Long-Term Personalization of an In-Home Socially Assistive Robot for Children With Autism Spectrum Disorders
Project Overview
The document explores the application of Socially Assistive Robotics (SAR) in education, particularly for children with Autism Spectrum Disorders (ASD). It emphasizes a personalized learning approach, employing reinforcement learning to customize the feedback and instructional challenges provided by the robot to meet the individual needs of each child. In a study involving 17 children aged 3 to 7 years, the SAR intervention took place over a month in the children's homes and yielded positive results, notably enhancing math learning and increasing user engagement. These findings suggest that generative AI, through its ability to adapt and personalize educational experiences, can significantly benefit children with ASD, providing tailored support that fosters both learning and interaction.
Key Applications
Socially Assistive Robot for personalized learning
Context: In-home educational support for children with Autism Spectrum Disorders (ASD)
Implementation: The SAR system utilized a hierarchical human-robot learning framework (hHRL) with reinforcement learning to adapt instruction and feedback based on individual learning patterns.
Outcomes: Children showed improvements in targeted math skills and long-term retention of intervention content. Families found the system useful and adaptable.
Challenges: Computational challenges in providing effective personalization and ensuring engagement in a real-world, noisy home environment.
Implementation Barriers
Technical Barrier
The challenges of real-world environments introduce noise and unpredictability that complicate the personalization of learning experiences.
Proposed Solutions: Utilization of reinforcement learning to adapt the robot's actions and feedback to the child's performance over time.
Affordability Barrier
Personalized services for children with ASD can be costly and are not universally accessible.
Proposed Solutions: Development of autonomous, personalized SAR systems that can provide scalable support without the need for extensive human intervention.
Project Team
Caitlyn Clabaugh
Researcher
Kartik Mahajan
Researcher
Shomik Jain
Researcher
Roxanna Pakkar
Researcher
David Becerra
Researcher
Zhonghao Shi
Researcher
Eric Deng
Researcher
Rhianna Lee
Researcher
Gisele Ragusa
Researcher
Maja Matarić
Researcher
Contact Information
For information about the paper, please contact the authors.
Authors: Caitlyn Clabaugh, Kartik Mahajan, Shomik Jain, Roxanna Pakkar, David Becerra, Zhonghao Shi, Eric Deng, Rhianna Lee, Gisele Ragusa, Maja Matarić
Source Publication: View Original PaperLink opens in a new window
Project Contact: Dr. Jianhua Yang
LLM Model Version: gpt-4o-mini-2024-07-18
Analysis Provider: Openai