A Qualitative Investigation to Design Empathetic Agents as Conversation Partners for People with Autism Spectrum Disorder (2407.20637v1)

Abstract: Autism Spectrum Disorder (ASD) can profoundly affect reciprocal social communication, resulting in substantial and challenging impairments. One aspect is that for people with ASD conversations in everyday life are challenging due to difficulties in understanding social cues, interpreting emotions, and maintaining social verbal exchanges. To address these challenges and enhance social skills, we propose the development of a learning game centered around social interaction and conversation, featuring Artificial Intelligence agents. Our initial step involves seven expert interviews to gain insight into the requirements for empathetic and conversational agents in the field of improving social skills for people with ASD in a gamified environment. We have identified two distinct use cases: (1) Conversation partners to discuss real-life issues and (2) Training partners to experience various scenarios to improve social skills. In the latter case, users will receive quests for interacting with the agent. Additionally, the agent can assign quests to the user, prompting specific conversations in real life and providing rewards for successful completion of quests.

Designing Empathetic Agents for Autism Spectrum Disorder

The paper "A Qualitative Investigation to Design Empathetic Agents as Conversation Partners for People with Autism Spectrum Disorder" presents a comprehensive investigation into the development of AI-driven conversational agents aimed at enhancing social skills for individuals with Autism Spectrum Disorder (ASD). The paper emphasizes the need for empathetic interactions in a controlled environment, leveraging game elements to foster motivation and engagement.

Introduction and Motivation

ASD is characterized by significant difficulties in social interactions, including impaired social perception, nonverbal communication, and rigid behavioral patterns. These challenges, often referred to as 'social blindness,' create substantial barriers for individuals with ASD in everyday settings. The paper posits that with appropriate interventions such as social skills training and assistive technologies, individuals with ASD can improve their conversational abilities.

Given the increasing prevalence of ASD and the limitations of traditional therapeutic approaches, the research explores the utility of LLMs integrated with visual and interactive elements for simulating human-like interactions. The aim is to provide a safe and controlled environment where adults with ASD can practice and hone their social skills.

Research Methodology

The paper undertook a qualitative investigation involving seven expert interviews to gather requirements for the design of empathetic agents. The experts included therapists, doctors, clinical psychologists, and individuals with ASD. The investigation centered around four key research questions: the characteristics of empathetic agents, their visualization, interaction modalities, and evaluation of their effectiveness.

Questionnaire Design and Data Collection

The researchers developed a structured interview protocol with open-ended questions covering various aspects such as the design, communication, visualization, feedback mechanisms, and evaluation of the empathetic agents. The interviews, conducted both in-person and online, were transcribed and systematically analyzed to identify recurring themes and insights.

Key Findings and Implications

Design Aspects and Communication

The experts proposed two primary use cases for the empathetic agents:

1. Conversational Partners: These agents act as supportive friends, providing constructive feedback and facilitating social interactions.
2. Training Partners: These agents simulate real-life scenarios, employing ambiguous communication techniques such as metaphors and irony to enhance the user’s social skills.

The agents should ideally incorporate both verbal and non-verbal communication, including sensory capabilities to interpret gestures and facial expressions. The minimum viable product, however, should include a chat-based system.

Visualization

The visualization of the agents remains a contested topic. While some experts advocate for realistic avatars to convey emotions and nonverbal cues effectively, others suggest a stylized approach to avoid the Uncanny Valley effect, where overly realistic avatars may elicit discomfort. This indicates the need for further research on the optimal visualization strategy.

Feedback and Adaptation

Feedback mechanisms are critical for the effectiveness of the agents. The agents should provide real-time feedback, clarify their statements, and gradually increase the difficulty of interaction scenarios. The inclusion of game elements, such as quests assigned by the agent, is recommended to maintain user engagement and provide tangible rewards for successful real-life interactions.

Evaluation Metrics

The effectiveness of the empathetic agents should be evaluated using established scales such as the Bot Usability Scale (BUS), System Usability Scale (SUS), and the Social Responsiveness Scale (SRS). These metrics will help assess both user satisfaction and the transfer of learned skills to real-life scenarios.

Future Directions

The findings underscore the potential of integrating LLMs with interactive and visual elements to create effective training tools for social skills in individuals with ASD. Future research should focus on refining the visualization of agents, optimizing feedback mechanisms, and systematically evaluating long-term learning outcomes. Additionally, the proactive assignment of real-life quests by the agents presents an innovative approach to bridging the gap between simulated and real-world interactions.

Overall, this paper contributes significantly to the development of assistive technologies for ASD, providing a foundation for further research and practical applications in enhancing social competence through AI-driven empathetic agents.