Challenging Futures: Using Chatbots to Reflect on Aging and Dementia

Research Background and Issues

What problems or challenges did the authors identify?

1. Difficulty in reflecting on challenging futures: People generally struggle to concretely think about challenging future scenarios, such as dementia, which may hinder planning and coping.
2. Social stigma of dementia: Dementia patients and their families often face negative attitudes and discrimination, significantly impacting patients' quality of life and social support.
3. Lack of user-centered reflection tools: While virtual reality or augmented reality has potential for fostering empathy, chatbots as flexible and accessible reflection tools remain underexplored.

Why is this issue important?

• Dementia is expected to continue growing globally, projected to affect 152 million people by 2050, necessitating innovative approaches to improve public awareness and acceptance.
• The stigmatization of dementia hinders patients and their families from receiving necessary support.
• Providing tools to help individuals reflect on personal aging and potential challenges can psychologically and culturally promote acceptance and awareness of aging and cognitive decline.

Research Motivation and Related Work

• Introducing chatbots driven by Large Language Models (LLMs) into cognitive reflection scenarios to explore their potential in fostering reflection on challenging futures.
• Existing research has limited exploration of LLMs in dementia care or patient training but has not systematically studied their role in simulating dementia symptoms to enhance public understanding.

Solution

What methods or solutions did the authors propose?

1. Designing a chatbot to simulate dementia symptoms: Developing a chatbot capable of simulating dementia symptoms such as memory loss, repetitive speech, and difficulty understanding.
2. Introducing self vs. other framing: Framing the chatbot as either "future self" or "a stranger" to study how identity differences influence reflection and attitudes.
3. Task-driven dialogue design: Participants are tasked with obtaining a list of materials for baking from the chatbot, a realistic scenario reflecting common challenges faced by dementia patients.

What is innovative about this solution?

• Interactive simulation: For the first time, dynamically simulating dementia symptoms using an LLM-based platform, surpassing traditional educational methods or lengthy textual explanations.
• Personalized framing: Binding dementia symptoms to an individual's future self to stimulate reflection and emotional connection, rather than merely providing general education.
• Mixed-method evaluation: Comprehensive analysis of reflection experiences through both quantitative and qualitative methods, allowing user feedback to directly inform design improvements.

What are the implementation steps? What key technologies were used?

1. Developing two chatbots: One without dementia symptoms as a control and another simulating dementia symptoms (e.g., repetition, confusion) using specific prompts.
2. Experimental design:
   • Participants divided into four groups (2 framing types × 2 chatbot types).
   • The task involves obtaining a list of ten baking ingredients from the chatbot, with the dialogue designed to reflect cognitive challenges faced by dementia patients.
3. Data collection and analysis:
   • Post-task questionnaires include three scales: public stigma of dementia, dementia-related worry, and fear of Alzheimer’s disease.
   • Qualitative data collected through open-ended questions to capture participants' experiences, combined with quantitative analysis to explore results.

Research Findings

What specific findings were obtained?

1. Reflective emotions: When the chatbot was framed as "future self" and simulated dementia symptoms, participants reported deeper emotional connections and reflections on aging.
2. Task frustration: Collaborating with a chatbot simulating dementia symptoms often led to frustration among task-oriented participants, particularly in interactions framed as "strangers."
3. Familiarity and attitudes: Participants who had cared for dementia patients demonstrated greater understanding of dementia symptoms, showing lower rejection and stronger empathy.

What advantages does it have compared to existing solutions?

• Compared to traditional methods (e.g., virtual reality or written education), LLM-driven chatbots offer a conversational, contextualized, and scalable approach that closely resembles real-world interaction experiences.
• Introducing the "future self" framing helps participants directly relate to potential personal experiences.

What were the experimental or evaluation results?

• No significant change in social attitudes: Regardless of whether the chatbot exhibited symptoms or was framed as future self or others, participants' scores on public stigma, worry, or fear of dementia did not significantly change.
• Differences in interaction quality: Participants showed some emotional connection, but many felt disconnected due to the chatbot's inability to meet expected conversational norms.
• Task completion difficulty: Groups interacting with the dementia-simulating chatbot had lower accuracy in completing the ingredient list task, indicating increased challenges in interaction.

Limitations and Future Directions

Limitations:

1. Limited interaction time: The four-minute interaction may not be sufficient to capture deeper reflections.
2. Lack of personalization: The chatbot did not align with participants' personal characteristics (e.g., language and style), affecting authenticity.
3. Insufficient psychological guidance: Participants were unable to fully shift from task-oriented interaction to deeper self-exploration.

Future Directions:

1. Enhancing interaction personalization: Tailoring the experience to individual user data and characteristics for a more authentic experience.
2. Extending experiment duration: Designing longer-term interactions to provide broader cognitive and emotional feedback.
3. Strengthening educational content: Providing more direct dementia education and cognitive content without compromising sensitivity.

This study offers valuable perspectives and methodological frameworks for utilizing chatbots to promote public reflection on challenging future scenarios, such as dementia.