LLM-Powered AI Tutors with Personas for d/Deaf and Hard-of-Hearing Online Learners (2411.09873v1)

Abstract: Intelligent tutoring systems (ITS) using AI technology have shown promise in supporting learners with diverse abilities; however, they often fail to meet the specific communication needs and cultural nuances needed by d/Deaf and Hard-of-Hearing (DHH) learners. As LLMs provide new opportunities to incorporate personas to AI-based tutors and support dynamic interactive dialogue, this paper explores how DHH learners perceive LLM-powered ITS with different personas and identified design suggestions for improving the interaction. We developed an interface that allows DHH learners to interact with ChatGPT and three LLM-powered AI tutors with different experiences in DHH education while the learners watch an educational video. A user study with 16 DHH participants showed that they perceived conversations with the AI tutors who had DHH education experiences to be more human-like and trustworthy due to the tutors' cultural knowledge of DHH communities. Participants also suggested providing more transparency regarding the tutors' background information to clarify each AI tutor's position within the DHH community. We discuss design implications for more inclusive LLM-based systems, such as supports for the multimodality of sign language.

• The paper explores how LLM-powered AI tutors using distinct personas can enhance the educational experience for d/Deaf and Hard-of-Hearing (DHH) online learners.
• Key findings indicate that DHH learners perceived tutors with personas claiming DHH educational experience as more trustworthy and engaging, facilitating deeper interaction.
• Design implications highlight the need for transparency in AI tutor backgrounds, careful consideration of potential biases in personas, and essential integration of sign language support for DHH users.

Insights into LLM-Powered AI Tutors for DHH Learners

The paper presented in the paper "LLM-Powered AI Tutors with Personas for d/Deaf and Hard-of-Hearing Online Learners" explores the role of LLMs in enhancing the educational experience for d/Deaf and Hard-of-Hearing (DHH) learners. Employing the capabilities of LLMs, particularly through the lens of ChatGPT-powered tutors, this research investigates how different AI personas can cater to the unique communication requirements and cultural considerations pertinent to DHH learners. The focal point of this investigation is the application of Intelligent Tutoring Systems (ITS) molded by LLMs to create personas with varying degrees of relevance to DHH education.

Methodology

The authors developed a prototype interface involving ChatGPT and three LLM-driven AI tutors—each projecting a distinct persona—ranging from no DHH educational experience to being explicitly embedded within a DHH-centric educational environment. The paper engaged 16 DHH participants, allowing them to interact with these four tutors while they accessed educational video content on Augmented Reality (AR), thereby providing a context for generating dialogue and probing the tutors' responses.

Key Findings

1. Perception and Trust: Participants perceived tutors with personas claiming experience in DHH education as more trustworthy and human-like in comparison to a generalized ChatGPT tutor. This perception underscores the importance of personas that resonate with the cultural and educational background of the learner group, enhancing both trust and engagement.
2. Feedback and Interaction: The AI tutors with contextual personas, such as those symbolizing roles from DHH educational backgrounds, facilitated a deeper conversational engagement by prompting the learners with relevant questions. This form of interaction was reported as more engaging and educationally valuable by the participants.
3. Design Suggestions: Participants identified several areas for improvement, particularly the need for greater transparency regarding the tutors' backgrounds and abilities. Furthermore, they emphasized the importance of sign language support in AI interfaces to fully accommodate the needs of DHH learners.

Implications for Design and Future Research

The findings from this paper highlight several critical directions for the future development of AI tutors in educational settings for DHH learners:

• Enhanced Credibility and Transparency: There's a necessity to ensure that AI tutors provide clear background information and demonstrate credible integration within the learners' cultural contexts. This involves designing adaptive and informative personas that can genuinely represent the background of the tutors.
• Bias and Representation: The paper reveals potential biases linked with personas that relay specific educational or cultural experiences. There's a need to carefully navigate these biases to prevent skewed perceptions or expectations from the users regarding the LLM-powered personas' capabilities.
• Sign Language Integration: The paper underlines the imperative need for LLM systems to support sign language, both as an input and output modality, to enhance accessibility for DHH learners. This aspect remains a critical area for technological development and research focus.

Conclusion

In conclusion, the incorporation of LLMs with specifically designed personas appears to be a promising avenue for advancing AI-based educational tools for DHH learners. The paper provides a nuanced understanding of how cultural and educational alignment through AI personas can significantly enhance learner engagement and trust. Future research endeavors should prioritize technological innovations that ensure comprehensive inclusivity and accessibility, particularly focusing on multifaceted language support functionalities. The paper’s insights into designing empathetic and contextually relevant AI systems offer valuable guidelines for educators and AI developers committed to fostering inclusive learning environments.