%. During the other half of the time, the children focused on the physical elements of the experiences such as manipulatives, physical environment, and parents.
- As opposed to traditional PC-based educational games, the children were physically active nearly 92% of the time.
The Kid Space at school pilots had the following impact:
- The children’s scores on a post-test implemented after the sessions were significantly higher than their scores on the pre-test implemented before the sessions (p<0.05).
- The students’ ratings of how they felt after Kid Space sessions were significantly more positive than their ratings before the sessions (p<0.05). More importantly, all 14 students reported that they wanted to play with Oscar again after the sessions.
- The three educators who observed multiple sessions highly rated Oscar as a conversational agent (with an average score of 9.2, on a scale of 1-10). They also rated Kid Space design principles on a scale of 1-5 as follows: Educational (5), social (4.7), magical (5), creative (4.7), physical (5), and renewable (4.7).
- A teacher shared the following during a post-session interview: “[T]he students are answering [but] not just sitting there listening, pointing and touching. They're very involved, active, and they were able to maintain engagement throughout the experience….”
Future Uses for Kid Space
New technologies leveraging AI can provide innovative ways to support and facilitate collaborative problem-solving in various learning environments, including classrooms, after-school programs, and home-based learning. Unlike text-only platforms or traditional gamified approaches, Kid Space creates an immersive learning environment specifically designed for young learners’ developmental needs with responsible AI principles. This combination of physical interaction, real-time adaptation, and collaborative learning positions Kid Space as a distinct contribution to educational technology. With support for natural collaborative behaviors, the program augments students’ academic skills while reducing screen time and increasing physical activity and social interactions. This multi-year research program has provided impactful thought leadership in the field of human AI collaboration in education.