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

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.

• The paper presents a qualitative investigation where expert interviews shaped guidelines for designing empathetic agents to improve social communication in ASD.
• It demonstrates a gamified learning approach that integrates verbal and non-verbal cues, personalization, and adaptive challenges to engage users.
• Findings emphasize that contextual feedback, dynamic visualization, and real-world quest assignments are crucial for effective agent performance.

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

Introduction

This paper explores the role of empathetic AI agents in assisting individuals with Autism Spectrum Disorder (ASD) to improve their social communication skills. The focus is on designing agents that can participate in gamified learning environments to provide both an interactive friend and a training partner. ASD is associated with challenges in understanding social cues, which can result in significant communication impairments. To tackle these, the paper proposes deploying empathetic agents in a game-based framework to enhance users' social abilities by simulating real-life interactions and assigning practical quests.

Figure 1: Proposed framework for initiating conversations with an agent, incorporating a starting quest.

Methodology

Utilizing a qualitative approach, the study conducted interviews with seven experts possessing diverse backgrounds in autism spectrum disorders, including therapists and clinical psychologists. This initial step is intended to gather insights on the requirements for designing effective empathetic agents.

Questionnaire Development: A structured interview protocol was crafted with open-ended questions targeting various aspects of agent design, such as their characteristics, visualization, interaction, and evaluation.

Expert Recruitment: Experts from medical and autism centers were recruited, ensuring a mix of theoretical and practical perspectives.

Data Analysis: Thematic analysis was employed, involving transcription and coding to identify patterns and themes. This process provided a rich understanding of the expert insights, informing agent development.

Results

The study identified critical aspects of empathetic agent design, covering several dimensions:

Design Aspects: Experts suggested two main roles for the agents: as supportive friends sharing common interests, and as training partners for practicing complex scenarios. Users would benefit from personalizing the agent’s character to enhance relatability and engagement.

Communication: A highly interactive system is essential, with verbal and non-verbal elements ensuring smooth human-like interactions. Sensory capabilities such as eye and facial detection are preferred to facilitate realistic communication.

Visualization: There is a debate between stylized and realistic visualization, with concerns over the Uncanny Valley effect influencing the design decisions.

Feedback and Adaptation: Adaptation involves providing contextual feedback and real-world quest assignments. Regular assessments on users' performance are necessary to adjust the agent's difficulty and maintain engagement.

Evaluation: Usability and success metrics include the Bot Usability Scale, Social Responsiveness Scale, and potential analysis of the transferability of learned skills to real-world interactions.

Discussion

This research underlines the potential of LLMs and virtual agents in creating immersive, supportive environments for ASD individuals to hone their social skills. By framing the agent as both a conversational partner and a proactive trainer, users are provided with motivation-driven activities that challenge their social deficits and offer tangible incentives through task completion.

Furthermore, the game-based approach fosters a sense of achievement and progress, pivotal in maintaining motivation over extended periods. Continued exploration is needed, particularly in the visualization domain, to ascertain the optimal balance between avatar realism and user comfort.

Conclusion

Ultimately, the insights garnered from the qualitative investigation lay a robust foundation for developing empathetic agents that address the nuances of social communication in people with ASD. By leveraging interactive technologies and adaptive learning systems, the study highlights the capacity to revolutionize support mechanisms for socially empowering individuals with ASD. Future work might involve quantitative studies to validate the effectiveness of these agents and further refine their design based on empirical results.