Skip to main content Skip to navigation

CARING-AI: Towards Authoring Context-aware Augmented Reality INstruction through Generative Artificial Intelligence

Project Overview

The document explores the application of Generative Artificial Intelligence (Gen-AI) in education, particularly through the CARING-AI system, which facilitates the creation of context-aware augmented reality (AR) instructions. By enabling the generation of humanoid avatar animations tailored to users' physical environments, CARING-AI addresses the limitations of existing AR instructional methods. The document outlines three main application scenarios: asynchronous instruction creation, ad hoc instruction creation, and remote instruction creation, all aimed at enhancing learning experiences. User studies indicate that CARING-AI significantly improves usability and effectiveness compared to traditional methods, leading to increased engagement and streamlined instructional processes. However, the document also notes challenges related to the effective implementation of these tools and the need for user adaptation. Overall, the findings suggest that generative AI has the potential to transform educational practices by offering more personalized and context-sensitive learning experiences.

Key Applications

CARING-AI for Augmented Reality Instruction

Context: Educational environments including research labs and virtual learning settings, targeting educators, researchers, and technicians. Users create AR instructions by capturing relevant contextual information from their surroundings, enhancing the educational experience with personalized and context-aware content.

Implementation: Users generate textual instructions using a speech interface while scanning their environment to capture context. The CARING-AI system then utilizes this data to create animations of humanoid avatars that demonstrate the instructions, making the content interactive and engaging.

Outcomes: Improved usability and effectiveness in creating AR instructions, with users reporting enhanced learning experiences, increased engagement, and effective communication of complex tasks. The personalized approach leads to better understanding and retention of information.

Challenges: Users may face adaptation challenges when using new technologies, and varying contexts can lead to potential confusion. Additionally, there are limitations in accurately rendering complex hand-object interactions and a need for further development in the underlying generative AI algorithms.

Implementation Barriers

Technical Limitations

Current Gen-AI lacks contextual awareness and fails to fully adapt to varying real-world scenarios. Additionally, variability in physical environments may lead to inconsistencies in the instructional content generated by the AI.

Proposed Solutions: Enhanced algorithms to incorporate contextual information, improve the quality of AI-generated content, and implement robust context recognition capabilities that allow users to customize instructions based on specific environments.

Hardware Limitations

Dependence on complex hardware setups for motion capture and AR interaction can restrict usability.

Proposed Solutions: Developing software that minimizes hardware requirements and allows for remote instruction authoring.

Technological Barrier

Users may struggle with adapting to new AR technologies and the complexity of creating contextualized instructions.

Proposed Solutions: Providing training sessions and resources to familiarize users with the technology, as well as simplifying the interface for easier interaction.

Project Team

Jingyu Shi

Researcher

Rahul Jain

Researcher

Seungguen Chi

Researcher

Hyungjun Doh

Researcher

Hyunggun Chi

Researcher

Alexander J. Quinn

Researcher

Karthik Ramani

Researcher

Contact Information

For information about the paper, please contact the authors.

Authors: Jingyu Shi, Rahul Jain, Seungguen Chi, Hyungjun Doh, Hyunggun Chi, Alexander J. Quinn, Karthik Ramani

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

Let us know you agree to cookies